ABP1 3.11 App for Windows available for Beta Test |
Message boards : Cruncher's Corner : ABP1 3.11 App for Windows available for Beta Test
| Author | Message |
|---|---|
|
A new Einstein@home ABP1 App for Windows is available for Beta Test at Beta Test Page. | |
| ID: 99835 | | |
|
The installation path on the Beta Test Page for Windows (and probably the other OS too) is still not correct (pre BOINC 6.x). See also Message 98950. | |
| ID: 99838 | | |
The installation path on the Beta Test Page for Windows (and probably the other OS too) is still not correct (pre BOINC 6.x). See also Message 98950. Thanks. Fixed. BM | |
| ID: 99842 | | |
|
Loaded and running. | |
| ID: 99843 | | |
Thanks. Fixed. Well, up to a point. That's the Windows 2000/XP default. There are others, both pre- and post-2K/XP. See The Big BOINC 6 Answer Thread. | |
| ID: 99844 | | |
Thanks. Fixed. Ok, I added a link to there in the instructions. BM | |
| ID: 99845 | | |
Is there any change on the (non-)SSE support - do you want me to try and fix the fan on the MMX? I don't think that's necessary as we didn't change things WRT SSE support. However, switching from GCC 3.4.5 to GCC 4.4 means quite a substantial change that might introduce unexpected problems... On the positive side of things: we found this version to increase performance by a factor of ~2 compared to 3.10 (current production version). As always, your mileage may vary. Cheers, Oliver | |
| ID: 99851 | | |
|
Host 1758399, Core2 Q9550, 4 GiB, Vista64 | |
| ID: 99861 | | |
|
Thanks. My rough impression is that the speed gain is larger on AMD than on Intel, at least with the Core2 architecture. | |
| ID: 99862 | | |
|
On my 9550 Quad Ver. 3.09 averages 17500 seconds, Ver. 3.11 is a little faster at 16200 average. | |
| ID: 99865 | | |
|
My E8400 needs 14500s for 3.09 and now 13500 for 3.11 on Win7 64bit. | |
| ID: 99874 | | |
|
FWIW, http://einstein.phys.uwm.edu/result.php?resultid=142544570 seems to have finished well. It took about 11.7 hours; with the previous version (3.10?), tasks took about 13 hours, as I recall. | |
| ID: 99888 | | |
|
BOINC won't download ANY E@H. APB1 or GW. | |
| ID: 99890 | | |
|
Message in Boinc says: | |
| ID: 99892 | | |
|
When you downloaded the task it was 'branded' for the 3.09 app. That will not change regardless off the actual app you use to crunch the task. New tasks should be branded for 3.11. | |
| ID: 99895 | | |
|
Speed increase on Win. XP Athlon X2 4800 is: | |
| ID: 99907 | | |
|
Intel Core 2 Quad Q9300, Windows XP SP2 task 142497586 completed and verified. Seems faster. | |
| ID: 99918 | | |
|
Athlon XP 2800+ (SSE) - WinXP - Time improved from ~55.0ksecs to ~53.2ksecs | |
| ID: 99919 | | |
|
Hi, | |
| ID: 99939 | | |
Hi, I had the same problem. I even did a detach, which pulled down 1 3.10 WU. I then reinstalled the Beta. Eventually, I removed everything from the app_info.xml file except for 3.10 and 3.11 (i.e. I got rid of GW 3.05) and several hours later it finally downloaded some WUs flagged as 3.10. (The original 3.10 WU was almost finished.) It seems to be OK for now. | |
| ID: 99942 | | |
I did everything exactly as described on the Beta Test Page to run the new ABP1 Beta app. However I got these errors: This is a message from the server and not necessarily an error. I get similar messages from time to time. The server is quite fond of sending me some work and then informing me that to get more I will need to delete app_info.xml. I just wait a minute, update the project and if I'm due more work, the server will send some and usually with the same warning repeated. The server appears to be somewhat misconfigured but it's a relatively harmless thing. When running the Beta app for 3.09 and 3.10 I did exactly the same and went without problems. What is wrong? Probably nothing much. What happens if you try updating the project? If you wish to stay with the 6.6.x series of BOINC, you would be quite wise to upgrade to the recommended version as there are problems with particular versions that may be worth avoiding. Can I delete the app_info.xml file without problems? Probably not. If you do, you will most likely trash your current cache of work unless you take some (undocumented and a bit tedious) precautions. If you don't have tasks on board, then delete app_info.xml and reset the project to clean up your state file and force the download of the stock app. However, there should be no reason why the AP mechanism shouldn't be working OK for you so it's worth persevering and giving us a bit more info with which to diagnose the situation. ____________ Cheers, Gary. | |
| ID: 99943 | | |
I did everything exactly as described on the Beta Test Page to run the new ABP1 Beta app. However I got these errors: It stopped with getting new taks last night and did not report the reday tasks automatically, but did a request for new work every 3 hours. When I looked at it this morning, I did a few manual updates with intervals of about 10 minutes with no luck of getting new work (the finished WU's where sent), therefore I posted this question. The pc is now working for my work, but this afternoon I will upgrade to 6.10.13 and see what happens. I let you know. Thanks all for the rapid response. ____________ Greetings from TJ | |
| ID: 99946 | | |
|
I get the same type of message when Boinc asks for CUDA-work, I don't have a CUDA-app in my app_info so the server won't send any. (witch is the way I want it to be) | |
| ID: 99954 | | |
|
Nice 3.11 throughput improvement on my hosts compared to 3.09. | |
| ID: 99956 | | |
|
First WU returned and validated; similar result. Typical 3.09 runs were low-to-mid 90k secs. This one took 83k (Old 1.8 GHz P4 WinXP). | |
| ID: 99983 | | |
I did everything exactly as described on the Beta Test Page to run the new ABP1 Beta app. However I got these errors: I did an upgrade to 6.6.38 and started getting messages that it was requesting GPU work and none was available. CPU work was not being requested. This morning I upgraded to 6.10.13 and it finally requested GPU and CPU work. I got a bunch of S5R6 WUs, but no ABP1. :( Note that I modified my app_info.xml file to allow S5R6 WUs. Better than nothing. | |
| ID: 100044 | | |
|
Although the speedup is not as impressive as we've seen on one of our test machines, this App isn't slower than 3.10 and apparently not causing trouble, so I made this the new "official" Windows App. Thanks for testing! | |
| ID: 100046 | | |
Although the speedup is not as impressive as we've seen on one of our test machines, this App isn't slower than 3.10 and apparently not causing trouble, so I made this the new "official" Windows App. Thanks for testing! So, time to remove the app_info.xml file? | |
| ID: 100071 | | |
Although the speedup is not as impressive as we've seen on one of our test machines, this App isn't slower than 3.10 and apparently not causing trouble, so I made this the new "official" Windows App. Thanks for testing! I did...very carefully though. Drain your cache first. After I drained my cache and reported all work I did a detach and attach for the project...probably not necessary, but clean anyway. | |
| ID: 100085 | | |
Drain your cache first. Yep, that's the recommended way. Although the app version is identical to the official one, the signature is not (there is no signature in the app_info.xml). If you just remove tha app_info.xml and restart the client, it will find the app binary doesn't match the signature, mark all results in progress as client error and start downloading the app again, this time with the correct signature. BM | |
| ID: 100096 | | |
|
What is the way do install the 3.11 now. | |
| ID: 100101 | | |
|
Thanks. Drain your cache first. | |
| ID: 100107 | | |
Drain your cache first. I've found that if I go and delete the beta app along with the app_info.xml, then shutdown and restart the client (having drained the cache first of course), that it doesn't automatically reload the right client. Instead, it fails to start since it can't find the executable!? (v 6.6.28). It seems I have to issue a reset to get it to refetch the executables (it will fetch all of them of course, not just the missing one). Is there some other way short of a reset to get it to download the missing executable? ____________ "Better is the enemy of the good." - Voltaire (should be memorized by every requirements lead) | |
| ID: 100160 | | |
|
I've found that we didn't compare this app with 3.10 CUDA app. It is important this time to understand - should we stick with 3.10 CUDA any longer or we should go further with the new app wich in fact may be faster and doesn't need any powerful graphic cards anymore. Are there any tests about this? | |
| ID: 100196 | | |
I've found that we didn't compare this app with 3.10 CUDA app. It is important this time to understand - should we stick with 3.10 CUDA any longer or we should go further with the new app wich in fact may be faster and doesn't need any powerful graphic cards anymore. Are there any tests about this? Could someone answer this question? I'm wondering if its worth it to me use the extra electricy and gpu time when I could use it for something else if its not speeding this up... | |
| ID: 100271 | | |
|
Hi! | |
| ID: 100281 | | |
I've found that we didn't compare this app with 3.10 CUDA app. It is important this time to understand - should we stick with 3.10 CUDA any longer or we should go further with the new app wich in fact may be faster and doesn't need any powerful graphic cards anymore. Are there any tests about this? Not sure Bikeman's answer is as clear as it could be ... :) The CUDA version is in test... and because the EaH project is cautious and careful in the fielding of its applications the process of getting a new version out the door can be slow and tedious. That said, people have to volunteer and run the test application on tasks so the project can find the problems that exist. In this case it is not a question of electricity or effectiveness ... it is a question of are you willing to take the risk to run an unstable application where all tasks may fail and not produce any credit at all ... oh, and require you to keep an eye on the computer ... if you don't mind THOSE costs, then running the CUDA application here is for you ... If you want stable applications and few failures, then you would want to steer clear ... As with running new alpha projects, being a beta tester can be rewarding in the sense that you know that you helped get the new stuff out the door ... but, it can take some additional time to read the forums and to check to see if any tasks failed last night ... | |
| ID: 100287 | | |
|
What if the goal is to get the most boinc points and you don't mind some extra work? I have been running the CUDA for ever since it came out a long time ago and have had 0 problems. So no use running it unless its faster in my book. I have a Q6600 @ 3ghz and a nvidia 275 GTX. Does anyone have any data on a machine like that which is faster? | |
| ID: 100615 | | |
What if the goal is to get the most boinc points and you don't mind some extra work? I have been running the CUDA for ever since it came out a long time ago and have had 0 problems. So no use running it unless its faster in my book. I have a Q6600 @ 3ghz and a nvidia 275 GTX. Does anyone have any data on a machine like that which is faster? If you're credit whoring get an ATI card and run milkyway with collatz as a backup project. IIRC the current E@H CUDA beta has a lower credit rate than GPUgrid; primarily because only one part of the calculation is done on the GPU most is still being crunched on your CPU. ____________ ![]() | |
| ID: 100622 | | |
Message boards :
Cruncher's Corner :
ABP1 3.11 App for Windows available for Beta Test