ABP1 CUDA applications


Advanced search

Message boards : Cruncher's Corner : ABP1 CUDA applications

AuthorMessage
Profile Bernd Machenschalk
Forum moderator
Project administrator
Project developer
Avatar
Send message
Joined: Oct 15 04
Posts: 3240
Credit: 90,353,666
RAC: 16,138
Message 100691 - Posted 26 Nov 2009 12:03:06 UTC

    Last modified: 26 Nov 2009 12:20:31 UTC

    We have finally begun to automatically deliver CUDA work & applications (plan class "ABP1cuda23") to machines that satisfy the following requirements:

    - enabled NVIDIA GPU work in Einstein@home preferences
    - NVidia GPU with at least 450MB of free memory
    - Display Driver version 190.38 (&up), i.e. CUDA 2.3 capability
    - BOINC Core Client version 6.10 (&up)

    CUDA Beta App testers should drain their work cache and switch back to the normal project work.

    BM

    Richard Haselgrove
    Send message
    Joined: Dec 10 05
    Posts: 1322
    Credit: 29,148,497
    RAC: 10,708
    Message 100694 - Posted 26 Nov 2009 12:07:06 UTC

      Are these the same 'hybrid' applications, requiring a full CPU core in support, that we were testing in Beta, or have you been able to tansfer more of the other calculations (apart from fft) onto the GPU?

      Profile Bernd Machenschalk
      Forum moderator
      Project administrator
      Project developer
      Avatar
      Send message
      Joined: Oct 15 04
      Posts: 3240
      Credit: 90,353,666
      RAC: 16,138
      Message 100695 - Posted 26 Nov 2009 12:13:31 UTC - in response to Message 100694.

        Are these the same 'hybrid' applications, requiring a full CPU core in support, that we were testing in Beta, or have you been able to tansfer more of the other calculations (apart from fft) onto the GPU?

        These are basically the same Apps as have been in Beta est. They still require a full CPU core.

        BM

        cristipurdel
        Send message
        Joined: Jul 19 07
        Posts: 17
        Credit: 85,452
        RAC: 0
        Message 100697 - Posted 26 Nov 2009 13:01:55 UTC - in response to Message 100695.

          Are these the same 'hybrid' applications, requiring a full CPU core in support, that we were testing in Beta, or have you been able to tansfer more of the other calculations (apart from fft) onto the GPU?

          These are basically the same Apps as have been in Beta est. They still require a full CPU core.

          BM

          They require a core at 100% or we can find a lower value so that other projects can use it?
          I'm running it with 0.3 CPU + 1.0 GPU.

          Profile Svenie25
          Send message
          Joined: Mar 21 05
          Posts: 139
          Credit: 2,436,862
          RAC: 0
          Message 100703 - Posted 26 Nov 2009 14:33:36 UTC

            The new apps are not shown on the apps-page: http://einstein.phys.uwm.edu/apps.php

            Now it would be nice, if you could select, that your CPU only gets tasks for S5R6. Because the ABP1-work is done at the GPU.
            ____________

            [B^S] Elphidieus
            Send message
            Joined: Feb 20 05
            Posts: 220
            Credit: 12,347,408
            RAC: 0
            Message 100704 - Posted 26 Nov 2009 14:52:37 UTC - in response to Message 100691.

              We have finally begun to automatically deliver CUDA work & applications (plan class "ABP1cuda23") to machines that satisfy the following requirements:

              - enabled NVIDIA GPU work in Einstein@home preferences
              - NVidia GPU with at least 450MB of free memory
              - Display Driver version 190.38 (&up), i.e. CUDA 2.3 capability
              - BOINC Core Client version 6.10 (&up)

              CUDA Beta App testers should drain their work cache and switch back to the normal project work.

              BM


              And did you forget to include a fifth piece of requirement - Not meant for Macs...!!!
              ____________

              Profile Bernd Machenschalk
              Forum moderator
              Project administrator
              Project developer
              Avatar
              Send message
              Joined: Oct 15 04
              Posts: 3240
              Credit: 90,353,666
              RAC: 16,138
              Message 100706 - Posted 26 Nov 2009 15:55:20 UTC - in response to Message 100704.

                And did you forget to include a fifth piece of requirement - Not meant for Macs...!!!

                Yep. Stupid as it is, NVidia hasn't yet delivered their promised 64Bit CUDA libraries for Mac OS X. As soon as they do, we could build and send out CUDA Apps for Macs, too.

                BM

                Profile Gundolf Jahn
                Send message
                Joined: Mar 1 05
                Posts: 1079
                Credit: 328,104
                RAC: 18
                Message 100709 - Posted 26 Nov 2009 16:20:51 UTC - in response to Message 100703.

                  Because the ABP1-work is done at the GPU.

                  It's not, at least not exclusively.

                  Grutte Pier [Wa Oars]~GP500
                  Avatar
                  Send message
                  Joined: May 18 09
                  Posts: 39
                  Credit: 2,462,785
                  RAC: 2,761
                  Message 100710 - Posted 26 Nov 2009 17:18:48 UTC - in response to Message 100697.

                    Last modified: 26 Nov 2009 17:22:10 UTC

                    Are these the same 'hybrid' applications, requiring a full CPU core in support, that we were testing in Beta, or have you been able to tansfer more of the other calculations (apart from fft) onto the GPU?

                    These are basically the same Apps as have been in Beta est. They still require a full CPU core.

                    BM

                    They require a core at 100% or we can find a lower value so that other projects can use it?
                    I'm running it with 0.3 CPU + 1.0 GPU.


                    I find that very bad that it needs a 100% core and that you set cuda on Enable as default.

                    It's a problem with that on a few pc.
                    I run Einstein on a pc and folding@home on the GPU.

                    It was only by accident that i discovered that cuda was enabled today.

                    Bedrich Hajek
                    Send message
                    Joined: Dec 9 05
                    Posts: 1
                    Credit: 6,977,924
                    RAC: 2,034
                    Message 100713 - Posted 26 Nov 2009 19:50:52 UTC - in response to Message 100691.

                      We have finally begun to automatically deliver CUDA work & applications (plan class "ABP1cuda23") to machines that satisfy the following requirements:

                      - enabled NVIDIA GPU work in Einstein@home preferences
                      - NVidia GPU with at least 450MB of free memory
                      - Display Driver version 190.38 (&up), i.e. CUDA 2.3 capability
                      - BOINC Core Client version 6.10 (&up)

                      CUDA Beta App testers should drain their work cache and switch back to the normal project work.

                      BM


                      The only problem I have with this, is, it is using 100% CPU, while using only about 4% of the GPU from a GTX 285 card.
                      ____________

                      Profile XJR-Maniac
                      Avatar
                      Send message
                      Joined: Dec 8 05
                      Posts: 3
                      Credit: 670,148
                      RAC: 0
                      Message 100714 - Posted 26 Nov 2009 19:57:17 UTC

                        This was my first AND my last ABP1 so called "CUDA" WU!

                        This application is a complete nonsense! It's occupying 1 CPU-Core AND a GPU for more than six hours now and most of the time the GPU is nearly IDLE, temperature is at 49°C now! In this amount of time the GPU could have crunched a hundred of Milky Way WUs or dozens of Seti or Collatz WUs.

                        Why doesn't it suspend after a given time? Are there no checkpoints in CUDA apps or what? Others are way too short to suspend, so there's not much to compare for me.

                        If that's what you call CUDA then cancel it and revert back to CPU only WUs. This is nothing more than a bad joke. Sorry but I didn't contribute in beta test phase, otherwise I would have mentioned this much earlier.
                        ____________

                        pelpolaris
                        Send message
                        Joined: Nov 5 06
                        Posts: 2
                        Credit: 4,055,251
                        RAC: 0
                        Message 100715 - Posted 26 Nov 2009 22:07:49 UTC - in response to Message 100691.

                          [quote]We have finally begun to automatically deliver CUDA work & applications (plan class "ABP1cuda23") to machines that satisfy the following requirements:

                          - enabled NVIDIA GPU work in Einstein@home preferences
                          - NVidia GPU with at least 450MB of free memory
                          - Display Driver version 190.38 (&up), i.e. CUDA 2.3 capability
                          - BOINC Core Client version 6.10 (&up)

                          Do you have any idea of what kind of "Linows" or "Windux" platforms that can accept to run the Einstein CUDA 23 ?


                          ____________

                          ML1
                          Send message
                          Joined: Feb 20 05
                          Posts: 320
                          Credit: 17,155,851
                          RAC: 10,895
                          Message 100716 - Posted 26 Nov 2009 22:11:07 UTC - in response to Message 100715.

                            Last modified: 26 Nov 2009 22:11:25 UTC

                            Do you have any idea of what kind of "Linows" or "Windux" platforms that can accept to run the Einstein CUDA 23 ?

                            Not sure what you mean by that, but their standard app is working fine on this Mandriva 2010.0 Linux system.

                            The GPU temps suggest that there is not much GPU utilisation, but then again this is their first attempt.

                            Happy crunchin',
                            Martin
                            ____________
                            Powered by Mandriva Linux A user friendly OS!
                            See the Boinc HELP Wiki

                            Profile XJR-Maniac
                            Avatar
                            Send message
                            Joined: Dec 8 05
                            Posts: 3
                            Credit: 670,148
                            RAC: 0
                            Message 100718 - Posted 26 Nov 2009 23:08:00 UTC

                              Now got my first (and probably last) ABP1 3.13 "CUDA" WU finished in over 10 hours time on a Q9650/GeForce GTX260 were CPU time was 8.3 hours and GPU time around 2 hours. This means more than 8 hours of wasted GPU time! Does that make any sense?

                              My last two ABP1 3.12 CPU only WUs took less than 5 hours on a Q9650.

                              Are the new WUs more complex or longer than the old ones or is this just another bad joke?
                              ____________

                              Profile Gary Roberts
                              Forum moderator
                              Send message
                              Joined: Feb 9 05
                              Posts: 3003
                              Credit: 1,008,163,047
                              RAC: 2,457,858
                              Message 100720 - Posted 27 Nov 2009 3:25:25 UTC - in response to Message 100718.

                                Now got my first (and probably last) ABP1 3.13 "CUDA" WU finished in over 10 hours time on a Q9650/GeForce GTX260 were CPU time was 8.3 hours and GPU time around 2 hours. This means more than 8 hours of wasted GPU time! Does that make any sense?

                                You don't have any CUDA tasks on your Q9650. The list of tasks for that machine shows 3 completed ABP1 tasks, all of which took around 17k secs and none of which used a GPU for crunching. I decided to look at your other machines and I found the GPU crunched ABP1 task on your Pentium D. There are no other ABP1 tasks still showing on that machine so there's no ability to do a comparison. Here is the list of tasks for your pentium D with the GPU crunched ABP1 task at the top. It's interesting to note that it took much the same time to crunch the ABP1 task (250 credits) as the two previous GW tasks (136 credits) - nearly double the credits for a tiny bit more crunch time.

                                What is even more interesting is the apparent dramatic slowdown after Nov 20. The three earlier GW tasks took around 11K secs each while the two after this date took 27K and 29k secs respectively. Now there is variability in the GW crunch times but there is usually a variation in credits to compensate - at least partially. Since all GW tasks were awarded the same credit, it's unusual to see such a huge variation in crunch time. Can you think of anything that might have happened to your machine after Nov 20? Something drastic like halving the CPU frequency might do it :-).

                                My last two ABP1 3.12 CPU only WUs took less than 5 hours on a Q9650.

                                It's not really fair to compare a Pentium D to a Q9650 :-).

                                Are the new WUs more complex or longer than the old ones or is this just another bad joke?

                                There aren't any new tasks - just the same old tasks being crunched with a new program which is (performance-wise) much the same as the beta test app it replaces.

                                It might help if you realise that just because one project can make hugely efficient use of a GPUs parallelism, other projects may struggle to do anything like the same even after considerable effort has been expended. You might take that into account when firing off your criticism.

                                ____________
                                Cheers,
                                Gary.

                                cyborg888
                                Send message
                                Joined: Jul 12 07
                                Posts: 1
                                Credit: 850,061
                                RAC: 0
                                Message 100723 - Posted 27 Nov 2009 5:27:33 UTC

                                  Hello
                                  I was happy to look APP Cuda optimisation in my boinc ...
                                  (config : Q9950 + 8800gt)

                                  But there is maybe a problem.

                                  My GPU iddle temperature is 48-49 deg ...
                                  And during cruching process Gpu temp is same : 48-49 deg

                                  So My gpu doesn't seem to be used.

                                  Best Regards.

                                  Profile Bikeman (Heinz-Bernd Eggenstein)
                                  Forum moderator
                                  Project administrator
                                  Project developer
                                  Avatar
                                  Send message
                                  Joined: Aug 28 06
                                  Posts: 3225
                                  Credit: 72,411,621
                                  RAC: 35,282
                                  Message 100724 - Posted 27 Nov 2009 5:57:05 UTC - in response to Message 100723.

                                    Last modified: 27 Nov 2009 5:59:30 UTC


                                    So My gpu doesn't seem to be used.

                                    Best Regards.

                                    It's normal that the GPU temp will not rise drastically when running this version of the CUDA app, the same was seen during the beta tests. This is because the app still makes heavy use of the CPU for certain parts of the computatations.

                                    This doesn't exclude a speedup of the app by using the GPU:

                                    Let's assume (it's just a simple example) that an app does computations consisting of two parts, A and B, where A has to be executed before B can start. Let's assume that on a CPU, A and B each take 50% of the runtime.

                                    Now assume that only part B can easily be ported to GPU-code, resulting in a (say) 25 fold speedup for this part B. A still has to be done on the CPU.

                                    The result: If the total runtime was 1000 sec before, it is now 520 s, almost doubling the performance. Only 20 seconds of these 520 seconds will be spent in the GPU, or below 4 %. So even small load factors on the GPU can result in reasonable speedups.
                                    Yes, it would be nicer if parts A and B in our example could be done on the GPU for a total speedup of (say) 25-fold performance, but that might be not so easy. Some types of computation lend themselves more easily to parallelization than others.

                                    CU
                                    Bikeman
                                    ____________

                                    Profile BigDaddyDave
                                    Send message
                                    Joined: Sep 6 09
                                    Posts: 15
                                    Credit: 8,968,100
                                    RAC: 6,158
                                    Message 100728 - Posted 27 Nov 2009 7:22:25 UTC

                                      Hi all,

                                      Question for you, if these are CUDA WU, why is my CPU crunching them and my GPU is just sitting idle?


                                      11/26/2009 11:18:21 PM Starting BOINC client version 6.10.18 for windows_intelx86
                                      11/26/2009 11:18:21 PM log flags: file_xfer, sched_ops, task
                                      11/26/2009 11:18:21 PM Libraries: libcurl/7.19.4 OpenSSL/0.9.8l zlib/1.2.3
                                      11/26/2009 11:18:21 PM Data directory: C:\Documents and Settings\All Users\Application Data\BOINC
                                      11/26/2009 11:18:21 PM Running under account User
                                      11/26/2009 11:18:21 PM Processor: 2 GenuineIntel Intel(R) Pentium(R) 4 CPU 3.20GHz [x86 Family 15 Model 4 Stepping 1]
                                      11/26/2009 11:18:21 PM Processor: 1.00 MB cache
                                      11/26/2009 11:18:21 PM Processor features: fpu tsc sse sse2 mmx
                                      11/26/2009 11:18:21 PM OS: Microsoft Windows XP: Professional x86 Edition, Service Pack 3, (05.01.2600.00)
                                      11/26/2009 11:18:21 PM Memory: 1.50 GB physical, 2.85 GB virtual
                                      11/26/2009 11:18:21 PM Disk: 179.31 GB total, 95.15 GB free
                                      11/26/2009 11:18:21 PM Local time is UTC -8 hours
                                      11/26/2009 11:18:21 PM NVIDIA GPU 0: GeForce GTX 260 (driver version 19107, CUDA version 2030, compute capability 1.3, 896MB, 675 GFLOPS peak)
                                      11/26/2009 11:18:21 PM Not using a proxy
                                      11/26/2009 11:18:21 PM Einstein@Home URL http://einstein.phys.uwm.edu/; Computer ID 2063582; resource share 100
                                      11/26/2009 11:18:21 PM SETI@home URL http://setiathome.berkeley.edu/; Computer ID 2378854; resource share 100
                                      11/26/2009 11:18:21 PM SETI@home General prefs: from SETI@home (last modified 04-Mar-2009 23:03:00)
                                      11/26/2009 11:18:21 PM SETI@home Computer location: home
                                      11/26/2009 11:18:21 PM SETI@home General prefs: no separate prefs for home; using your defaults
                                      11/26/2009 11:18:21 PM Preferences limit memory usage when active to 767.36MB
                                      11/26/2009 11:18:21 PM Preferences limit memory usage when idle to 1381.25MB
                                      11/26/2009 11:18:21 PM Preferences limit disk usage to 89.65GB
                                      11/26/2009 11:18:22 PM Einstein@Home Restarting task p2030_54471_60586_0034_G46.39-00.47.S_1.dm_499_1 using einsteinbinary_ABP1 version 313
                                      11/26/2009 11:18:22 PM Einstein@Home Restarting task h1_1085.60_S5R4__1050_S5R6a_1 using einstein_S5R6 version 301


                                      Thanks!


                                      BDDave
                                      ____________

                                      cristipurdel
                                      Send message
                                      Joined: Jul 19 07
                                      Posts: 17
                                      Credit: 85,452
                                      RAC: 0
                                      Message 100729 - Posted 27 Nov 2009 7:38:33 UTC - in response to Message 100724.


                                        So My gpu doesn't seem to be used.

                                        Best Regards.

                                        It's normal that the GPU temp will not rise drastically when running this version of the CUDA app, the same was seen during the beta tests. This is because the app still makes heavy use of the CPU for certain parts of the computatations.

                                        This doesn't exclude a speedup of the app by using the GPU:

                                        Let's assume (it's just a simple example) that an app does computations consisting of two parts, A and B, where A has to be executed before B can start. Let's assume that on a CPU, A and B each take 50% of the runtime.

                                        Now assume that only part B can easily be ported to GPU-code, resulting in a (say) 25 fold speedup for this part B. A still has to be done on the CPU.

                                        The result: If the total runtime was 1000 sec before, it is now 520 s, almost doubling the performance. Only 20 seconds of these 520 seconds will be spent in the GPU, or below 4 %. So even small load factors on the GPU can result in reasonable speedups.
                                        Yes, it would be nicer if parts A and B in our example could be done on the GPU for a total speedup of (say) 25-fold performance, but that might be not so easy. Some types of computation lend themselves more easily to parallelization than others.

                                        CU
                                        Bikeman

                                        Are this the actual numbers for ABP1cuda23 ?
                                        If not, could you post the actual speedup on your machine between the GPU and CPU version?

                                        Oliver Bock
                                        Forum moderator
                                        Project administrator
                                        Project developer
                                        Send message
                                        Joined: Sep 4 07
                                        Posts: 359
                                        Credit: 20,915,781
                                        RAC: 1,674
                                        Message 100733 - Posted 27 Nov 2009 10:09:00 UTC - in response to Message 100710.

                                          Last modified: 27 Nov 2009 10:14:40 UTC


                                          I find that very bad that it needs a 100% core and that you set cuda on Enable as default.


                                          I understand and we tried not to enable the CUDA app by default. Unfortunately that would have involved a change in the BOINC core client code which is not under our direct control. Please note that this is one of the reasons why we set quite a few minimum requirements. This way the number of volunteers who receive CUDA work is as limited as possible.

                                          WRT the efficiency of the current implementation: we are working on a number of improvements. The CPU part of the radio pulsar search received quite a few changes that will not only benefit the CPU-only application but also the CUDA version, thereby moving the computational ratio towards the GPU. These changes will be released as a new application called "ABP2" - probably in the next 1-2 weeks. In parallel to that we are currently working hard to move the remaining CPU part of the CUDA version more or less completely to the GPU.

                                          Please note that even today the CUDA app wouldn't actually require a full CPU. However, as soon as you tell the client you use less than 100% it doesn't renice the process (reduce it's priority) anymore. From our point of view it's better to have the process claiming one CPU at the lowest priority than using, say, 60% at normal priority.


                                          Hope this gives a small insight...

                                          Oliver

                                          Oliver Bock
                                          Forum moderator
                                          Project administrator
                                          Project developer
                                          Send message
                                          Joined: Sep 4 07
                                          Posts: 359
                                          Credit: 20,915,781
                                          RAC: 1,674
                                          Message 100734 - Posted 27 Nov 2009 10:20:26 UTC - in response to Message 100729.

                                            Last modified: 27 Nov 2009 10:26:39 UTC


                                            Are this the actual numbers for ABP1cuda23 ?
                                            If not, could you post the actual speedup on your machine between the GPU and CPU version?


                                            During our final tests last week we observed up to 100% speedup for the Windows version and about 20% for the Linux version. The reason for this rather huge difference is that the CPU parts (as for the CPU-only version) are still faster on Linux than on Windows. As always, your mileage may very depending on the GPU and CPU (Intel vs. AMD) used.

                                            Cheers,
                                            Oliver

                                            cristipurdel
                                            Send message
                                            Joined: Jul 19 07
                                            Posts: 17
                                            Credit: 85,452
                                            RAC: 0
                                            Message 100735 - Posted 27 Nov 2009 10:43:59 UTC - in response to Message 100733.

                                              Last modified: 27 Nov 2009 10:47:28 UTC


                                              Please note that even today the CUDA app wouldn't actually require a full CPU. However, as soon as you tell the client you use less than 100% it doesn't renice the process (reduce it's priority) anymore. From our point of view it's better to have the process claiming one CPU at the lowest priority than using, say, 60% at normal priority.


                                              Thanks for the info, very useful.

                                              From my point of view, while I'm running boinc on my xeon + nvidia, I can run only 3 other projects + 1 einstein gpu. Instead of sending tasks which use 1.00 CPU + 1.00 NVIDIA, can you modify that to 0.99 & 0.99 so that another project can be freed to utilize the rest of the CPU core used on einstein?

                                              I hope I'm not mistaken, but when I used the app_info, I put 0.3 & 0.9 and I could run a total of 5 projects, now I run obviously just 4.

                                              If the cpu or gpu is idling some of the time, or all the time, don't give it 100%, take a little bit off, like 99% :)

                                              I'm saying this thinking in order not to make another app_info, which I'll have to edit again when the ABP2 gets out.

                                              Mar.io
                                              Send message
                                              Joined: Oct 9 08
                                              Posts: 3
                                              Credit: 2,030,184
                                              RAC: 671
                                              Message 100736 - Posted 27 Nov 2009 11:42:26 UTC - in response to Message 100691.


                                                - NVidia GPU with at least 450MB of free memory
                                                - Display Driver version 190.38 (&up), i.e. CUDA 2.3 capability
                                                BM


                                                1. Are 450 MB really necessary? I only have 256 MB...
                                                2. Why CUDA 2.2 is not enough? It is the Laptop version of these drivers. Therefore no Laptop can use the GPU or am I wrong?

                                                Thanks
                                                Mario

                                                Oliver Bock
                                                Forum moderator
                                                Project administrator
                                                Project developer
                                                Send message
                                                Joined: Sep 4 07
                                                Posts: 359
                                                Credit: 20,915,781
                                                RAC: 1,674
                                                Message 100737 - Posted 27 Nov 2009 12:00:32 UTC - in response to Message 100736.


                                                  1. Are 450 MB really necessary? I only have 256 MB...


                                                  Yes, we really need that much memory.


                                                  2. Why CUDA 2.2 is not enough? It is the Laptop version of these drivers. Therefore no Laptop can use the GPU or am I wrong?


                                                  I think you are wrong as there are no "laptop" versions AFAIK. You may just download the latest driver for your operating system. CUDA 2.3 is superior to 2.2 in a number of details...

                                                  Oliver

                                                  Oliver Bock
                                                  Forum moderator
                                                  Project administrator
                                                  Project developer
                                                  Send message
                                                  Joined: Sep 4 07
                                                  Posts: 359
                                                  Credit: 20,915,781
                                                  RAC: 1,674
                                                  Message 100738 - Posted 27 Nov 2009 12:11:23 UTC - in response to Message 100735.

                                                    Instead of sending tasks which use 1.00 CPU + 1.00 NVIDIA, can you modify that to 0.99 & 0.99 so that another project can be freed to utilize the rest of the CPU core used on einstein?
                                                    I hope I'm not mistaken, but when I used the app_info, I put 0.3 & 0.9 and I could run a total of 5 projects, now I run obviously just 4.
                                                    If the cpu or gpu is idling some of the time, or all the time, don't give it 100%, take a little bit off, like 99% :)


                                                    That's what I tried to explain: as soon as we assign less than 100% (this includes 99%) BOINC will run the app with normal priority. Please keep in mind that BOINC is meant to run in the background, affecting the user as little as possible - and there are still volunteers out there running Einstein@Home on single or dual core CPUs...

                                                    Oliver

                                                    pelpolaris
                                                    Send message
                                                    Joined: Nov 5 06
                                                    Posts: 2
                                                    Credit: 4,055,251
                                                    RAC: 0
                                                    Message 100740 - Posted 27 Nov 2009 12:40:47 UTC - in response to Message 100716.

                                                      [quote]Do you have any idea of what kind of "Linows" or "Windux" platforms that can accept to run the Einstein CUDA 23 ?

                                                      Not sure what you mean by that, but their standard app is working fine on this Mandriva 2010.0 Linux system.

                                                      I only ment something as common as W32 or/and W64; Linux 32 or/and Linux64.

                                                      Profile XJR-Maniac
                                                      Avatar
                                                      Send message
                                                      Joined: Dec 8 05
                                                      Posts: 3
                                                      Credit: 670,148
                                                      RAC: 0
                                                      Message 100741 - Posted 27 Nov 2009 13:26:08 UTC - in response to Message 100720.

                                                        Now got my first (and probably last) ABP1 3.13 "CUDA" WU finished in over 10 hours time on a Q9650/GeForce GTX260 were CPU time was 8.3 hours and GPU time around 2 hours. This means more than 8 hours of wasted GPU time! Does that make any sense?

                                                        You don't have any CUDA tasks on your Q9650. The list of tasks for that machine shows 3 completed ABP1 tasks, all of which took around 17k secs and none of which used a GPU for crunching. I decided to look at your other machines and I found the GPU crunched ABP1 task on your Pentium D. There are no other ABP1 tasks still showing on that machine so there's no ability to do a comparison. Here is the list of tasks for your pentium D with the GPU crunched ABP1 task at the top. It's interesting to note that it took much the same time to crunch the ABP1 task (250 credits) as the two previous GW tasks (136 credits) - nearly double the credits for a tiny bit more crunch time.


                                                        Oh, I'm sorry for that. Seems that I got a little bit confused. Of course you're right when you say that it's not fair to compare a Pentium D with a Q9650. I thought that the CUDA app has been running on the Q9650 machine. My fault.

                                                        What is even more interesting is the apparent dramatic slowdown after Nov 20. The three earlier GW tasks took around 11K secs each while the two after this date took 27K and 29k secs respectively. Now there is variability in the GW crunch times but there is usually a variation in credits to compensate - at least partially. Since all GW tasks were awarded the same credit, it's unusual to see such a huge variation in crunch time. Can you think of anything that might have happened to your machine after Nov 20? Something drastic like halving the CPU frequency might do it :-).


                                                        Regarding the dramatic slowdown there is a very simple solution for that mystery. I had to swap my CPUs between those two machines because I got a new mainboard that doesn't support that Pentium D anymore, but the machine that has the Q9650 before has a mainboard "old" enough to run with that "old" Pentium D chip.

                                                        My last two ABP1 3.12 CPU only WUs took less than 5 hours on a Q9650.

                                                        It's not really fair to compare a Pentium D to a Q9650 :-).

                                                        Are the new WUs more complex or longer than the old ones or is this just another bad joke?

                                                        There aren't any new tasks - just the same old tasks being crunched with a new program which is (performance-wise) much the same as the beta test app it replaces.

                                                        It might help if you realise that just because one project can make hugely efficient use of a GPUs parallelism, other projects may struggle to do anything like the same even after considerable effort has been expended. You might take that into account when firing off your criticism.


                                                        OK, but if you come to a point where it's obvious that it makes no sense to continue a project like this disastrous CUDA application, I think someone should have the courage to stop it, instead of firing it out to the public with might and main, just to be able to say "look here, we have a CUDA app, too!".

                                                        I always say that BOINC is not a one man show. If you create a new application you should always keep in mind that there is a world outside your lab and you have to share the resources, that your volunteers are donating to you, with other projects.

                                                        And hey, don't tell me that I can leave the project if I don't like it. This is the most antisocial attitude I've heard of. So if you don't like to share the resources of this planet with others, maybe it's time for you to leave it!
                                                        ____________

                                                        Profile Megacruncher TSBT
                                                        Send message
                                                        Joined: Feb 11 05
                                                        Posts: 7
                                                        Credit: 15,042,091
                                                        RAC: 2,433
                                                        Message 100743 - Posted 27 Nov 2009 13:48:20 UTC

                                                          Dearie me, keep it civil folks, we are all working for the good of humankind after all.

                                                          Can I just clarify something. I gather from above that the CUDA application might only use the GPU for a small part of the time - 4% was mentioned. Does this mean that it can be working on Collatz etc for the the other 96%? Or is it not that simple and the GPU needs to hang around waiting for Einstein?

                                                          If it is the case that my GPU will be in a state of enforced idleness then I think I might chose to set it to work on something that keeps it a little busier.
                                                          ____________
                                                          The Scottish Boinc Team

                                                          cristipurdel
                                                          Send message
                                                          Joined: Jul 19 07
                                                          Posts: 17
                                                          Credit: 85,452
                                                          RAC: 0
                                                          Message 100745 - Posted 27 Nov 2009 14:11:59 UTC - in response to Message 100743.

                                                            Last modified: 27 Nov 2009 14:12:30 UTC

                                                            Dearie me, keep it civil folks, we are all working for the good of humankind after all.

                                                            Can I just clarify something. I gather from above that the CUDA application might only use the GPU for a small part of the time - 4% was mentioned. Does this mean that it can be working on Collatz etc for the the other 96%? Or is it not that simple and the GPU needs to hang around waiting for Einstein?

                                                            If it is the case that my GPU will be in a state of enforced idleness then I think I might chose to set it to work on something that keeps it a little busier.

                                                            You can set your resources smth like ... x CPU + y GPU for Einstein, and (1-x) CPU + (1-y) GPU for Collatz, but that would require a lot of 'attention' to new version updates, thus tweaking the app_info, and the boinc processes would get a 'normal' priority, which would most probably freeze your screen 'most of the time'.

                                                            Michael Goetz
                                                            Avatar
                                                            Send message
                                                            Joined: Feb 11 05
                                                            Posts: 15
                                                            Credit: 150,848
                                                            RAC: 0
                                                            Message 100746 - Posted 27 Nov 2009 14:43:42 UTC

                                                              I got my first three Einstein CUDA tasks last night. The computer is a C2D Q6600 (2.4GHz) with a factory-OC GTX280 GPU. Link to the machine's tasks is here: 1093664

                                                              As of the time I'm posting this, 5 ABP1 tasks have been crunched; 2 on CPU only and 3 on CPU/GPU. Since the credit given is 250 for both the CPU and CPU/GPU tasks are both the same, I'm assuming (perhaps incorrectly) that the same amount of work is being done by both.

                                                              CPU only tasks:
                                                              148600607
                                                              148695929

                                                              CPU/GPU tasks:
                                                              148948845
                                                              148986070
                                                              149018218

                                                              The interesting part is that CPU/GPU task ran only moderately faster than the CPU only tasks, despite consuming the expensive GPU resource. Also of note is that that GPU was running approximately 10 degrees Celsious cooler than it does when running GPUGRID. (The fan was running significantly slower with the Einstein app too.) This, of course, indicates that the GPU is being utilized much, much less when running the Einstein app than it is with GPUGRID.

                                                              My conclusion is that the GPU version of the app is making very inefficient use of the GPU -- in fact, judging by the time comparisons, it appears to be doing less work on the GPU than on the CPU!

                                                              As such, at least on this computer, it appears to be a waste of computing hardware to run the CUDA version of this app.

                                                              ____________
                                                              My lucky number is 75898^524288+1

                                                              Profile Bikeman (Heinz-Bernd Eggenstein)
                                                              Forum moderator
                                                              Project administrator
                                                              Project developer
                                                              Avatar
                                                              Send message
                                                              Joined: Aug 28 06
                                                              Posts: 3225
                                                              Credit: 72,411,621
                                                              RAC: 35,282
                                                              Message 100747 - Posted 27 Nov 2009 15:10:46 UTC - in response to Message 100729.

                                                                Last modified: 27 Nov 2009 16:13:45 UTC


                                                                ....
                                                                Let's assume (it's just a simple example) that an app does computations consisting of two parts, A and B, where A has to be executed before B can start. Let's assume that on a CPU, A and B each take 50% of the runtime.

                                                                Now assume that only part B can easily be ported to GPU-code, resulting in a (say) 25 fold speedup for this part B. A still has to be done on the CPU.

                                                                The result: If the total runtime was 1000 sec before, it is now 520 s, almost doubling the performance. Only 20 seconds of these 520 seconds will be spent in the GPU, or below 4 %. So even small load factors on the GPU can result in reasonable speedups.
                                                                ...
                                                                Bikeman

                                                                Are this the actual numbers for ABP1cuda23 ?
                                                                If not, could you post the actual speedup on your machine between the GPU and CPU version?


                                                                Hi!

                                                                This was just an example, and more important there IS no single speedup factor for this kind of app. Relative performance compared with running on CPU only will depend drastically on the combination of CPU and GPU. E.g. running on an AMD CPU with a rather powerful GPU will give you a better speedup, because the part that is done on the GPU executes rather not-so-well on AMD CPUs (the CPU version is slower than on Intels)....etc. On a fast Intel Core2 or i7 the speedup will be less, even with powerful GPUs.

                                                                I'm currently travelling and have not yet managed to deactivate my app_info.xml on my only CUDA machine that is capable to run this app, so I can't provide numbers.

                                                                CU
                                                                Bikeman
                                                                ____________

                                                                John Clark
                                                                Avatar
                                                                Send message
                                                                Joined: May 4 07
                                                                Posts: 1092
                                                                Credit: 3,143,193
                                                                RAC: 0
                                                                Message 100748 - Posted 27 Nov 2009 15:46:25 UTC - in response to Message 100743.

                                                                  Dearie me, keep it civil folks, we are all working for the good of humankind after all.

                                                                  Can I just clarify something. I gather from above that the CUDA application might only use the GPU for a small part of the time - 4% was mentioned. Does this mean that it can be working on Collatz etc for the the other 96%? Or is it not that simple and the GPU needs to hang around waiting for Einstein?

                                                                  If it is the case that my GPU will be in a state of enforced idleness then I think I might chose to set it to work on something that keeps it a little busier.



                                                                  You are possibly already aware of this, but I will state it anyway.

                                                                  I find that GPU crunching (mainly based on ATI 38xx and 48xx creatures) will take 1 project on the GPU and a single, or more, on the CPUs. But crunching Collatz and MW on the same GPU seems impossible and the GPU remains idle.

                                                                  Outside of using the CPU for other projects, the way I go is to attach to say Collatz and GPU crunch. I then set the project to NNT, and attach to MM and GPU crunch here. If problems remain on the project then it;s a simple case of moving from Collatz to MW and back by suspending the one I don't want to crunch.

                                                                  ____________
                                                                  Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!

                                                                  Grutte Pier [Wa Oars]~GP500
                                                                  Avatar
                                                                  Send message
                                                                  Joined: May 18 09
                                                                  Posts: 39
                                                                  Credit: 2,462,785
                                                                  RAC: 2,761
                                                                  Message 100757 - Posted 27 Nov 2009 21:43:04 UTC - in response to Message 100743.

                                                                    Last modified: 27 Nov 2009 21:49:43 UTC

                                                                    Dearie me, keep it civil folks, we are all working for the good of humankind after all.

                                                                    Can I just clarify something. I gather from above that the CUDA application might only use the GPU for a small part of the time - 4% was mentioned. Does this mean that it can be working on Collatz etc for the the other 96%? Or is it not that simple and the GPU needs to hang around waiting for Einstein?

                                                                    If it is the case that my GPU will be in a state of enforced idleness then I think I might chose to set it to work on something that keeps it a little busier.


                                                                    Nope, that's the problem with a GPU you can't devide 2 cuda project on 1 card.

                                                                    That's why EINSTEIn should run on CPU totally or only CPU.

                                                                    When you run E@H on both you would also keep a better return from the total machine. but atleast you could switch between projects on the GPU.

                                                                    Now you can only get a & gain for the total of EInstein or run another project on the GPU. Wich would do more good for the machine.



                                                                    Ps: EInstein team: there is a cuda 3 version coming, maybe make it so that the wu's are only GPU.
                                                                    CUDA 3.0

                                                                    Profile Megacruncher TSBT
                                                                    Send message
                                                                    Joined: Feb 11 05
                                                                    Posts: 7
                                                                    Credit: 15,042,091
                                                                    RAC: 2,433
                                                                    Message 100767 - Posted 28 Nov 2009 0:48:36 UTC

                                                                      Okay then. Nice try but possibly only relevant to an Einstein mono-cruncher with a buckshee high end graphics card going to waste.
                                                                      The rest of us will run Einstein on our CPUs and use our GPUs for something more generous and thought out. Thanks anyway!
                                                                      ____________
                                                                      The Scottish Boinc Team

                                                                      Profile MarkJ
                                                                      Avatar
                                                                      Send message
                                                                      Joined: Feb 28 08
                                                                      Posts: 212
                                                                      Credit: 25,053,374
                                                                      RAC: 287
                                                                      Message 100768 - Posted 28 Nov 2009 3:46:38 UTC

                                                                        Well I got one unannounced but let it run through to see how it went anyway. The GPU in the machine is a GTX275. The workunit is here

                                                                        It took approx 1/3rd off the "normal" run time, so its faster. As the others have pointed out at the expense of using both a cpu and a gpu for 5.5 hours. A 1.25 hour cpu time saving for 5.5 hours GPU time doesn't seem a reasonable use of the GPU to me.
                                                                        ____________
                                                                        BOINC blog

                                                                        Profile Kty
                                                                        Avatar
                                                                        Send message
                                                                        Joined: Oct 18 09
                                                                        Posts: 21
                                                                        Credit: 8,278,893
                                                                        RAC: 1
                                                                        Message 100769 - Posted 28 Nov 2009 9:49:40 UTC - in response to Message 100768.

                                                                          Hello,

                                                                          I have a Pentium IV 3.06 (2 logical cores) with nVidia 9400 GT.
                                                                          My first cuda WU took 15 hours to be treated with 1.00 CPU + 1.00 NVIDIA GPU.
                                                                          And I got... 250 credits.

                                                                          I had the same amount of credits (250) for one Pulsar Search treated with 1.00 CPU only running on my laptop - Core DUO at 2.00 GHz...

                                                                          It's not worth it !

                                                                          So lets resume the situation : Einstein cuda WU runs fine but the claimed credits is too low. That really doesn't help to get involved in the project. The amount of credits for Cuda WU using 2 cores (CPU and GPU) should be raised up.

                                                                          Lionel

                                                                          John Clark
                                                                          Avatar
                                                                          Send message
                                                                          Joined: May 4 07
                                                                          Posts: 1092
                                                                          Credit: 3,143,193
                                                                          RAC: 0
                                                                          Message 100771 - Posted 28 Nov 2009 10:22:54 UTC

                                                                            Last modified: 28 Nov 2009 10:23:21 UTC

                                                                            Perhaps the best approach to GPU crunching (whether CUDA - nVidia - or ATI) is to use the cards on projects which can use the GPU properly (MW, Collatz, etc) and use the rest of the CPUs on Einstein. That is the approach I am taking.
                                                                            ____________
                                                                            Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!

                                                                            Profile JockMacMad TSBT
                                                                            Send message
                                                                            Joined: Jun 5 09
                                                                            Posts: 1
                                                                            Credit: 48,658
                                                                            RAC: 0
                                                                            Message 100772 - Posted 28 Nov 2009 12:23:29 UTC

                                                                              The is also a change to the CUDA drivers coming when they support DX11 I beliee that will allow multiple CUDA processes on to a GPU the way ATI does at the moment.

                                                                              Michael Goetz
                                                                              Avatar
                                                                              Send message
                                                                              Joined: Feb 11 05
                                                                              Posts: 15
                                                                              Credit: 150,848
                                                                              RAC: 0
                                                                              Message 100773 - Posted 28 Nov 2009 13:00:52 UTC - in response to Message 100769.

                                                                                Einstein cuda WU runs fine but the claimed credits is too low.


                                                                                Actually, the credits are correct given the amount of work that's actualy being done. The problem is that the application is barely using the GPU at all. Even though the GPU is tied up for hours, most of the crunching is still being done on the CPU.

                                                                                It's not that you're not getting credit for what your GPU is doing; the problem is that E@H is wasting the GPU time you're giving it and not doing much on the GPU.

                                                                                Do what I (and lots of others) have done -- go to the E@H preferences and turn off GPU processing. Run E@H only on the CPU, and let some other project use the GPU.
                                                                                ____________
                                                                                My lucky number is 75898^524288+1

                                                                                Mark Rush
                                                                                Send message
                                                                                Joined: Jan 18 05
                                                                                Posts: 1
                                                                                Credit: 49,095,200
                                                                                RAC: 25,583
                                                                                Message 100776 - Posted 28 Nov 2009 16:26:12 UTC - in response to Message 100773.

                                                                                  I agree with others that using the GPU only a fraction of the time is a waste of resources. I have been using my GPU exclusively for SETI. "Fortunately" SETI went down this weekend and I have worked through my backlog of SETI Cuda WUs, so the Einstein Cuda WU's impose no opportunity cost to me by forcing my GPU to set idel while I have uncrunched SETI Cuda WUs. When SETI is back up, I shall (sadly) select no GPU for Einstein. But I will keep an active eye out for ABP2 application to see if it more efficiently uses the GPU. And I'll be really eager for the "all Cuda" port because I would very much like to have both SETI and Einstein splitting time on my GPU.

                                                                                  Mark
                                                                                  ____________

                                                                                  Profile Kty
                                                                                  Avatar
                                                                                  Send message
                                                                                  Joined: Oct 18 09
                                                                                  Posts: 21
                                                                                  Credit: 8,278,893
                                                                                  RAC: 1
                                                                                  Message 100777 - Posted 28 Nov 2009 19:10:27 UTC - in response to Message 100773.

                                                                                    Hello,

                                                                                    Is this the way we know that time is wasted :

                                                                                    My Cuda WU was created at 02:18:30 UTC and was crunched straight to 15:25:36. That makes 47226 sec. The result indicate 52288 sec.

                                                                                    Using one CPU and one GPU - two processing unit - should give
                                                                                    47226 x 2 = 94452, no ?

                                                                                    that makes a difference of 42164 sec... wasted ?

                                                                                    Is it correct ?

                                                                                    Bsts rgrds.

                                                                                    Lionel.

                                                                                    Profile Bryan
                                                                                    Avatar
                                                                                    Send message
                                                                                    Joined: Aug 12 07
                                                                                    Posts: 7
                                                                                    Credit: 870,833
                                                                                    RAC: 0
                                                                                    Message 100779 - Posted 28 Nov 2009 20:22:33 UTC

                                                                                      Well, it seems that my piddling little 256MB Nvidia card (though big enough to run 2 big fat monitors) isn't enough to run the new app. Meanwhile, I'm not getting any CPU units either. Are all the new units CUDA only?

                                                                                      Michael Goetz
                                                                                      Avatar
                                                                                      Send message
                                                                                      Joined: Feb 11 05
                                                                                      Posts: 15
                                                                                      Credit: 150,848
                                                                                      RAC: 0
                                                                                      Message 100780 - Posted 28 Nov 2009 21:29:53 UTC - in response to Message 100777.


                                                                                        Is this the way we know that time is wasted :


                                                                                        There's two ways you can tell the GPU isn't being used effectively.

                                                                                        Firstly, if you have a utility that displays the GPU temperature, that's a very good indication of how much the GPU is doing. My GPU is running MUCH cooler running E@H than when it's running any other application.

                                                                                        Secondly, you can compare the run-time of CPU and CPU/GPU workunits of the same application (Einstein, SETI, and Milkyway all offer the same app in both CPU and GPU versions). Depending on your hardware, your GPU is usually 10 to 20 times faster than your CPU, so you should expect the GPU WUs to finish in only 5% o 10% of the time it takes the same CPU WU to finish (which represents a 1000% to 2000% speed increase).

                                                                                        On Einstein, everyone is seeing cold GPU temps and only about a 33% increase in speed -- while using hardware that's physically around 2000% faster.

                                                                                        Or, let me put it this way: If it normally takes your machine about 10 hours to complete an E@H WU, if it was making good use of the GPU, it would finish WUs in around 30 to 60 minutes, give or take.

                                                                                        To put things in perspective, if I run Milkyway@home on my CPU, it take a bit under four hours to run. M@H pushes my GPU harder than any other app, as evidenced by higher temperatures. When I run M@H WUs on my GPU, they complete in slightly under 4 minutes. A 6000% increase. Compared to the 33% increase I'm seeing on E@H.
                                                                                        ____________
                                                                                        My lucky number is 75898^524288+1

                                                                                        Grutte Pier [Wa Oars]~GP500
                                                                                        Avatar
                                                                                        Send message
                                                                                        Joined: May 18 09
                                                                                        Posts: 39
                                                                                        Credit: 2,462,785
                                                                                        RAC: 2,761
                                                                                        Message 100786 - Posted 29 Nov 2009 1:00:35 UTC

                                                                                          Last modified: 29 Nov 2009 1:02:24 UTC

                                                                                          All we can say is, there is allot more potential For CUDA @ Einstein@home.


                                                                                          And these are the first steps to moving on.
                                                                                          Without a start there is no motion.

                                                                                          @ EINstein@home ask for more support from the big community.
                                                                                          There are people here that can help with the development of a cuda AND stream client for this.

                                                                                          You can multiply the total TFLOPS of troughput with that.

                                                                                          Profile Tom95134
                                                                                          Send message
                                                                                          Joined: Aug 21 07
                                                                                          Posts: 18
                                                                                          Credit: 1,842,773
                                                                                          RAC: 0
                                                                                          Message 100788 - Posted 29 Nov 2009 6:17:00 UTC

                                                                                            First let me say I'm crunching CUDA numbers just fine for both SETI@Home and Einstein@Home.

                                                                                            I only have one comment. SETI@Home (which also feeds CUDA tasks) breaks their CUDA tasks down into much smaller "chunks", i.e., up to about 30 minutes. Because of the longer tasks coming from Einstein@Home this will capture the GPU which essentially means it doesn't play well with others.

                                                                                            Right now SETI is going through some problems keeping tasks fed but once these issues are sorted out I suspect that the longer tasks from Einstein will mean that everyone that participates in both programs will either have to do some very fine tweaking of the percentages allocated to each project or abandon accepting CUDA tasks from one of the projects. Keep in mind that you cannot tweak CPU project use separate from GU project use.

                                                                                            Profile Tom95134
                                                                                            Send message
                                                                                            Joined: Aug 21 07
                                                                                            Posts: 18
                                                                                            Credit: 1,842,773
                                                                                            RAC: 0
                                                                                            Message 100789 - Posted 29 Nov 2009 6:22:00 UTC - in response to Message 100780.

                                                                                              Last modified: 29 Nov 2009 6:22:25 UTC


                                                                                              Is this the way we know that time is wasted :


                                                                                              There's two ways you can tell the GPU isn't being used effectively.

                                                                                              Firstly, if you have a utility that displays the GPU temperature, that's a very good indication of how much the GPU is doing. My GPU is running MUCH cooler running E@H than when it's running any other application.

                                                                                              Secondly, you can compare the run-time of CPU and CPU/GPU workunits of the same application (Einstein, SETI, and Milkyway all offer the same app in both CPU and GPU versions). Depending on your hardware, your GPU is usually 10 to 20 times faster than your CPU, so you should expect the GPU WUs to finish in only 5% o 10% of the time it takes the same CPU WU to finish (which represents a 1000% to 2000% speed increase).

                                                                                              On Einstein, everyone is seeing cold GPU temps and only about a 33% increase in speed -- while using hardware that's physically around 2000% faster.

                                                                                              Or, let me put it this way: If it normally takes your machine about 10 hours to complete an E@H WU, if it was making good use of the GPU, it would finish WUs in around 30 to 60 minutes, give or take.

                                                                                              To put things in perspective, if I run Milkyway@home on my CPU, it take a bit under four hours to run. M@H pushes my GPU harder than any other app, as evidenced by higher temperatures. When I run M@H WUs on my GPU, they complete in slightly under 4 minutes. A 6000% increase. Compared to the 33% increase I'm seeing on E@H.


                                                                                              I agree that something needs to be done to optimize the CUDA tasks. Because of this long processing time for GPU targeted tasks it is critical that something be done about the size of the GPU tasks coming from Einstein@Home. If using the GPU is not that much more efficient in processing tasks then one might as well not allow CUDA for Einstein@Home.
                                                                                              ____________

                                                                                              Profile Kty
                                                                                              Avatar
                                                                                              Send message
                                                                                              Joined: Oct 18 09
                                                                                              Posts: 21
                                                                                              Credit: 8,278,893
                                                                                              RAC: 1
                                                                                              Message 100794 - Posted 29 Nov 2009 9:04:15 UTC - in response to Message 100789.

                                                                                                Hello,

                                                                                                One of my computers is a laptop using a NVIDIA 9600M GS.
                                                                                                E@H refuse to upload Cuda WU on this laptop because the display driver 190.38 is requested.
                                                                                                After checking on NVIDIA web site, it appears that this driver 190.38 is not for laptops or M graphics cards series. The latest display driver available for GPU M series is the 186.81.

                                                                                                So if you're using a laptop with NVIDIA GPU, don't waste your time triying to run E@H cuda WU. it is not possible right now.

                                                                                                Lionel.

                                                                                                Ver Greeneyes
                                                                                                Send message
                                                                                                Joined: Mar 26 09
                                                                                                Posts: 139
                                                                                                Credit: 4,088,916
                                                                                                RAC: 0
                                                                                                Message 100797 - Posted 29 Nov 2009 12:52:29 UTC - in response to Message 100794.

                                                                                                  Last modified: 29 Nov 2009 12:53:02 UTC

                                                                                                  Lionel, try one of the drivers from LaptopVideo2Go. Be sure to read the instructions on how to use a modified inf, it's simple when you know how!

                                                                                                  Michael Goetz
                                                                                                  Avatar
                                                                                                  Send message
                                                                                                  Joined: Feb 11 05
                                                                                                  Posts: 15
                                                                                                  Credit: 150,848
                                                                                                  RAC: 0
                                                                                                  Message 100798 - Posted 29 Nov 2009 13:05:13 UTC - in response to Message 100788.

                                                                                                    I only have one comment. SETI@Home (which also feeds CUDA tasks) breaks their CUDA tasks down into much smaller "chunks", i.e., up to about 30 minutes. Because of the longer tasks coming from Einstein@Home this will capture the GPU which essentially means it doesn't play well with others.


                                                                                                    The problem isn't the length of the WU. S@H's WU are actually longer (i.e., more computation is done) than Einstein's. On my computer, SETI's WUs take about 10 hours vs. 6 hours for Einstein. The difference is that those same identical work units, run on my GPU, drop to 10 minutes for SETI but only 4 hours for Einstein.

                                                                                                    For both SETI and Einstein, the same WUs are sent to both CPU and GPU computers. If you look at the results for both projects, you'll see this -- for any of your GPU WUs, chances are likely that the other computer ran it on the CPU.

                                                                                                    It's not the size of the WU that's the problem -- it's that the GPU is barely being used.

                                                                                                    As for BOINC scheduling of short vs. long tasks on the GPU -- that *should* be fine. "Should* is the important word there. I don't actually know how well the BOINC client schedules GPU tasks. But how it's supposed to work is clear: If the work percentages are the same, two projects should each get 50% of the GPU time, regardless of the size of the WUs. More of the shorter WUs should run, as compared to the longer WUs.

                                                                                                    Oh, I want to make something clear. While it's painfully obvious that the E@H application isn't utilizing the GPU very well, I am in no way implying that the people who wrote the application did a poor job. Some problems simply do not lend themselves to being solved on a massively parallel computer (i.e., a GPU). It may be that Einstein is one of those problems, and it just can't be run efficiently on a GPU.
                                                                                                    ____________
                                                                                                    My lucky number is 75898^524288+1

                                                                                                    Walker
                                                                                                    Send message
                                                                                                    Joined: Nov 29 05
                                                                                                    Posts: 1
                                                                                                    Credit: 218,578
                                                                                                    RAC: 65
                                                                                                    Message 100799 - Posted 29 Nov 2009 13:36:12 UTC - in response to Message 100741.

                                                                                                      OK, but if you come to a point where it's obvious that it makes no sense to continue a project like this disastrous CUDA application, I think someone should have the courage to stop it, instead of firing it out to the public with might and main, just to be able to say "look here, we have a CUDA app, too!".

                                                                                                      I always say that BOINC is not a one man show. If you create a new application you should always keep in mind that there is a world outside your lab and you have to share the resources, that your volunteers are donating to you, with other projects.

                                                                                                      And hey, don't tell me that I can leave the project if I don't like it. This is the most antisocial attitude I've heard of. So if you don't like to share the resources of this planet with others, maybe it's time for you to leave it!


                                                                                                      i agree with that !!
                                                                                                      ____________

                                                                                                      Profile Gundolf Jahn
                                                                                                      Send message
                                                                                                      Joined: Mar 1 05
                                                                                                      Posts: 1079
                                                                                                      Credit: 328,104
                                                                                                      RAC: 18
                                                                                                      Message 100800 - Posted 29 Nov 2009 15:00:33 UTC - in response to Message 100799.

                                                                                                        ...And hey, don't tell me that I can leave the project if I don't like it. This is the most antisocial attitude I've heard of. So if you don't like to share the resources of this planet with others, maybe it's time for you to leave it!


                                                                                                        i agree with that !!

                                                                                                        No one told anyone to leave the project. XJR-Maniac was just quite rude without reason.

                                                                                                        If you don't want Einstein CUDA tasks, deselect them in your preferences. Nothing easier than that.

                                                                                                        Gruß,
                                                                                                        Gundolf
                                                                                                        ____________
                                                                                                        Computer sind nicht alles im Leben. (Kleiner Scherz)

                                                                                                        [B^S] Elphidieus
                                                                                                        Send message
                                                                                                        Joined: Feb 20 05
                                                                                                        Posts: 220
                                                                                                        Credit: 12,347,408
                                                                                                        RAC: 0
                                                                                                        Message 100803 - Posted 29 Nov 2009 15:54:48 UTC

                                                                                                          Curious... Just noticed the option for ABP Search (SP) on my Preferences. Am I to assume it's referring to the CUDA App...?
                                                                                                          ____________

                                                                                                          Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                          Forum moderator
                                                                                                          Project administrator
                                                                                                          Project developer
                                                                                                          Avatar
                                                                                                          Send message
                                                                                                          Joined: Aug 28 06
                                                                                                          Posts: 3225
                                                                                                          Credit: 72,411,621
                                                                                                          RAC: 35,282
                                                                                                          Message 100804 - Posted 29 Nov 2009 16:08:57 UTC - in response to Message 100803.

                                                                                                            Curious... Just noticed the option for ABP Search (SP) on my Preferences. Am I to assume it's referring to the CUDA App...?


                                                                                                            Not exactly.

                                                                                                            The two options for "Arecibo Pulsar Search" and "ABP Search (SP)" refer to two different searches, ABP1 and ABP2.

                                                                                                            Currently only ABP1 work is distributed, there are ABP1 apps for CPU and GPU (CUDA). If you want to continue crunching ABP1 on the CPU but would like not to crunch them on the GPU, you should deselect the "Use GPU" option in the same configuration screen.

                                                                                                            While the ABP1 search is doing some of its processing in single precision (e.g. the part that also gets executed on the GPU in the CUDA app), other parts are still done in double precision. The new ABP2 apps will be doing nearly everything in single precision, which will also allow to put more load on the GPU with high efficiency.

                                                                                                            CU
                                                                                                            Bikeman

                                                                                                            ____________

                                                                                                            [B^S] Elphidieus
                                                                                                            Send message
                                                                                                            Joined: Feb 20 05
                                                                                                            Posts: 220
                                                                                                            Credit: 12,347,408
                                                                                                            RAC: 0
                                                                                                            Message 100805 - Posted 29 Nov 2009 16:47:28 UTC - in response to Message 100804.

                                                                                                              Curious... Just noticed the option for ABP Search (SP) on my Preferences. Am I to assume it's referring to the CUDA App...?


                                                                                                              Not exactly.

                                                                                                              The two options for "Arecibo Pulsar Search" and "ABP Search (SP)" refer to two different searches, ABP1 and ABP2.

                                                                                                              Currently only ABP1 work is distributed, there are ABP1 apps for CPU and GPU (CUDA). If you want to continue crunching ABP1 on the CPU but would like not to crunch them on the GPU, you should deselect the "Use GPU" option in the same configuration screen.

                                                                                                              While the ABP1 search is doing some of its processing in single precision (e.g. the part that also gets executed on the GPU in the CUDA app), other parts are still done in double precision. The new ABP2 apps will be doing nearly everything in single precision, which will also allow to put more load on the GPU with high efficiency.

                                                                                                              CU
                                                                                                              Bikeman


                                                                                                              Nah... I'm not too worry about CUDA apps, Macs can't even support CUDA. Just curious about the new SP option.

                                                                                                              Any idea when will the new ABP2 search be implemented...?
                                                                                                              ____________

                                                                                                              Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                              Forum moderator
                                                                                                              Project administrator
                                                                                                              Project developer
                                                                                                              Avatar
                                                                                                              Send message
                                                                                                              Joined: Aug 28 06
                                                                                                              Posts: 3225
                                                                                                              Credit: 72,411,621
                                                                                                              RAC: 35,282
                                                                                                              Message 100807 - Posted 29 Nov 2009 17:33:58 UTC - in response to Message 100805.

                                                                                                                Last modified: 29 Nov 2009 21:50:49 UTC


                                                                                                                Any idea when will the new ABP2 search be implemented...?


                                                                                                                According to Oliver's message here, probably in the next 1-2 weeks. I guess the road-map will be something like

                                                                                                                - Beta-Test for ABP2 CPU app
                                                                                                                - Beta-Test for ABP2 CUDA app
                                                                                                                - Release of ABP2 CPU app
                                                                                                                - Release of ABP2 CUDA app

                                                                                                                CU
                                                                                                                Bikeman
                                                                                                                ____________

                                                                                                                Profile Olaf
                                                                                                                Send message
                                                                                                                Joined: Sep 16 06
                                                                                                                Posts: 23
                                                                                                                Credit: 112,819,327
                                                                                                                RAC: 124,231
                                                                                                                Message 100822 - Posted 30 Nov 2009 9:58:11 UTC - in response to Message 100691.


                                                                                                                  CUDA Beta App testers should drain their work cache and switch back to the normal project work.

                                                                                                                  BM


                                                                                                                  This looks like an opt out condition ;o) With the beta version it worked on my
                                                                                                                  computers. First I simply removed the app_info.xml (after I finished all work
                                                                                                                  and stopped boinc). But starting it again resulted in the message, that BOINC
                                                                                                                  was too old (installed just one month ago) to get some work from einstein@home.
                                                                                                                  Surprising, because other computers without GPUs still use BOINC 5.10 with
                                                                                                                  getting work from einstein@home (I cannot update this, because the home
                                                                                                                  directory is the same for several computers, what seems to stop newer versions
                                                                                                                  of BOINC working, independently from the fact that they are installed in
                                                                                                                  different directories for each computer).
                                                                                                                  I switched to BOINC 6.10 for this new computer and the result is, that the
                                                                                                                  GPU gets no work anymore, because of the driver/CUDA2.3 condition - BOINC was
                                                                                                                  not able to get the correct driver version and it obviously uses CUDA2.2.
                                                                                                                  Well, no big problem, because there is anyway no big difference, whether it
                                                                                                                  works with or without the GPU. Now the CPUs start to crunch the ABP again.
                                                                                                                  The installed driver is already from derived from NVidia, not that what is
                                                                                                                  indicated from Debian to be the current stable version.

                                                                                                                  Because I do not want to reinstall every month new experimental drivers on an
                                                                                                                  else stable computer, for now this CORE i7 crunches without the GPU again.
                                                                                                                  For another new notebook I think I will continue to crunch with this
                                                                                                                  app_info.xml, just because it worked without problems and for this notebook
                                                                                                                  GPU those ABP are ok (looks like it is already to small or to slow for GPUgrid ;o)

                                                                                                                  Profile Mike Hewson
                                                                                                                  Forum moderator
                                                                                                                  Avatar
                                                                                                                  Send message
                                                                                                                  Joined: Dec 1 05
                                                                                                                  Posts: 3444
                                                                                                                  Credit: 27,813,552
                                                                                                                  RAC: 5,485
                                                                                                                  Message 100824 - Posted 30 Nov 2009 11:41:04 UTC - in response to Message 100800.

                                                                                                                    ...And hey, don't tell me that I can leave the project if I don't like it. This is the most antisocial attitude I've heard of. So if you don't like to share the resources of this planet with others, maybe it's time for you to leave it!


                                                                                                                    i agree with that !!

                                                                                                                    No one told anyone to leave the project. XJR-Maniac was just quite rude without reason.

                                                                                                                    If you don't want Einstein CUDA tasks, deselect them in your preferences. Nothing easier than that.

                                                                                                                    Yes, I too am unsure how that deduction was made from Gary's post. :-)

                                                                                                                    In any case the primary requirement for CUDA to yield significant benefit is that the problem must lend itself to massive parallelism ( ideally thousands of threads, plus other restrictions ). This is a basic reason ( plus of course issues like compiler technology ) that leads to variable success with apps.

                                                                                                                    The development here at E@H is quite cautious, with a considerable user pool feeding back via beta testing. CUDA is no exception. While not always successful ( a failure outcome is within the definition of testing ), one hopes to be able to productively generalise beyond the test participants. One can opt out of CUDA if it doesn't fly well enough. In fact that is likely to be a common response for those with unsuitable hardware for optimal CUDA use. Alas as Oliver pointed out, without changing BOINC code ( not under E@H control ) then a default setting of opt-out was/is not available.

                                                                                                                    Cheers, Mike.
                                                                                                                    ____________
                                                                                                                    "I have made this letter longer than usual, because I lack the time to make it short." - Blaise Pascal

                                                                                                                    Jim Kleine
                                                                                                                    Send message
                                                                                                                    Joined: Jan 25 06
                                                                                                                    Posts: 3
                                                                                                                    Credit: 1,053,126
                                                                                                                    RAC: 0
                                                                                                                    Message 100825 - Posted 30 Nov 2009 11:56:15 UTC

                                                                                                                      Hmmm ... the CUDA app also seems to have a similar problem to GPUGrid, in that it fails to correctly detect CUDA hardware on a Windows host that has been connected to with Remote Desktop and has not yet been accessed again from the console:

                                                                                                                      <core_client_version>6.10.17</core_client_version>
                                                                                                                      <snip>
                                                                                                                      <stderr_txt>
                                                                                                                      Activated exception handling...
                                                                                                                      [20:00:57][4508][INFO ] Starting data processing...
                                                                                                                      [20:00:57][4508][INFO ] Using CUDA device #0 "Device Emulation (CPU)" (518.40 GFLOPS)
                                                                                                                      [20:00:57][4508][INFO ] Checkpoint file unavailable: status.cpt (No such file or directory).
                                                                                                                      ------> Starting from scratch...
                                                                                                                      <snip>
                                                                                                                      [20:00:58][4508][WARN ] Couldn't allocate 25165824 bytes of CUDA pinned host memory for resampled time series! Trying fallback...
                                                                                                                      [20:00:58][4508][WARN ] Couldn't allocate 25165832 bytes of CUDA pinned host memory for resampled time series FFT! Trying fallback...
                                                                                                                      [20:00:58][4508][ERROR] Error allocating CUDA device memory: 25165832 bytes (no CUDA-capable device is available)
                                                                                                                      [20:00:58][4508][ERROR] Demodulation failed (error: 3)!
                                                                                                                      20:00:58 (4508): called boinc_finish
                                                                                                                      </stderr_txt>

                                                                                                                      The GPU in this host is a 512MB 9800GT:

                                                                                                                      <core_client_version>6.10.17</core_client_version>
                                                                                                                      <![CDATA[
                                                                                                                      <stderr_txt>
                                                                                                                      # Using CUDA device 0
                                                                                                                      # There is 1 device supporting CUDA
                                                                                                                      # Device 0: "GeForce 9800 GT"
                                                                                                                      # Clock rate: 1.62 GHz
                                                                                                                      # Total amount of global memory: 536543232 bytes
                                                                                                                      # Number of multiprocessors: 14
                                                                                                                      # Number of cores: 112

                                                                                                                      GPUGrid work units that are already running will continue to run without issue across a remote desktop session, but no new work units can be started until you have logged in again from the console first. This may be related to GPUGrid seemingly using a deprecated function to test for the presence of CUDA hardware: I'm not sure if you have the same problem.

                                                                                                                      My current workaround for GPUGrid is to suspend work fetch when I am expecting to be away and to suspend all tasks except the currently active one. This strategy, of course, regularly results in GPU idle time.

                                                                                                                      Unfortunately, the "stealth" release of the CUDA version caught me unawares and resulted in a swag of errored work units within the space of ~90 seconds (about five hours ago, while I was out) due to this problem and that, in turn, has reduced this host's quota to 2/day, so it seems I won't be doing much Einstein work on this host (CPU or GPU) for some time :-(

                                                                                                                      In an ideal world (a) the CUDA hardware would continue to be available despite the Remote Desktop video driver having been invoked and (b) if a work unit fails for this reason, the queue of GPU WUs needs to be paused, since once the first has failed, all the others in the queue *are* going to suffer a similar fate.
                                                                                                                      ____________

                                                                                                                      Profile Gundolf Jahn
                                                                                                                      Send message
                                                                                                                      Joined: Mar 1 05
                                                                                                                      Posts: 1079
                                                                                                                      Credit: 328,104
                                                                                                                      RAC: 18
                                                                                                                      Message 100830 - Posted 30 Nov 2009 12:56:53 UTC - in response to Message 100825.

                                                                                                                        That has nothing to do with any projects's application but all with Remote Desktop and how Microsoft handles the graphics drivers.

                                                                                                                        There've been plenty threads at SETI and BOINC dev about that topic.

                                                                                                                        Gruß,
                                                                                                                        Gundolf
                                                                                                                        ____________
                                                                                                                        Computer sind nicht alles im Leben. (Kleiner Scherz)

                                                                                                                        Jim Kleine
                                                                                                                        Send message
                                                                                                                        Joined: Jan 25 06
                                                                                                                        Posts: 3
                                                                                                                        Credit: 1,053,126
                                                                                                                        RAC: 0
                                                                                                                        Message 100831 - Posted 30 Nov 2009 13:40:48 UTC - in response to Message 100830.

                                                                                                                          Last modified: 30 Nov 2009 13:46:56 UTC

                                                                                                                          That has nothing to do with any projects's application but all with Remote Desktop and how Microsoft handles the graphics drivers.

                                                                                                                          There've been plenty threads at SETI and BOINC dev about that topic.

                                                                                                                          Gruß,
                                                                                                                          Gundolf


                                                                                                                          I don't wish to appear rude, but did you actually read my post?

                                                                                                                          a) "... GPUGrid appears to be using a deprecated function ..."

                                                                                                                          b) If it is "unfixable" (and I'm not yet convinced, but I don't frequent the SETI board), then my suggestion that "if a work unit fails for this reason, the queue of GPU WUs needs to be paused" seems all the more relevant.

                                                                                                                          More importantly, running work units don't spontaneously abort upon a Remote Desktop connection being initiated, so the CUDA hardware clearly remains accessible to the *running* app, which leads me to question whether there is an assertion in the start-up code that isn't testing what it thinks it is testing.

                                                                                                                          As a follow up:

                                                                                                                          [BOINC] #936: CUDA devices not detected when logged in through Remote Desktop
                                                                                                                          ---------------------------+------------------------------------------------ Reporter: mart0258 | Owner: Type: Enhancement | Status: closed Priority: Undetermined | Milestone: Undetermined Component: Undetermined | Version: 6.6.31 Resolution: fixed | Keywords: ---------------------------+------------------------------------------------

                                                                                                                          Changes (by romw):
                                                                                                                          * status: new => closed
                                                                                                                          * resolution: => fixed
                                                                                                                          Comment:
                                                                                                                          This is now fixed in the 6.10 version of the BOINC client.

                                                                                                                          "resolution: => fixed" doesn't seem consistent with your comment, although this issue doesn't actually seem to be fixed as of 6.10.17.

                                                                                                                          Oliver Bock
                                                                                                                          Forum moderator
                                                                                                                          Project administrator
                                                                                                                          Project developer
                                                                                                                          Send message
                                                                                                                          Joined: Sep 4 07
                                                                                                                          Posts: 359
                                                                                                                          Credit: 20,915,781
                                                                                                                          RAC: 1,674
                                                                                                                          Message 100832 - Posted 30 Nov 2009 15:19:45 UTC - in response to Message 100825.

                                                                                                                            Hi Jim,

                                                                                                                            Hmmm ... the CUDA app also seems to have a similar problem to GPUGrid, in that it fails to correctly detect CUDA hardware on a Windows host that has been connected to with Remote Desktop and has not yet been accessed again from the console:


                                                                                                                            We (and the BOINC team) are already aware of that problem. Do you use Windows Vista or Windows 7 and run BOINC as a service?

                                                                                                                            Thanks,
                                                                                                                            Oliver

                                                                                                                            |MatMan|
                                                                                                                            Send message
                                                                                                                            Joined: Jan 22 05
                                                                                                                            Posts: 11
                                                                                                                            Credit: 9,397,158
                                                                                                                            RAC: 12,036
                                                                                                                            Message 100835 - Posted 30 Nov 2009 18:03:23 UTC

                                                                                                                              As the GPU load is so low how about allowing two WUs to be processed on one GPU at the same time?
                                                                                                                              ____________

                                                                                                                              Profile Paul D. Buck
                                                                                                                              Send message
                                                                                                                              Joined: Jan 17 05
                                                                                                                              Posts: 754
                                                                                                                              Credit: 4,075,867
                                                                                                                              RAC: 10,266
                                                                                                                              Message 100837 - Posted 30 Nov 2009 19:05:45 UTC - in response to Message 100733.


                                                                                                                                I find that very bad that it needs a 100% core and that you set cuda on Enable as default.


                                                                                                                                I understand and we tried not to enable the CUDA app by default. Unfortunately that would have involved a change in the BOINC core client code which is not under our direct control. Please note that this is one of the reasons why we set quite a few minimum requirements. This way the number of volunteers who receive CUDA work is as limited as possible.

                                                                                                                                Of course you could have changed the site adding the option to not do CUDA, made an announcement, given us a week to make our choices and then made the switch. That would have been far more friendly.

                                                                                                                                WRT the efficiency of the current implementation: we are working on a number of improvements. The CPU part of the radio pulsar search received quite a few changes that will not only benefit the CPU-only application but also the CUDA version, thereby moving the computational ratio towards the GPU. These changes will be released as a new application called "ABP2" - probably in the next 1-2 weeks. In parallel to that we are currently working hard to move the remaining CPU part of the CUDA version more or less completely to the GPU.

                                                                                                                                Please note that even today the CUDA app wouldn't actually require a full CPU. However, as soon as you tell the client you use less than 100% it doesn't renice the process (reduce it's priority) anymore. From our point of view it's better to have the process claiming one CPU at the lowest priority than using, say, 60% at normal priority.

                                                                                                                                I am not sure how the other projects are doing this but GPU Grid never posed that kind of issue when I have run it ... as in causing problems with lag and load ... but fundamentally, you are "wasting" resources for a BOINC bug which I have not seen being discussed on the mailing lists as a problem (then again I may have missed it, in which case apologies), so, why no effort to get the real problem fixed? If this is the issue then it is going to affect all GPU projects not just yours...

                                                                                                                                And, in running MW, Collatz and GPU Grid though I have seen higher loading and some lag (acceptable to me because I don't do anything but BOINC on my windows systems) it still does not seem to be a viable trade-off for the low performance gain.

                                                                                                                                At any rate, I look forward to the changes ...

                                                                                                                                Jim Kleine
                                                                                                                                Send message
                                                                                                                                Joined: Jan 25 06
                                                                                                                                Posts: 3
                                                                                                                                Credit: 1,053,126
                                                                                                                                RAC: 0
                                                                                                                                Message 100847 - Posted 1 Dec 2009 0:46:07 UTC - in response to Message 100832.

                                                                                                                                  Hi Jim,

                                                                                                                                  We (and the BOINC team) are already aware of that problem. Do you use Windows Vista or Windows 7 and run BOINC as a service?

                                                                                                                                  Thanks,
                                                                                                                                  Oliver


                                                                                                                                  Hi Oliver,

                                                                                                                                  This host is running Windows XP x64 edition. BOINC is running in the standard (boinc_master/boinc_project) service configuration.

                                                                                                                                  I'm quite prepared to invest some testing time if it helps to advance this problem. In the meantime, I have unchecked the GPU units account preference.

                                                                                                                                  Regards,
                                                                                                                                  Jim

                                                                                                                                  ____________

                                                                                                                                  robertmiles
                                                                                                                                  Send message
                                                                                                                                  Joined: Oct 8 09
                                                                                                                                  Posts: 108
                                                                                                                                  Credit: 2,420,922
                                                                                                                                  RAC: 2,861
                                                                                                                                  Message 100855 - Posted 1 Dec 2009 6:44:41 UTC - in response to Message 100794.

                                                                                                                                    Hello,

                                                                                                                                    One of my computers is a laptop using a NVIDIA 9600M GS.
                                                                                                                                    E@H refuse to upload Cuda WU on this laptop because the display driver 190.38 is requested.
                                                                                                                                    After checking on NVIDIA web site, it appears that this driver 190.38 is not for laptops or M graphics cards series. The latest display driver available for GPU M series is the 186.81.

                                                                                                                                    So if you're using a laptop with NVIDIA GPU, don't waste your time triying to run E@H cuda WU. it is not possible right now.

                                                                                                                                    Lionel.


                                                                                                                                    I have a similar problem, but with a G 105 M for which the latest driver available is 186.44. That means it's idle for now because no BOINC project I'm interested in can use it.

                                                                                                                                    robertmiles
                                                                                                                                    Send message
                                                                                                                                    Joined: Oct 8 09
                                                                                                                                    Posts: 108
                                                                                                                                    Credit: 2,420,922
                                                                                                                                    RAC: 2,861
                                                                                                                                    Message 100856 - Posted 1 Dec 2009 6:53:07 UTC - in response to Message 100737.


                                                                                                                                      1. Are 450 MB really necessary? I only have 256 MB...


                                                                                                                                      Yes, we really need that much memory.


                                                                                                                                      2. Why CUDA 2.2 is not enough? It is the Laptop version of these drivers. Therefore no Laptop can use the GPU or am I wrong?


                                                                                                                                      I think you are wrong as there are no "laptop" versions AFAIK. You may just download the latest driver for your operating system. CUDA 2.3 is superior to 2.2 in a number of details...

                                                                                                                                      Oliver


                                                                                                                                      The Nvidia site says that they don't have a 190.* driver suitable for my laptop's G 105 M. I suspect that it needs some power-saving features that 190.38 does not support. At least it has 512 MB memory.

                                                                                                                                      Ver Greeneyes
                                                                                                                                      Send message
                                                                                                                                      Joined: Mar 26 09
                                                                                                                                      Posts: 139
                                                                                                                                      Credit: 4,088,916
                                                                                                                                      RAC: 0
                                                                                                                                      Message 100858 - Posted 1 Dec 2009 7:00:39 UTC - in response to Message 100856.

                                                                                                                                        Last modified: 1 Dec 2009 7:02:21 UTC

                                                                                                                                        Instructions on getting the latest known official driver for your laptop card.

                                                                                                                                        This website also offers a modified inf for all the latest drivers which make them work on all laptop cards.

                                                                                                                                        robertmiles
                                                                                                                                        Send message
                                                                                                                                        Joined: Oct 8 09
                                                                                                                                        Posts: 108
                                                                                                                                        Credit: 2,420,922
                                                                                                                                        RAC: 2,861
                                                                                                                                        Message 100860 - Posted 1 Dec 2009 7:14:08 UTC

                                                                                                                                          I assume that you've seen the various messages on GPUGRID saying that at least one fft instruction tends to work on some GTX260 cards and not on others. You might want to ask them just which one, so you can think about making a separate version of the CUDA code for GTX260 cards, possibly after you make one for CUDA 2.2 with compute capability 1.2 (all that's now available on many laptops, including mine).

                                                                                                                                          Oliver Bock
                                                                                                                                          Forum moderator
                                                                                                                                          Project administrator
                                                                                                                                          Project developer
                                                                                                                                          Send message
                                                                                                                                          Joined: Sep 4 07
                                                                                                                                          Posts: 359
                                                                                                                                          Credit: 20,915,781
                                                                                                                                          RAC: 1,674
                                                                                                                                          Message 100869 - Posted 1 Dec 2009 15:32:08 UTC - in response to Message 100837.


                                                                                                                                            Of course you could have changed the site adding the option to not do CUDA, made an announcement, given us a week to make our choices and then made the switch. That would have been far more friendly.

                                                                                                                                            Right, we also considered that approach. However, the BOINC web preferences implementation only shows the opt-out check boxes for deployed/enabled application types (i.e. NVIDIA GPU).


                                                                                                                                            but fundamentally, you are "wasting" resources for a BOINC bug which I have not seen being discussed on the mailing lists as a problem (then again I may have missed it, in which case apologies), so, why no effort to get the real problem fixed? If this is the issue then it is going to affect all GPU projects not just yours...

                                                                                                                                            The usual question, bug or feature...


                                                                                                                                            At any rate, I look forward to the changes ...

                                                                                                                                            Thanks, so do we.


                                                                                                                                            Best,
                                                                                                                                            Oliver

                                                                                                                                            Profile Paul D. Buck
                                                                                                                                            Send message
                                                                                                                                            Joined: Jan 17 05
                                                                                                                                            Posts: 754
                                                                                                                                            Credit: 4,075,867
                                                                                                                                            RAC: 10,266
                                                                                                                                            Message 100892 - Posted 2 Dec 2009 10:52:38 UTC

                                                                                                                                              I posted this:

                                                                                                                                              On Einstein we have had a mini-debate about the transition of the CUDA application mostly because of the heavy use of the CPU along with the GPU meaning that throughput is going to suffer. Many of us would have wished to opt out of running these tasks but were caught by the surprise of the migration. One day no CUDA, next day it was live.

                                                                                                                                              This is part of an exchange:

                                                                                                                                              Of course you could have changed the site adding the option to not do CUDA, made an announcement, given us a week to make our choices and then made the switch. That would have been far more friendly.

                                                                                                                                              Right, we also considered that approach. However, the BOINC web preferences implementation only shows the opt-out check boxes for deployed/enabled application types (i.e. NVIDIA GPU).

                                                                                                                                              The point here is that the current server side code does not allow a transition time where the project can announce the change to have a CUDA/GPU application and allow for the users to select their opt-in or opt-out choice before the application is fielded. There should be a way that the settings can be made manifest before going live.

                                                                                                                                              Profile Rabinovitch
                                                                                                                                              Send message
                                                                                                                                              Joined: Oct 2 07
                                                                                                                                              Posts: 13
                                                                                                                                              Credit: 9,903,975
                                                                                                                                              RAC: 14,399
                                                                                                                                              Message 100924 - Posted 3 Dec 2009 20:11:33 UTC

                                                                                                                                                Last modified: 3 Dec 2009 20:25:05 UTC

                                                                                                                                                Hi all!

                                                                                                                                                Which apps should I uncheck in "Run only the selected applications" section of preferences edition page to receive only cuda-related tasks, and stop to receive pure-CPU WUs?

                                                                                                                                                btw, somewhy I can't uncheck Hierarchical S5 all-sky GW search #5 and #6....
                                                                                                                                                ____________

                                                                                                                                                Michael Goetz
                                                                                                                                                Avatar
                                                                                                                                                Send message
                                                                                                                                                Joined: Feb 11 05
                                                                                                                                                Posts: 15
                                                                                                                                                Credit: 150,848
                                                                                                                                                RAC: 0
                                                                                                                                                Message 100925 - Posted 3 Dec 2009 20:36:36 UTC - in response to Message 100924.

                                                                                                                                                  Hi all!
                                                                                                                                                  Which apps should I uncheck in "Run only the selected applications" section of preferences edition page to receive only cuda-related tasks, and stop to receive pure-CPU WUs?


                                                                                                                                                  The applications you choose aren't important; just leave them all checked.

                                                                                                                                                  At the top of that section, there's two boxes "Use CPU" and "Use NVIDIA GPU". Check the NVIDIA GPU box and uncheck the CPU box. As noted on the form, you'll need BOINC client 6.10+ for that to work.

                                                                                                                                                  That being said... I'd strongly recommend just the opposite -- turn off the GPU and only accept CPU work from Einstein, at least for now. At least if there's any other GPU project you want to do work on. At this time, the Einstein CUDA application hardly uses the GPU at all -- it *barely* runs any faster than the pure CPU app (about 33%), but it uses up both a full CPU core and the GPU. By turning on GPU tasks you're preventing your computer from doing other meaningful work on the GPU, not speeding up Einstein a whole lot, and still using a full CPU core to do the Einstein crunching.
                                                                                                                                                  ____________
                                                                                                                                                  My lucky number is 75898^524288+1

                                                                                                                                                  Profile Gundolf Jahn
                                                                                                                                                  Send message
                                                                                                                                                  Joined: Mar 1 05
                                                                                                                                                  Posts: 1079
                                                                                                                                                  Credit: 328,104
                                                                                                                                                  RAC: 18
                                                                                                                                                  Message 100926 - Posted 3 Dec 2009 20:37:00 UTC - in response to Message 100924.

                                                                                                                                                    Last modified: 3 Dec 2009 20:39:17 UTC

                                                                                                                                                    Hi all!

                                                                                                                                                    Which apps should I uncheck in "Run only the selected applications" section of preferences edition page to receive only cuda-related tasks, and stop to receive pure-CPU WUs?

                                                                                                                                                    You can't. You'd have to uncheck the GW tasks, and that's not possible.

                                                                                                                                                    btw, somewhy I can't uncheck Hierarchical S5 all-sky GW search #5 and #6....

                                                                                                                                                    That's a feature, not a bug ;-)

                                                                                                                                                    Gruß,
                                                                                                                                                    Gundolf
                                                                                                                                                    [edit]This time, I was too slow (again :-)[/edit]

                                                                                                                                                    K6XT
                                                                                                                                                    Send message
                                                                                                                                                    Joined: Sep 11 05
                                                                                                                                                    Posts: 7
                                                                                                                                                    Credit: 3,088,306
                                                                                                                                                    RAC: 19
                                                                                                                                                    Message 100943 - Posted 4 Dec 2009 16:18:27 UTC - in response to Message 100925.

                                                                                                                                                      [quote]Hi all!
                                                                                                                                                      [Snip]
                                                                                                                                                      That being said... I'd strongly recommend just the opposite -- turn off the GPU and only accept CPU work from Einstein, at least for now. At least if there's any other GPU project you want to do work on. At this time, the Einstein CUDA application hardly uses the GPU at all -- it *barely* runs any faster than the pure CPU app (about 33%), but it uses up both a full CPU core and the GPU. By turning on GPU tasks you're preventing your computer from doing other meaningful work on the GPU, not speeding up Einstein a whole lot, and still using a full CPU core to do the Einstein crunching.


                                                                                                                                                      Some of these comments are really, and unnecessarily, snippy (not including the above). Just turn off GPU at Einstein. After reading this thread I turned off GPU. Immediately the GPU switched back to SETI as I expected. I also had to reset the Einstein project because all I had were CUDA WU.

                                                                                                                                                      ____________


                                                                                                                                                      Profile Mike Hewson
                                                                                                                                                      Forum moderator
                                                                                                                                                      Avatar
                                                                                                                                                      Send message
                                                                                                                                                      Joined: Dec 1 05
                                                                                                                                                      Posts: 3444
                                                                                                                                                      Credit: 27,813,552
                                                                                                                                                      RAC: 5,485
                                                                                                                                                      Message 100957 - Posted 5 Dec 2009 0:10:07 UTC

                                                                                                                                                        Last modified: 5 Dec 2009 0:10:30 UTC

                                                                                                                                                        Just in case of misconception. E@H is about GW detection. The ABP is an optional sideline if you're keen. That's why there isn't a GW opt out. This is the policy of the project ( no doubt because GW detection is the project ). And no, that's not likely to be re-defined. :-)

                                                                                                                                                        Cheers, Mike.
                                                                                                                                                        ____________
                                                                                                                                                        "I have made this letter longer than usual, because I lack the time to make it short." - Blaise Pascal

                                                                                                                                                        cristipurdel
                                                                                                                                                        Send message
                                                                                                                                                        Joined: Jul 19 07
                                                                                                                                                        Posts: 17
                                                                                                                                                        Credit: 85,452
                                                                                                                                                        RAC: 0
                                                                                                                                                        Message 100966 - Posted 5 Dec 2009 9:48:30 UTC

                                                                                                                                                          Any news on ABP2 or maybe an ATI application?

                                                                                                                                                          DanNeely
                                                                                                                                                          Send message
                                                                                                                                                          Joined: Sep 4 05
                                                                                                                                                          Posts: 1072
                                                                                                                                                          Credit: 63,074,199
                                                                                                                                                          RAC: 92,058
                                                                                                                                                          Message 100990 - Posted 6 Dec 2009 6:28:51 UTC - in response to Message 100966.

                                                                                                                                                            Any news on ABP2 or maybe an ATI application?


                                                                                                                                                            An improved ABP app that does most of the calculations in single precision and which will allow offloading most of the work to the GPU is in work; ETA when it's done.

                                                                                                                                                            In the past ATI's declined to provide any assistance in developing a GPU app; so an ATI app isn't likely anytime soon.
                                                                                                                                                            ____________

                                                                                                                                                            rroonnaalldd
                                                                                                                                                            Send message
                                                                                                                                                            Joined: Dec 12 05
                                                                                                                                                            Posts: 116
                                                                                                                                                            Credit: 537,221
                                                                                                                                                            RAC: 0
                                                                                                                                                            Message 100991 - Posted 6 Dec 2009 14:26:42 UTC - in response to Message 100990.

                                                                                                                                                              In the past ATI's declined to provide any assistance in developing a GPU app; so an ATI app isn't likely anytime soon.
                                                                                                                                                              Maybe the user "gipsel" from milkyway can help in developing a GPU app.
                                                                                                                                                              ____________

                                                                                                                                                              hotze33
                                                                                                                                                              Send message
                                                                                                                                                              Joined: Nov 10 04
                                                                                                                                                              Posts: 91
                                                                                                                                                              Credit: 75,092,734
                                                                                                                                                              RAC: 12,259
                                                                                                                                                              Message 101003 - Posted 7 Dec 2009 15:31:42 UTC

                                                                                                                                                                I recently switch from the beta app to this new official one. I noticed some discontinuity between the actual used processing time and the time shown in the CPU Time column for a wu.
                                                                                                                                                                In the beta app it was roughly 4 hours real processing time and time reported
                                                                                                                                                                (149011603 ) and every wu took the same amount of time.
                                                                                                                                                                Now with the new app the real time is up to 18k s and reported times are much lower more like 13k s.
                                                                                                                                                                Also the variance is much more pronounced. I have short ABP units which only take 10.5k s and claim less credit (70) but get also 250cr like the ones which take 14k s and claim 90 cr. The beta app usually claimed 90 cr and took also 14k s.
                                                                                                                                                                Is there a new type of wu out?
                                                                                                                                                                ____________

                                                                                                                                                                Profile Gary Roberts
                                                                                                                                                                Forum moderator
                                                                                                                                                                Send message
                                                                                                                                                                Joined: Feb 9 05
                                                                                                                                                                Posts: 3003
                                                                                                                                                                Credit: 1,008,163,047
                                                                                                                                                                RAC: 2,457,858
                                                                                                                                                                Message 101007 - Posted 8 Dec 2009 2:25:23 UTC - in response to Message 101003.

                                                                                                                                                                  I recently switch from the beta app to this new official one. I noticed some discontinuity between the actual used processing time and the time shown in the CPU Time column for a wu.

                                                                                                                                                                  I don't have an NVIDIA GPU so some of the following comments are guesses about what might be happening.

                                                                                                                                                                  With modern BOINCs, I think that the value reported in the CPU time column of BOINC manager is wall clock time rather than actual CPU time. If a task was waiting for the GPU, the wall clock time would keep ticking but the CPU time wouldn't.

                                                                                                                                                                  In your tasks list, all the tasks showing as completed on Dec 6 and Dec 7 are ABP1 tasks and all seem to show use of the GPU - I didn't check every single one. As you have a quad core - 4 simultaneous tasks - and 1 GPU, I presume there may be quite a few times where some tasks are waiting for another to release the GPU. There didn't seem to be any GW tasks crunched during those 2 days.

                                                                                                                                                                  In the beta app it was roughly 4 hours real processing time and time reported
                                                                                                                                                                  (149011603 ) and every wu took the same amount of time.

                                                                                                                                                                  Since the beta test used AP, presumably your app_info.xml specified both the GW app and the APB1 app. Perhaps you had plenty of GW tasks running and perhaps the APB1 tasks weren't competing for the GPU. I didn't really spend time analysing the older parts of your task list but I did notice a lot more GW tasks there.

                                                                                                                                                                  Now with the new app the real time is up to 18k s and reported times are much lower more like 13k s.

                                                                                                                                                                  I think you can discount the 18K values as being wall clock time. The real variation seems to be largely in the 12-14K range. My own impressions are that even on CPU only crunching, there is some variability that seems to come in batches from time to time. It's only an impression as I haven't looked in any detail.

                                                                                                                                                                  Also the variance is much more pronounced. I have short ABP units which only take 10.5k s and claim less credit (70) but get also 250cr like the ones which take 14k s and claim 90 cr. The beta app usually claimed 90 cr and took also 14k s.

                                                                                                                                                                  I don't know why you got a value as low as 10.5K. Perhaps there is more variation in what the GPU use can save and perhaps that was a task with little or no GPU contention. I'm only guessing.

                                                                                                                                                                  The variation in credit claim is quite normal since the claim by the client is based on benchmarks and CPU time. The variation in true CPU time will give a variation in claim. The credits are granted server side and completely ignore the claim.

                                                                                                                                                                  Is there a new type of wu out?

                                                                                                                                                                  Not that I'm aware of.

                                                                                                                                                                  ____________
                                                                                                                                                                  Cheers,
                                                                                                                                                                  Gary.

                                                                                                                                                                  Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                                                                                  Forum moderator
                                                                                                                                                                  Project administrator
                                                                                                                                                                  Project developer
                                                                                                                                                                  Avatar
                                                                                                                                                                  Send message
                                                                                                                                                                  Joined: Aug 28 06
                                                                                                                                                                  Posts: 3225
                                                                                                                                                                  Credit: 72,411,621
                                                                                                                                                                  RAC: 35,282
                                                                                                                                                                  Message 101010 - Posted 8 Dec 2009 3:02:16 UTC - in response to Message 101007.

                                                                                                                                                                    Last modified: 8 Dec 2009 3:02:49 UTC


                                                                                                                                                                    I don't know why you got a value as low as 10.5K. Perhaps there is more variation in what the GPU use can save and perhaps that was a task with little or no GPU contention. I'm only guessing.


                                                                                                                                                                    Indeed, this is curious. Not all WUs are alike, there is some data dependency in the WUs' runtime, but this is the most noticeable I've ever seen. I suggest to wait if it validates and then see if the wingman also spent a less-than-average runtime on it.

                                                                                                                                                                    CU
                                                                                                                                                                    Bikeman
                                                                                                                                                                    ____________

                                                                                                                                                                    Profile Mike Hewson
                                                                                                                                                                    Forum moderator
                                                                                                                                                                    Avatar
                                                                                                                                                                    Send message
                                                                                                                                                                    Joined: Dec 1 05
                                                                                                                                                                    Posts: 3444
                                                                                                                                                                    Credit: 27,813,552
                                                                                                                                                                    RAC: 5,485
                                                                                                                                                                    Message 101012 - Posted 8 Dec 2009 3:28:01 UTC - in response to Message 101010.


                                                                                                                                                                      I don't know why you got a value as low as 10.5K. Perhaps there is more variation in what the GPU use can save and perhaps that was a task with little or no GPU contention. I'm only guessing.


                                                                                                                                                                      Indeed, this is curious. Not all WUs are alike, there is some data dependency in the WUs' runtime, but this is the most noticeable I've ever seen. I suggest to wait if it validates and then see if the wingman also spent a less-than-average runtime on it.

                                                                                                                                                                      My humble suggestion is contention yes, but what aspect of GPU? GPU thread time per se? Memory on the graphics card? GPU/CPU bandwidth? What option selected for "Suspend GPU work while computer is in use?" .....

                                                                                                                                                                      Those all vary widely. Then throw in a first person shooter and it all becomes moot. :-)

                                                                                                                                                                      Cheers, Mike.

                                                                                                                                                                      ____________
                                                                                                                                                                      "I have made this letter longer than usual, because I lack the time to make it short." - Blaise Pascal

                                                                                                                                                                      Profile Gary Roberts
                                                                                                                                                                      Forum moderator
                                                                                                                                                                      Send message
                                                                                                                                                                      Joined: Feb 9 05
                                                                                                                                                                      Posts: 3003
                                                                                                                                                                      Credit: 1,008,163,047
                                                                                                                                                                      RAC: 2,457,858
                                                                                                                                                                      Message 101017 - Posted 8 Dec 2009 4:45:49 UTC - in response to Message 101012.

                                                                                                                                                                        ... Those all vary widely. Then throw in a first person shooter and it all becomes moot. :-)

                                                                                                                                                                        It wouldn't even need to be a game, would it? I'd imagine an animated 3D screensaver kicking in while he's sleeping would have some effect ...?

                                                                                                                                                                        ____________
                                                                                                                                                                        Cheers,
                                                                                                                                                                        Gary.

                                                                                                                                                                        Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                                                                                        Forum moderator
                                                                                                                                                                        Project administrator
                                                                                                                                                                        Project developer
                                                                                                                                                                        Avatar
                                                                                                                                                                        Send message
                                                                                                                                                                        Joined: Aug 28 06
                                                                                                                                                                        Posts: 3225
                                                                                                                                                                        Credit: 72,411,621
                                                                                                                                                                        RAC: 35,282
                                                                                                                                                                        Message 101018 - Posted 8 Dec 2009 4:52:41 UTC

                                                                                                                                                                          The funny thing is that we have here an outlier (the 10k sec job) on the fast end. I can understand outliers on the slow end, but if using an ego-shooter accelerates ABP1 jobs we are sending the wrong message to kids :-).

                                                                                                                                                                          CU
                                                                                                                                                                          Bikeman
                                                                                                                                                                          ____________

                                                                                                                                                                          Profile Mike Hewson
                                                                                                                                                                          Forum moderator
                                                                                                                                                                          Avatar
                                                                                                                                                                          Send message
                                                                                                                                                                          Joined: Dec 1 05
                                                                                                                                                                          Posts: 3444
                                                                                                                                                                          Credit: 27,813,552
                                                                                                                                                                          RAC: 5,485
                                                                                                                                                                          Message 101021 - Posted 8 Dec 2009 5:41:38 UTC - in response to Message 101017.

                                                                                                                                                                            ... Those all vary widely. Then throw in a first person shooter and it all becomes moot. :-)

                                                                                                                                                                            It wouldn't even need to be a game, would it? I'd imagine an animated 3D screensaver kicking in while he's sleeping would have some effect ...?

                                                                                                                                                                            With my young lads - "Online Call of Duty 4" or some such - the GPU temp can go up 20+ degrees real quick, real easy. :-)

                                                                                                                                                                            I think it's polygon rate related? [ thus scene complexity and rate of change thereof ]

                                                                                                                                                                            Cheers, Mike.
                                                                                                                                                                            ____________
                                                                                                                                                                            "I have made this letter longer than usual, because I lack the time to make it short." - Blaise Pascal

                                                                                                                                                                            hotze33
                                                                                                                                                                            Send message
                                                                                                                                                                            Joined: Nov 10 04
                                                                                                                                                                            Posts: 91
                                                                                                                                                                            Credit: 75,092,734
                                                                                                                                                                            RAC: 12,259
                                                                                                                                                                            Message 101023 - Posted 8 Dec 2009 8:02:20 UTC

                                                                                                                                                                              Last modified: 8 Dec 2009 8:03:27 UTC

                                                                                                                                                                              @ Gary Roberts
                                                                                                                                                                              My system is a einstein dedicated one. It has a Q6600@3.6GHz and 2 9800GX2, so every cpu core has one gpu.
                                                                                                                                                                              One 9800GX2 is in a PCIe x16 slot and one in PCIex4. There was a difference in run time earlier. The gpu in PCIex4 took 15.5k s and the one in PCIx16 took 14.3k s. Now for the 90 cr claimed units I dont get times above 14.4k. So there is some kind of speed up.
                                                                                                                                                                              100cr claimed 15.2k s
                                                                                                                                                                              70cr claimed 10.7k s
                                                                                                                                                                              92 cr claimed 14k s
                                                                                                                                                                              67 cr claimed 10.1k s
                                                                                                                                                                              All units granted 250cr.
                                                                                                                                                                              The run times seem more consistently now. The last transition between the cuda apps gave me also some weird cpu numbers.

                                                                                                                                                                              In conclusion I have to get a better mainboard with a least two PCIex16 slots ;)
                                                                                                                                                                              ____________

                                                                                                                                                                              Profile Tom95134
                                                                                                                                                                              Send message
                                                                                                                                                                              Joined: Aug 21 07
                                                                                                                                                                              Posts: 18
                                                                                                                                                                              Credit: 1,842,773
                                                                                                                                                                              RAC: 0
                                                                                                                                                                              Message 101054 - Posted 8 Dec 2009 19:44:04 UTC - in response to Message 100824.

                                                                                                                                                                                ...And hey, don't tell me that I can leave the project if I don't like it. This is the most antisocial attitude I've heard of. So if you don't like to share the resources of this planet with others, maybe it's time for you to leave it!


                                                                                                                                                                                i agree with that !!

                                                                                                                                                                                No one told anyone to leave the project. XJR-Maniac was just quite rude without reason.

                                                                                                                                                                                If you don't want Einstein CUDA tasks, deselect them in your preferences. Nothing easier than that.

                                                                                                                                                                                Yes, I too am unsure how that deduction was made from Gary's post. :-)

                                                                                                                                                                                In any case the primary requirement for CUDA to yield significant benefit is that the problem must lend itself to massive parallelism ( ideally thousands of threads, plus other restrictions ). This is a basic reason ( plus of course issues like compiler technology ) that leads to variable success with apps.

                                                                                                                                                                                The development here at E@H is quite cautious, with a considerable user pool feeding back via beta testing. CUDA is no exception. While not always successful ( a failure outcome is within the definition of testing ), one hopes to be able to productively generalise beyond the test participants. One can opt out of CUDA if it doesn't fly well enough. In fact that is likely to be a common response for those with unsuitable hardware for optimal CUDA use. Alas as Oliver pointed out, without changing BOINC code ( not under E@H control ) then a default setting of opt-out was/is not available.

                                                                                                                                                                                Cheers, Mike.


                                                                                                                                                                                My understanding is that you can opt out of CUDA work for E@H. The problem is how do you do it and not lose downloaded WUs.

                                                                                                                                                                                Here is the approach I am taking:

                                                                                                                                                                                #1. Not allow new E@H tasks (WU) to be downloaded.

                                                                                                                                                                                #2. When all E@H tasks have been completed and uploaded then change the E@H preferences to not allow CUDA tasks.

                                                                                                                                                                                #3. Allow new E@H tasks.

                                                                                                                                                                                You may ask why I am taking these steps. Mainly because E@H does not make efficient use of the GPU. I can see this simply by watching the Elapsed Time and the "To Completion" time on the Tasks tab. SETI@Home shows a decrement of the To Completion time of about 10~15 seconds for every second of CPU time while E@H only decrements the "To Completion" value by about 1~2 seconds for every second of Elapsed time.

                                                                                                                                                                                I look forward to running E@H tasks on the GPU in the future but currently it is not the best use of my hardware. My GPU is an NVIDIA card GeForce 8800 GTS 512 and the mobo is an Intel D975XBX2 with an Intel Core2 Quad 2.40 GHz.
                                                                                                                                                                                ____________

                                                                                                                                                                                Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                                                                                                Forum moderator
                                                                                                                                                                                Project administrator
                                                                                                                                                                                Project developer
                                                                                                                                                                                Avatar
                                                                                                                                                                                Send message
                                                                                                                                                                                Joined: Aug 28 06
                                                                                                                                                                                Posts: 3225
                                                                                                                                                                                Credit: 72,411,621
                                                                                                                                                                                RAC: 35,282
                                                                                                                                                                                Message 101070 - Posted 9 Dec 2009 2:15:54 UTC - in response to Message 101054.

                                                                                                                                                                                  Last modified: 9 Dec 2009 13:45:35 UTC

                                                                                                                                                                                  Mainly because E@H does not make efficient use of the GPU. I can see this simply by watching the Elapsed Time and the "To Completion" time on the Tasks tab. SETI@Home shows a decrement of the To Completion time of about 10~15 seconds for every second of CPU time while E@H only decrements the "To Completion" value by about 1~2 seconds for every second of Elapsed time.


                                                                                                                                                                                  I'm not at all contesting your statement about the degree to which ABP1 CUDA currently uses the GPU, but you arrive at this conclusion for the wrong reason.

                                                                                                                                                                                  The rate at which the "time to completion" is diminishing is NOT a good indicator of app efficiency. When BOINC downloads a new workunit, it tries to predict the runtime of it, based on information that is embedded in the workunit itself and based on statistics BOINC gathered during earlier WUs for the same project.

                                                                                                                                                                                  So the rate at which the "time to completion" changes for a job in execution is just a measure of how good this runtime prediction was. If BOINC made a good guess, the rate will diminish at a rate of 1:1 and in the end the predicted runtime will turn out to be about right.

                                                                                                                                                                                  If the initial guess was way too high, BOINC will sense during the execution of the task that the progress (in % as shown in the boincmanager) is actually faster than predicted and it will correct the time to completion slowly.

                                                                                                                                                                                  If the initial guess was way too low, you will even see the "Time to completion" going UP instead of down for some time during the execution of the task.

                                                                                                                                                                                  CU
                                                                                                                                                                                  Bikeman
                                                                                                                                                                                  ____________

                                                                                                                                                                                  Profile Gary Roberts
                                                                                                                                                                                  Forum moderator
                                                                                                                                                                                  Send message
                                                                                                                                                                                  Joined: Feb 9 05
                                                                                                                                                                                  Posts: 3003
                                                                                                                                                                                  Credit: 1,008,163,047
                                                                                                                                                                                  RAC: 2,457,858
                                                                                                                                                                                  Message 101082 - Posted 9 Dec 2009 9:51:40 UTC - in response to Message 101023.

                                                                                                                                                                                    @ Gary Roberts
                                                                                                                                                                                    My system is a einstein dedicated one. It has a Q6600@3.6GHz and 2 9800GX2, so every cpu core has one gpu.


                                                                                                                                                                                    As your computers are hidden, it wasn't possible for me to see what hardware you had. You didn't comment on how many cards or that they were duals so I just made the (wrong) assumption of a single GPU. I did warn that I don't have any NVIDIA experience and that I was just guessing :-).

                                                                                                                                                                                    So just ignore my comments about possible contention.


                                                                                                                                                                                    ____________
                                                                                                                                                                                    Cheers,
                                                                                                                                                                                    Gary.

                                                                                                                                                                                    Profile Tom95134
                                                                                                                                                                                    Send message
                                                                                                                                                                                    Joined: Aug 21 07
                                                                                                                                                                                    Posts: 18
                                                                                                                                                                                    Credit: 1,842,773
                                                                                                                                                                                    RAC: 0
                                                                                                                                                                                    Message 101094 - Posted 9 Dec 2009 19:49:06 UTC - in response to Message 101070.

                                                                                                                                                                                      I'm not at all contesting your statement about the degree to which ABP1 CUDA currently uses the GPU, but you arrive at this conclusion for the wrong reason.

                                                                                                                                                                                      The rate at which the "time to completion" is diminishing is NOT a good indicator of app efficiency. When BOINC downloads a new workunit, it tries to predict the runtime of it, based on information that is embedded in the workunit itself and based on statistics BOINC gathered during earlier WUs for the same project.

                                                                                                                                                                                      So the rate at which the "time to completion" changes for a job in execution is just a measure of how good this runtime prediction was. If BOINC made a good guess, the rate will diminish at a rate of 1:1 and in the end the predicted runtime will turn out to be about right.

                                                                                                                                                                                      If the initial guess was way too high, BOINC will sense during the execution of the task that the progress (in % as shown in the boincmanager) is actually faster than predicted and it will correct the time to completion slowly.

                                                                                                                                                                                      If the initial guess was way too low, you will even see the "Time to completion" going UP instead of down for some time during the execution of the task.

                                                                                                                                                                                      CU
                                                                                                                                                                                      Bikeman


                                                                                                                                                                                      Bikeman,

                                                                                                                                                                                      Thanks for the information. You may be right on the reason for slow decrementing of the To Completion time. It just seemed to me a good indication since the BOINC client is the same for both E@H and S@H. In any case, I'm flushing all the E@H tasks and will "disallow" GPU tasks for E@H until this gets resolved. I also hope by then the size of the E@H CUDA/GPU tasks will also be less than 3 hours (a limit which appears to be what the people at S@H are using) so E@H plays better with others on the GPU.

                                                                                                                                                                                      Tom

                                                                                                                                                                                      Oliver Bock
                                                                                                                                                                                      Forum moderator
                                                                                                                                                                                      Project administrator
                                                                                                                                                                                      Project developer
                                                                                                                                                                                      Send message
                                                                                                                                                                                      Joined: Sep 4 07
                                                                                                                                                                                      Posts: 359
                                                                                                                                                                                      Credit: 20,915,781
                                                                                                                                                                                      RAC: 1,674
                                                                                                                                                                                      Message 101108 - Posted 10 Dec 2009 10:36:25 UTC - in response to Message 101094.

                                                                                                                                                                                        Last modified: 10 Dec 2009 12:27:22 UTC

                                                                                                                                                                                        I also hope by then the size of the E@H CUDA/GPU tasks will also be less than 3 hours (a limit which appears to be what the people at S@H are using) so E@H plays better with others on the GPU.


                                                                                                                                                                                        Our tests indicate that the upcoming ABP2 GPU tasks typically take ~0.6 hours per WU. This is still "just" a factor of 2-3 faster than the ABP2 CPU version, but after the ABP2 release we are going to concentrate on improving the GPU code.


                                                                                                                                                                                        Cheers,
                                                                                                                                                                                        Oliver

                                                                                                                                                                                        ML1
                                                                                                                                                                                        Send message
                                                                                                                                                                                        Joined: Feb 20 05
                                                                                                                                                                                        Posts: 320
                                                                                                                                                                                        Credit: 17,155,851
                                                                                                                                                                                        RAC: 10,895
                                                                                                                                                                                        Message 101110 - Posted 10 Dec 2009 12:38:27 UTC - in response to Message 101108.

                                                                                                                                                                                          Last modified: 10 Dec 2009 12:39:12 UTC

                                                                                                                                                                                          Our tests indicate that the upcoming ABP2 GPU tasks typically take ~0.6 hours per WU. This is still "just" a factor of 2-3 faster than the ABP2 CPU version, but after the ABP2 release we are going to concentrate on improving the GPU code

                                                                                                                                                                                          Looking good.

                                                                                                                                                                                          Is this going to be an automated rollout or do we need to start playing with the app_info file?


                                                                                                                                                                                          Just a thought: Is there anything 'interesting' to see with visualising the CUDA code in operation? Or would it all just be a noisy mass of coloured pixels?

                                                                                                                                                                                          Happy crunchin',
                                                                                                                                                                                          Martin
                                                                                                                                                                                          ____________
                                                                                                                                                                                          Powered by Mandriva Linux A user friendly OS!
                                                                                                                                                                                          See the Boinc HELP Wiki

                                                                                                                                                                                          [B^S] Elphidieus
                                                                                                                                                                                          Send message
                                                                                                                                                                                          Joined: Feb 20 05
                                                                                                                                                                                          Posts: 220
                                                                                                                                                                                          Credit: 12,347,408
                                                                                                                                                                                          RAC: 0
                                                                                                                                                                                          Message 101111 - Posted 10 Dec 2009 12:40:45 UTC - in response to Message 101108.



                                                                                                                                                                                            Our tests indicate that the upcoming ABP2 GPU tasks typically take ~0.6 hours per WU. This is still "just" a factor of 2-3 faster than the ABP2 CPU version, but after the ABP2 release we are going to concentrate on improving the GPU code.


                                                                                                                                                                                            Cheers,
                                                                                                                                                                                            Oliver


                                                                                                                                                                                            Sounds like Macs and ATI-equipped PCs are on the losing side...
                                                                                                                                                                                            ____________

                                                                                                                                                                                            cristipurdel
                                                                                                                                                                                            Send message
                                                                                                                                                                                            Joined: Jul 19 07
                                                                                                                                                                                            Posts: 17
                                                                                                                                                                                            Credit: 85,452
                                                                                                                                                                                            RAC: 0
                                                                                                                                                                                            Message 101115 - Posted 10 Dec 2009 15:03:42 UTC - in response to Message 101108.


                                                                                                                                                                                              Our tests indicate that the upcoming ABP2 GPU tasks typically take ~0.6 hours per WU. This is still "just" a factor of 2-3 faster than the ABP2 CPU version, but after the ABP2 release we are going to concentrate on improving the GPU code.


                                                                                                                                                                                              Cheers,
                                                                                                                                                                                              Oliver

                                                                                                                                                                                              Which GPU card, and how much did it took with the ABP1 GPU?

                                                                                                                                                                                              Oliver Bock
                                                                                                                                                                                              Forum moderator
                                                                                                                                                                                              Project administrator
                                                                                                                                                                                              Project developer
                                                                                                                                                                                              Send message
                                                                                                                                                                                              Joined: Sep 4 07
                                                                                                                                                                                              Posts: 359
                                                                                                                                                                                              Credit: 20,915,781
                                                                                                                                                                                              RAC: 1,674
                                                                                                                                                                                              Message 101116 - Posted 10 Dec 2009 15:09:47 UTC - in response to Message 101110.


                                                                                                                                                                                                Is this going to be an automated rollout or do we need to start playing with the app_info file?

                                                                                                                                                                                                We're currently running ABP2 workunits on our test project server. These tests typically cover the CPU/GPU versions of all supported platforms. We haven't yet decided whether a public beta test (using app_info.xml) will be necessary or not. If we choose not to run a public beta test, ABP2 apps and workunits are likely to be rolled out very soon (1-2 weeks from now). As soon as ABP2 is running smoothly we'll stop publishing ABP1 work.


                                                                                                                                                                                                Just a thought: Is there anything 'interesting' to see with visualising the CUDA code in operation? Or would it all just be a noisy mass of coloured pixels?

                                                                                                                                                                                                No, it does the same as the CPU version. However, the displayed search parameters and the power spectrum will be updated more frequently. Maybe we'll be adding a few GPU-specific details in the future - for the time being we focus on performance though. IMHO, the screensaver should be disabled when running GPU workunits anyway, unless you have a dedicated GPU for crunching...

                                                                                                                                                                                                Best,
                                                                                                                                                                                                Oliver

                                                                                                                                                                                                Oliver Bock
                                                                                                                                                                                                Forum moderator
                                                                                                                                                                                                Project administrator
                                                                                                                                                                                                Project developer
                                                                                                                                                                                                Send message
                                                                                                                                                                                                Joined: Sep 4 07
                                                                                                                                                                                                Posts: 359
                                                                                                                                                                                                Credit: 20,915,781
                                                                                                                                                                                                RAC: 1,674
                                                                                                                                                                                                Message 101117 - Posted 10 Dec 2009 15:40:07 UTC - in response to Message 101115.


                                                                                                                                                                                                  Our tests indicate that the upcoming ABP2 GPU tasks typically take ~0.6 hours per WU. This is still "just" a factor of 2-3 faster than the ABP2 CPU version, but after the ABP2 release we are going to concentrate on improving the GPU code.

                                                                                                                                                                                                  Which GPU card, and how much did it took with the ABP1 GPU?


                                                                                                                                                                                                  Sorry, forgot to mention that: we use GTX200 series cards (GTX 285, GTX 295, Tesla C1060). The improvement in comparison to ABP1 stems from changes not related to the GPU, so ABP2 CPU tasks will also finish much more quickly than ABP1 CPU tasks. The CPU/GPU ratio should still be slightly improved however.

                                                                                                                                                                                                  Oliver

                                                                                                                                                                                                  Profile Stranger7777
                                                                                                                                                                                                  Avatar
                                                                                                                                                                                                  Send message
                                                                                                                                                                                                  Joined: Mar 17 05
                                                                                                                                                                                                  Posts: 321
                                                                                                                                                                                                  Credit: 92,844,028
                                                                                                                                                                                                  RAC: 76,959
                                                                                                                                                                                                  Message 101137 - Posted 11 Dec 2009 19:57:00 UTC

                                                                                                                                                                                                    Tonight I have a great time for a E@H pause reaching daily quota for 4 results for Q9400 8[ ].
                                                                                                                                                                                                    Yes, I did make a mistake when I deleted app_info.xml before the cache has drained out. But!
                                                                                                                                                                                                    Suddendly after stopping and upgrading BOINC to the latest 6.10.18, I received 4 last S5R5 tasks and unfortunelly didn't find any *.exe for it neither in my BOINC directory nor in E@H (the link for S5R5 tasks obviously had been removed already). So my BOINC was not able to download any S5R exe and I aborted the download for this file. Now I'm in "reached daily quota" mode for 4 hours. And even 9800GTX suffering because of that.
                                                                                                                                                                                                    So, can somebody explain this so strange behaviour of BOINC?

                                                                                                                                                                                                    Profile Gundolf Jahn
                                                                                                                                                                                                    Send message
                                                                                                                                                                                                    Joined: Mar 1 05
                                                                                                                                                                                                    Posts: 1079
                                                                                                                                                                                                    Credit: 328,104
                                                                                                                                                                                                    RAC: 18
                                                                                                                                                                                                    Message 101141 - Posted 11 Dec 2009 22:30:12 UTC - in response to Message 101137.

                                                                                                                                                                                                      The easiest way is to reset the project after deleting app_info.xml, so that the standard application is downloaded again. There have been several threads about that topic.

                                                                                                                                                                                                      Gruß,
                                                                                                                                                                                                      Gundolf
                                                                                                                                                                                                      ____________
                                                                                                                                                                                                      Computer sind nicht alles im Leben. (Kleiner Scherz)

                                                                                                                                                                                                      Profile Stranger7777
                                                                                                                                                                                                      Avatar
                                                                                                                                                                                                      Send message
                                                                                                                                                                                                      Joined: Mar 17 05
                                                                                                                                                                                                      Posts: 321
                                                                                                                                                                                                      Credit: 92,844,028
                                                                                                                                                                                                      RAC: 76,959
                                                                                                                                                                                                      Message 101142 - Posted 11 Dec 2009 22:52:17 UTC

                                                                                                                                                                                                        Yes, I did it. But it doesn't help, because it tries to download S5R5....x86.exe that doesn't exist anymore :)

                                                                                                                                                                                                        Profile Gary Roberts
                                                                                                                                                                                                        Forum moderator
                                                                                                                                                                                                        Send message
                                                                                                                                                                                                        Joined: Feb 9 05
                                                                                                                                                                                                        Posts: 3003
                                                                                                                                                                                                        Credit: 1,008,163,047
                                                                                                                                                                                                        RAC: 2,457,858
                                                                                                                                                                                                        Message 101145 - Posted 12 Dec 2009 0:02:10 UTC - in response to Message 101137.

                                                                                                                                                                                                          ... Suddendly after stopping and upgrading BOINC to the latest 6.10.18, I received 4 last S5R5 tasks and unfortunelly didn't find any *.exe for it neither in my BOINC directory nor in E@H (the link for S5R5 tasks obviously had been removed already).

                                                                                                                                                                                                          The S5R5 app is still readily available. For example, here is the 'switcher' app but the _0, _1 and _2 apps are also there as well.

                                                                                                                                                                                                          So my BOINC was not able to download any S5R exe and I aborted the download for this file.

                                                                                                                                                                                                          The mistake you made was probably that you deleted the app files at the same time as you deleted the app_info.xml file. When BOINC starts, it already has previously copied the contents of app_info.xml into the state file so if you have just deleted app_info.xml, it still has the previous contents (not yet removed but will be auto-removed later) in the state file so it still expects to find the old apps and it gets upset if it doesn't find them. Under the AP mechanism, there is no legally available url for BOINC to use since the files are supposed to be 'user supplied'.

                                                                                                                                                                                                          There are several ways to terminate AP participation but I find the easiest is to just delete app_info.xml and leave the apps for BOINC itself to delete (which it will do at a later stage) when it fully realises that it is no longer running under AP. Jord (Ageless) promotes the (perhaps more user friendly) procedure of resetting the project as well (which removes the AP stuff from the state file immediately) but this can entail a lot of unnecessary re-downloading.

                                                                                                                                                                                                          Now I'm in "reached daily quota" mode for 4 hours. And even 9800GTX suffering because of that.
                                                                                                                                                                                                          So, can somebody explain this so strange behaviour of BOINC?

                                                                                                                                                                                                          I believe that the 4 S5R5 tasks would have been marked as 'download error' when the app download failed. Let me know if you still have those tasks in your list as it would be possible to 'resurrect' them as long as they hadn't been reported. If you have reported the results (which you probably have by now) it will be too late. I have done this before but the memory is a bit vague :-).

                                                                                                                                                                                                          ____________
                                                                                                                                                                                                          Cheers,
                                                                                                                                                                                                          Gary.

                                                                                                                                                                                                          Profile Tom95134
                                                                                                                                                                                                          Send message
                                                                                                                                                                                                          Joined: Aug 21 07
                                                                                                                                                                                                          Posts: 18
                                                                                                                                                                                                          Credit: 1,842,773
                                                                                                                                                                                                          RAC: 0
                                                                                                                                                                                                          Message 101156 - Posted 12 Dec 2009 20:58:59 UTC - in response to Message 101108.

                                                                                                                                                                                                            I also hope by then the size of the E@H CUDA/GPU tasks will also be less than 3 hours (a limit which appears to be what the people at S@H are using) so E@H plays better with others on the GPU.


                                                                                                                                                                                                            Our tests indicate that the upcoming ABP2 GPU tasks typically take ~0.6 hours per WU. This is still "just" a factor of 2-3 faster than the ABP2 CPU version, but after the ABP2 release we are going to concentrate on improving the GPU code.


                                                                                                                                                                                                            Cheers,
                                                                                                                                                                                                            Oliver


                                                                                                                                                                                                            Thank you for the update. When do you expect to start making ABP2 tasks available and is there a way to identify them when they appear in the task list?

                                                                                                                                                                                                            Thanks.
                                                                                                                                                                                                            Tom

                                                                                                                                                                                                            SeersantLoom
                                                                                                                                                                                                            Send message
                                                                                                                                                                                                            Joined: Nov 11 04
                                                                                                                                                                                                            Posts: 3
                                                                                                                                                                                                            Credit: 7,779,653
                                                                                                                                                                                                            RAC: 4,337
                                                                                                                                                                                                            Message 101186 - Posted 14 Dec 2009 10:15:37 UTC - in response to Message 100691.

                                                                                                                                                                                                              We have finally begun to automatically deliver CUDA work & applications (plan class "ABP1cuda23") to machines that satisfy the following requirements:

                                                                                                                                                                                                              - enabled NVIDIA GPU work in Einstein@home preferences
                                                                                                                                                                                                              - NVidia GPU with at least 450MB of free memory
                                                                                                                                                                                                              - Display Driver version 190.38 (&up), i.e. CUDA 2.3 capability
                                                                                                                                                                                                              - BOINC Core Client version 6.10 (&up)

                                                                                                                                                                                                              CUDA Beta App testers should drain their work cache and switch back to the normal project work.

                                                                                                                                                                                                              BM


                                                                                                                                                                                                              My box has all noted requirements: 8800GT card w/ 512MB RAM (don't know how much of it is exactly free, though, resolution is set to 1920x1200 24bpp), nvidia-drivers-190.42-r3, CUDA 2.3 and BOINC 6.10.18.

                                                                                                                                                                                                              System and OS: AMD Athlon(tm) 64 X2 Dual Core Processor 6000+, 8GB RAM and runs Gentoo Linux.

                                                                                                                                                                                                              Problem: ABPIcuda23 won't run on it. It stops right in the beginning of computation with error: "[ERROR] Error creating CUDA FFT plan (error code: 2)".

                                                                                                                                                                                                              All the E@H CUDA tasks this client was assigned to failed with the same message.

                                                                                                                                                                                                              One example: http://einstein.phys.uwm.edu/result.php?resultid=150643235
                                                                                                                                                                                                              ____________

                                                                                                                                                                                                              Profile Gundolf Jahn
                                                                                                                                                                                                              Send message
                                                                                                                                                                                                              Joined: Mar 1 05
                                                                                                                                                                                                              Posts: 1079
                                                                                                                                                                                                              Credit: 328,104
                                                                                                                                                                                                              RAC: 18
                                                                                                                                                                                                              Message 101187 - Posted 14 Dec 2009 11:13:16 UTC - in response to Message 101186.

                                                                                                                                                                                                                My box has all noted requirements: 8800GT card w/ 512MB RAM (don't know how much of it is exactly free, though, resolution is set to 1920x1200 24bpp), nvidia-drivers-190.42-r3, CUDA 2.3 and BOINC 6.10.18.

                                                                                                                                                                                                                Could you reduce the resolution, just to check if it's the culprit?

                                                                                                                                                                                                                If it isn't or you don't want to leave the resolution at low, I think that you'll have to disallow CUDA processing for Einstein (Einstein@Home preferences, Use NVIDIA GPU if present?), unless someone else knows another reason for the errors.

                                                                                                                                                                                                                To the developers:
                                                                                                                                                                                                                It should be possible to give informational output in stdout about total/remaining GPU memory (SETI could do it at least ;-).

                                                                                                                                                                                                                Perhaps (not so easily done:-) a "fallback to CPU mode" could be implemented when insufficient memory is detected.

                                                                                                                                                                                                                Gruß,
                                                                                                                                                                                                                Gundolf

                                                                                                                                                                                                                Profile Gary Roberts
                                                                                                                                                                                                                Forum moderator
                                                                                                                                                                                                                Send message
                                                                                                                                                                                                                Joined: Feb 9 05
                                                                                                                                                                                                                Posts: 3003
                                                                                                                                                                                                                Credit: 1,008,163,047
                                                                                                                                                                                                                RAC: 2,457,858
                                                                                                                                                                                                                Message 101199 - Posted 14 Dec 2009 20:01:25 UTC - in response to Message 101186.

                                                                                                                                                                                                                  All the E@H CUDA tasks this client was assigned to failed with the same message.

                                                                                                                                                                                                                  One example: http://einstein.phys.uwm.edu/result.php?resultid=150643235

                                                                                                                                                                                                                  To make it easier for others, I've made the link clickable.

                                                                                                                                                                                                                  The full error message is

                                                                                                                                                                                                                  [19:35:55][4928][ERROR] Error creating CUDA FFT plan (error code: 2)
                                                                                                                                                                                                                  [19:35:55][4928][ERROR] Demodulation failed (error: 3)!
                                                                                                                                                                                                                  19:35:55 (4928): called boinc_finish


                                                                                                                                                                                                                  If you ckeck out this response from Oliver in the CUDA app beta test thread, to an identical message, you can guess that you are running out of memory. Be aware that your other activities may have tied up graphics memory and perhaps that memory is still unavailable when E@H tries to start a task.
                                                                                                                                                                                                                  ____________
                                                                                                                                                                                                                  Cheers,
                                                                                                                                                                                                                  Gary.

                                                                                                                                                                                                                  Cupojoe
                                                                                                                                                                                                                  Send message
                                                                                                                                                                                                                  Joined: Feb 24 05
                                                                                                                                                                                                                  Posts: 25
                                                                                                                                                                                                                  Credit: 25,047,335
                                                                                                                                                                                                                  RAC: 211
                                                                                                                                                                                                                  Message 101212 - Posted 15 Dec 2009 5:50:19 UTC

                                                                                                                                                                                                                    Last modified: 15 Dec 2009 5:51:38 UTC

                                                                                                                                                                                                                    All I know is, I've gotten way too many CUDA units (which I don't mind, though if that's all I have, and I'm using my computer, nothing gets done -- no work is done on my GPU when the computer is in use, and when my computer is not in use, only one task is processed at a time -- though I have a dual core system -- since I only have one video card). If there were a better balance of CUDA to non-CUDA tasks, einstein@home would be getting a lot more math out of my computer...

                                                                                                                                                                                                                    _badger
                                                                                                                                                                                                                    Avatar
                                                                                                                                                                                                                    Send message
                                                                                                                                                                                                                    Joined: Mar 8 05
                                                                                                                                                                                                                    Posts: 12
                                                                                                                                                                                                                    Credit: 4,623,547
                                                                                                                                                                                                                    RAC: 0
                                                                                                                                                                                                                    Message 101217 - Posted 15 Dec 2009 7:52:52 UTC - in response to Message 101199.

                                                                                                                                                                                                                      All the E@H CUDA tasks this client was assigned to failed with the same message.

                                                                                                                                                                                                                      One example: http://einstein.phys.uwm.edu/result.php?resultid=150643235

                                                                                                                                                                                                                      To make it easier for others, I've made the link clickable.

                                                                                                                                                                                                                      The full error message is

                                                                                                                                                                                                                      [19:35:55][4928][ERROR] Error creating CUDA FFT plan (error code: 2)
                                                                                                                                                                                                                      [19:35:55][4928][ERROR] Demodulation failed (error: 3)!
                                                                                                                                                                                                                      19:35:55 (4928): called boinc_finish


                                                                                                                                                                                                                      If you ckeck out this response from Oliver in the CUDA app beta test thread, to an identical message, you can guess that you are running out of memory. Be aware that your other activities may have tied up graphics memory and perhaps that memory is still unavailable when E@H tries to start a task.


                                                                                                                                                                                                                      elmet

                                                                                                                                                                                                                      I have had the same error message 'Error creating CUDA FFT plan' downloading a block of ABP1cuda23 WU's. Have processed these successfully before and since.

                                                                                                                                                                                                                      Setup (kinda) similar to yours; Ubuntu 9.10 x64 on AMD Athlon 64 X2 4400, 2GB RAM, 2 x Asus 9800GT 512MB (not in SLI mode!). I'm running at a Philips CRT at 2048x1536 resolution so this not likely your problem.

                                                                                                                                                                                                                      BIONC 6.10.17 (manual install into /home/<user>/BOINC, not from the ubuntu respository).

                                                                                                                                                                                                                      nVidia driver 190.42 (downloaded from nVidia and manually installed).

                                                                                                                                                                                                                      My problem was a kernel upgrade, the nVidia driver ceased to work properly. Re-installed (followed these instructions, all has been well since).

                                                                                                                                                                                                                      Notes to Developers:

                                                                                                                                                                                                                      • Dosen't like the GPU's in SLI mode, got 'GPU device missing' on partially completed WU's, ignored any others still in the queue.
                                                                                                                                                                                                                      • With 2 core CPU and 2 GPUs, the BOINC scheduler is now ONLY running ABP1cuda23 WU's. Any CPU WUs are being ignored. Have watched another thread with a discussion on whether the BOINC scheduler will enter panic mode and process these. I waited until < 24 hrs before deadline, chickened out, set 'no new work' for the project, suspended the GPU WU's and processed the stragglers. I have more with a deadline of 24th Dec, I'll see what happens this time.
                                                                                                                                                                                                                      • Something thats not being said enough in this forum: Cheers on the work you guys are doing. Niggles like this minor! GPU integration is a major step forward, well done!




                                                                                                                                                                                                                      Michael Goetz
                                                                                                                                                                                                                      Avatar
                                                                                                                                                                                                                      Send message
                                                                                                                                                                                                                      Joined: Feb 11 05
                                                                                                                                                                                                                      Posts: 15
                                                                                                                                                                                                                      Credit: 150,848
                                                                                                                                                                                                                      RAC: 0
                                                                                                                                                                                                                      Message 101219 - Posted 15 Dec 2009 9:08:39 UTC - in response to Message 101217.


                                                                                                                                                                                                                        * Dosen't like the GPU's in SLI mode, got 'GPU device missing' on partially completed WU's, ignored any others still in the queue.


                                                                                                                                                                                                                        That's a known bug/feature. You can't use SLI with BOINC. Actually, it may be that you can't use SLI with CUDA (I don't remember.)


                                                                                                                                                                                                                        * With 2 core CPU and 2 GPUs, the BOINC scheduler is now ONLY running ABP1cuda23 WU's. Any CPU WUs are being ignored. Have watched another thread with a discussion on whether the BOINC scheduler will enter panic mode and process these. I waited until < 24 hrs before deadline, chickened out, set 'no new work' for the project, suspended the GPU WU's and processed the stragglers. I have more with a deadline of 24th Dec, I'll see what happens this time.


                                                                                                                                                                                                                        It would have been interesting to see if it actually did let the CPU tasks go beyond deadline. I would have chickened out too, however!

                                                                                                                                                                                                                        I know that there's been a lot of work on VPU vs. GPU scheduling in the BOINC client -- or at least a lot of discussion and angst about it in the project forums. To be honest, I have no idea how well BOINC handles it, and it possibly (probably?) varies significantly depending on which version of the client you're running.
                                                                                                                                                                                                                        ____________
                                                                                                                                                                                                                        My lucky number is 75898^524288+1

                                                                                                                                                                                                                        Profile Gary Roberts
                                                                                                                                                                                                                        Forum moderator
                                                                                                                                                                                                                        Send message
                                                                                                                                                                                                                        Joined: Feb 9 05
                                                                                                                                                                                                                        Posts: 3003
                                                                                                                                                                                                                        Credit: 1,008,163,047
                                                                                                                                                                                                                        RAC: 2,457,858
                                                                                                                                                                                                                        Message 101220 - Posted 15 Dec 2009 9:11:15 UTC - in response to Message 101212.

                                                                                                                                                                                                                          You have two computers currently active on E@H. Firstly, there is a Core 2 Duo whose task list shows tasks that have used a GPU for crunching. Secondly, there is an AMD64 X2 whose task list shows both GW and APB1 tasks but isn't listed as having a suitable GPU. So I guess your following comments refer to the C2D.

                                                                                                                                                                                                                          As a general observation, both hosts have quite large caches - perhaps in the order of 6 days or so, or even more if you aren't crunching 24/7 and about double that if you are only crunching on 1 CPU as your comments seem to suggest. Your C2D has lots of aborted tasks and lots of 'client detached' tasks so it might be a good idea to lower your cache a bit if you have tasks excessive to your requirements.

                                                                                                                                                                                                                          All I know is, I've gotten way too many CUDA units (which I don't mind, though if that's all I have, and I'm using my computer, nothing gets done -- no work is done on my GPU when the computer is in use, and when my computer is not in use, only one task is processed at a time -- though I have a dual core system -- since I only have one video card).

                                                                                                                                                                                                                          Let's see if I've got this straight. Your C2D gets both types of tasks, perhaps more GPU tasks than CPU tasks and you don't really mind that because it makes up for when you can't get CPU tasks? And you only ever have one task crunching at a time? Is this what you are saying?

                                                                                                                                                                                                                          You can control most things with preferences so it should be possible to have pretty much what you want (within reason) :-). The GPU app actually takes up 1 CPU + 1 GPU while a GPU task is crunching. The GPU is only quite lightly used for the duration. You should be able to have 2 tasks crunching simultaneously if you wish. If you are not using your computer and there is still only 1 task crunching, something is most likely at fault with your preference settings.

                                                                                                                                                                                                                          If there were a better balance of CUDA to non-CUDA tasks, einstein@home would be getting a lot more math out of my computer...

                                                                                                                                                                                                                          You create that balance with your preferences so can you please go to your account page and tell us what the following settings are

                                                                                                                                                                                                                          1. Under computing preferences

                                                                                                                                                                                                                          • Suspend work while computer is in use?
                                                                                                                                                                                                                          • Suspend GPU work while computer is in use?
                                                                                                                                                                                                                          • On multiprocessors, use at most -- processors
                                                                                                                                                                                                                          • On multiprocessors, use at most --% of the processors
                                                                                                                                                                                                                          • Use at most -- percent of CPU time


                                                                                                                                                                                                                          2. Under E@H preferences


                                                                                                                                                                                                                          • Use CPU -- I imagine this is 'yes'?
                                                                                                                                                                                                                          • Use NVIDIA GPU -- I imagine this is 'yes'?
                                                                                                                                                                                                                          • Run only the selected applications -- tell us the 4 values here


                                                                                                                                                                                                                          Once you give us the answers we can take this further, thanks.

                                                                                                                                                                                                                          ____________
                                                                                                                                                                                                                          Cheers,
                                                                                                                                                                                                                          Gary.

                                                                                                                                                                                                                          hotze33
                                                                                                                                                                                                                          Send message
                                                                                                                                                                                                                          Joined: Nov 10 04
                                                                                                                                                                                                                          Posts: 91
                                                                                                                                                                                                                          Credit: 75,092,734
                                                                                                                                                                                                                          RAC: 12,259
                                                                                                                                                                                                                          Message 101222 - Posted 15 Dec 2009 11:35:11 UTC - in response to Message 101219.

                                                                                                                                                                                                                            ***It would have been interesting to see if it actually did let the CPU tasks go beyond deadline. I would have chickened out too, however!***

                                                                                                                                                                                                                            The scheduler definitive kills the units. I have seen this in the beta app before.
                                                                                                                                                                                                                            ____________

                                                                                                                                                                                                                            SeersantLoom
                                                                                                                                                                                                                            Send message
                                                                                                                                                                                                                            Joined: Nov 11 04
                                                                                                                                                                                                                            Posts: 3
                                                                                                                                                                                                                            Credit: 7,779,653
                                                                                                                                                                                                                            RAC: 4,337
                                                                                                                                                                                                                            Message 101277 - Posted 17 Dec 2009 8:47:11 UTC - in response to Message 101187.

                                                                                                                                                                                                                              My box has all noted requirements: 8800GT card w/ 512MB RAM (don't know how much of it is exactly free, though, resolution is set to 1920x1200 24bpp), nvidia-drivers-190.42-r3, CUDA 2.3 and BOINC 6.10.18.

                                                                                                                                                                                                                              Could you reduce the resolution, just to check if it's the culprit?

                                                                                                                                                                                                                              If it isn't or you don't want to leave the resolution at low, I think that you'll have to disallow CUDA processing for Einstein (Einstein@Home preferences, Use NVIDIA GPU if present?), unless someone else knows another reason for the errors.

                                                                                                                                                                                                                              To the developers:
                                                                                                                                                                                                                              It should be possible to give informational output in stdout about total/remaining GPU memory (SETI could do it at least ;-).

                                                                                                                                                                                                                              Perhaps (not so easily done:-) a "fallback to CPU mode" could be implemented when insufficient memory is detected.

                                                                                                                                                                                                                              Gruß,
                                                                                                                                                                                                                              Gundolf


                                                                                                                                                                                                                              My 24" LCD display has this "native" resolution, I don't like the thought of lowering it. CUDA packages and nvidia-drivers should be OK. GPUGRID is doing fine (even with 3D screensaver and EVE-Online running at the same time). No compiz packages in the system, guess Xorg/KDE are not using as much 3D here.

                                                                                                                                                                                                                              E@H GPU tasks are now set to disabled. Did it soon after I saw them failing one after another. SETI hasn't sent me any GPU tasks, I think the cause is same as with E@H, only difference being it detected that in advance.

                                                                                                                                                                                                                              Elmet
                                                                                                                                                                                                                              ____________

                                                                                                                                                                                                                              DanNeely
                                                                                                                                                                                                                              Send message
                                                                                                                                                                                                                              Joined: Sep 4 05
                                                                                                                                                                                                                              Posts: 1072
                                                                                                                                                                                                                              Credit: 63,074,199
                                                                                                                                                                                                                              RAC: 92,058
                                                                                                                                                                                                                              Message 101315 - Posted 19 Dec 2009 0:07:48 UTC - in response to Message 101219.


                                                                                                                                                                                                                                That's a known bug/feature. You can't use SLI with BOINC. Actually, it may be that you can't use SLI with CUDA (I don't remember.)


                                                                                                                                                                                                                                That might've been true at one point, but my SLI 260's have been running 2x collatz and before that 2x milkyway WU's for months. (Dunno If I was still doing GPUgrid when I bought the 2nd card or not.)
                                                                                                                                                                                                                                ____________

                                                                                                                                                                                                                                Oliver Bock
                                                                                                                                                                                                                                Forum moderator
                                                                                                                                                                                                                                Project administrator
                                                                                                                                                                                                                                Project developer
                                                                                                                                                                                                                                Send message
                                                                                                                                                                                                                                Joined: Sep 4 07
                                                                                                                                                                                                                                Posts: 359
                                                                                                                                                                                                                                Credit: 20,915,781
                                                                                                                                                                                                                                RAC: 1,674
                                                                                                                                                                                                                                Message 101354 - Posted 21 Dec 2009 10:52:37 UTC

                                                                                                                                                                                                                                  For your information,

                                                                                                                                                                                                                                  We proposed an additional GPU memory checking procedure before CUDA tasks are launched. Backround: today, among other things, the total device memory is used to decide whether we sent work to a client requesting GPU work or not. This is, however, not enough to ensure that there's enough memory available on the device when the task is actually launched. The new method (already added to BOINC, pending release 6.10.25) will check the available memory and compare it to the minimum requirements of a given task. If it's not sufficient the task will be deferred for five minutes and other GPU tasks might be launched in the meantime. It's up to the user to decide whether this memory limitation is temporary or permanent (e.g. caused by multi-head setup and/or screen resolution) in which case one should opt-out GPU work. This opt-out process should eventually be made more fine-grained such that this decision can be made on the application rather than the project level.


                                                                                                                                                                                                                                  Cheers,
                                                                                                                                                                                                                                  Oliver

                                                                                                                                                                                                                                  Ver Greeneyes
                                                                                                                                                                                                                                  Send message
                                                                                                                                                                                                                                  Joined: Mar 26 09
                                                                                                                                                                                                                                  Posts: 139
                                                                                                                                                                                                                                  Credit: 4,088,916
                                                                                                                                                                                                                                  RAC: 0
                                                                                                                                                                                                                                  Message 101357 - Posted 21 Dec 2009 14:40:01 UTC - in response to Message 101354.

                                                                                                                                                                                                                                    The check-in notes mentioned both the scheduler and the client needed to be updated.. so this will only help Einstein@Home when they upgrade their server-side software, right?

                                                                                                                                                                                                                                    Profile MarkJ
                                                                                                                                                                                                                                    Avatar
                                                                                                                                                                                                                                    Send message
                                                                                                                                                                                                                                    Joined: Feb 28 08
                                                                                                                                                                                                                                    Posts: 212
                                                                                                                                                                                                                                    Credit: 25,053,374
                                                                                                                                                                                                                                    RAC: 287
                                                                                                                                                                                                                                    Message 101478 - Posted 28 Dec 2009 3:59:03 UTC - in response to Message 101108.

                                                                                                                                                                                                                                      I also hope by then the size of the E@H CUDA/GPU tasks will also be less than 3 hours (a limit which appears to be what the people at S@H are using) so E@H plays better with others on the GPU.


                                                                                                                                                                                                                                      Our tests indicate that the upcoming ABP2 GPU tasks typically take ~0.6 hours per WU. This is still "just" a factor of 2-3 faster than the ABP2 CPU version, but after the ABP2 release we are going to concentrate on improving the GPU code.


                                                                                                                                                                                                                                      Cheers,
                                                                                                                                                                                                                                      Oliver


                                                                                                                                                                                                                                      You mentioned elsewhere that this one uses single precision a lot more. Would there be a speed benefit to having an x64 app?

                                                                                                                                                                                                                                      Milkyway have one for ATI/Cal.
                                                                                                                                                                                                                                      ____________
                                                                                                                                                                                                                                      BOINC blog

                                                                                                                                                                                                                                      Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                                                                                                                                                      Forum moderator
                                                                                                                                                                                                                                      Project administrator
                                                                                                                                                                                                                                      Project developer
                                                                                                                                                                                                                                      Avatar
                                                                                                                                                                                                                                      Send message
                                                                                                                                                                                                                                      Joined: Aug 28 06
                                                                                                                                                                                                                                      Posts: 3225
                                                                                                                                                                                                                                      Credit: 72,411,621
                                                                                                                                                                                                                                      RAC: 35,282
                                                                                                                                                                                                                                      Message 101480 - Posted 28 Dec 2009 8:34:32 UTC - in response to Message 101478.



                                                                                                                                                                                                                                        You mentioned elsewhere that this one uses single precision a lot more. Would there be a speed benefit to having an x64 app?

                                                                                                                                                                                                                                        Milkyway have one for ATI/Cal.


                                                                                                                                                                                                                                        I did some experiments with x64 builds (under Linux) and found no significant performance increase so far.

                                                                                                                                                                                                                                        CU
                                                                                                                                                                                                                                        Bikeman
                                                                                                                                                                                                                                        ____________

                                                                                                                                                                                                                                        Profile MarkJ
                                                                                                                                                                                                                                        Avatar
                                                                                                                                                                                                                                        Send message
                                                                                                                                                                                                                                        Joined: Feb 28 08
                                                                                                                                                                                                                                        Posts: 212
                                                                                                                                                                                                                                        Credit: 25,053,374
                                                                                                                                                                                                                                        RAC: 287
                                                                                                                                                                                                                                        Message 101492 - Posted 28 Dec 2009 23:19:04 UTC - in response to Message 101480.



                                                                                                                                                                                                                                          You mentioned elsewhere that this one uses single precision a lot more. Would there be a speed benefit to having an x64 app?

                                                                                                                                                                                                                                          Milkyway have one for ATI/Cal.


                                                                                                                                                                                                                                          I did some experiments with x64 builds (under Linux) and found no significant performance increase so far.

                                                                                                                                                                                                                                          CU
                                                                                                                                                                                                                                          Bikeman


                                                                                                                                                                                                                                          Hi Bikeman,

                                                                                                                                                                                                                                          I presume that is the CPU app you are referring to?

                                                                                                                                                                                                                                          Cheers
                                                                                                                                                                                                                                          ____________
                                                                                                                                                                                                                                          BOINC blog

                                                                                                                                                                                                                                          Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                                                                                                                                                          Forum moderator
                                                                                                                                                                                                                                          Project administrator
                                                                                                                                                                                                                                          Project developer
                                                                                                                                                                                                                                          Avatar
                                                                                                                                                                                                                                          Send message
                                                                                                                                                                                                                                          Joined: Aug 28 06
                                                                                                                                                                                                                                          Posts: 3225
                                                                                                                                                                                                                                          Credit: 72,411,621
                                                                                                                                                                                                                                          RAC: 35,282
                                                                                                                                                                                                                                          Message 101507 - Posted 29 Dec 2009 6:54:54 UTC - in response to Message 101492.


                                                                                                                                                                                                                                            Hi Bikeman,

                                                                                                                                                                                                                                            I presume that is the CPU app you are referring to?

                                                                                                                                                                                                                                            Cheers


                                                                                                                                                                                                                                            Yes. If even the CPU is not benefitting from x64 compilation, the GPU app should not benefit either.

                                                                                                                                                                                                                                            CU
                                                                                                                                                                                                                                            Bikeman

                                                                                                                                                                                                                                            ____________

                                                                                                                                                                                                                                            cristipurdel
                                                                                                                                                                                                                                            Send message
                                                                                                                                                                                                                                            Joined: Jul 19 07
                                                                                                                                                                                                                                            Posts: 17
                                                                                                                                                                                                                                            Credit: 85,452
                                                                                                                                                                                                                                            RAC: 0
                                                                                                                                                                                                                                            Message 101611 - Posted 4 Jan 2010 7:59:42 UTC

                                                                                                                                                                                                                                              Any news on ABP2 CUDA? or a 'possible' ATI client?

                                                                                                                                                                                                                                              rroonnaalldd
                                                                                                                                                                                                                                              Send message
                                                                                                                                                                                                                                              Joined: Dec 12 05
                                                                                                                                                                                                                                              Posts: 116
                                                                                                                                                                                                                                              Credit: 537,221
                                                                                                                                                                                                                                              RAC: 0
                                                                                                                                                                                                                                              Message 101649 - Posted 6 Jan 2010 12:51:51 UTC

                                                                                                                                                                                                                                                Last modified: 6 Jan 2010 12:53:37 UTC

                                                                                                                                                                                                                                                Could the inefficiencies of the actually CUDA app have to do with the high values of page faults from the cpu-app???
                                                                                                                                                                                                                                                Here a log from sysinternals process-explorer:

                                                                                                                                                                                                                                                Process_________PID___CPU___Virtual Size__Working Set__Page Faults____PF Delta__CPU Time
                                                                                                                                                                                                                                                einsteinbinary____2276_100.00_V138.348___K100.908____K39.551.450__4.990____2:11:58.562
                                                                                                                                                                                                                                                Amolqc-preRC1_5_2840_000.00_V261.496___K075.600____K18.882______0.000____1:37:35.015
                                                                                                                                                                                                                                                Okay QMC is not running, but look at the cpu times and compare the values of page faults.
                                                                                                                                                                                                                                                ____________

                                                                                                                                                                                                                                                Profile Bikeman (Heinz-Bernd Eggenstein)
                                                                                                                                                                                                                                                Forum moderator
                                                                                                                                                                                                                                                Project administrator
                                                                                                                                                                                                                                                Project developer
                                                                                                                                                                                                                                                Avatar
                                                                                                                                                                                                                                                Send message
                                                                                                                                                                                                                                                Joined: Aug 28 06
                                                                                                                                                                                                                                                Posts: 3225
                                                                                                                                                                                                                                                Credit: 72,411,621
                                                                                                                                                                                                                                                RAC: 35,282
                                                                                                                                                                                                                                                Message 101683 - Posted 7 Jan 2010 18:07:29 UTC

                                                                                                                                                                                                                                                  Discussion of ABP2 CUDA app continues here:

                                                                                                                                                                                                                                                  http://einstein.phys.uwm.edu/forum_thread.php?id=7763
                                                                                                                                                                                                                                                  ____________

                                                                                                                                                                                                                                                  Post to thread

                                                                                                                                                                                                                                                  Message boards : Cruncher's Corner : ABP1 CUDA applications


                                                                                                                                                                                                                                                  Home · Your account · Message boards

                                                                                                                                                                                                                                                  This material is based upon work supported by the National Science Foundation (NSF) under Grants PHY-1104902, PHY-1104617 and PHY-1105572 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

                                                                                                                                                                                                                                                  Copyright © 2014 Bruce Allen