Hi!

We're near the end of the intermediate S5RI run and thus about to start the long-awaited S5R2. Some properties of S5R2 still change faster than I can type and would be hopelessly outdated when you read them, but I'd like to give you some impression of what's more or less fixed now.

The "science run #5" of the LIGO instruments, or S5 for short, gives us not only the most sensitive data, but also the largest amount of data we ever had. This is very good in terms of science, as the sensitivity of the search for continuous signals increases with "observation time", i.e. the length of the data available.

However, with our present analysis tool, the computation time needed grows to the power of six over the amount of data. It's obvious that we need something more clever to deal with a larger amount of data, so we developed a new program which we call "Hierarchical Search". The basic idea is to first scan the parameter space with a coarse "grid" and then only take a closer look at the areas that have been detected as interesting, in a "follow-up stage". The current code is designed to do only the first stage, it had not been decided yet if we will do a follow-up stage on Einstein@Home as well or e.g. on LSC clusters.

The application we'll use for this run is all new and has never been used before. The algorithm used in the old App is still a part of the new one, and other parts have also been used before, but they have never been used in the present combination, and in particular not in a distributed computing project of that scale. We expect some problems to arise from this.

One of the issues we are still working on is some "overhead", i.e. calculations that are performed for technical reasons, but don't actually contribute to the result, thus wasting computing power.

We therefore will set up S5R2 as a short, experimental run that limits the search to parts of the parameter space where the overhead is well under control. During this short run we will improve the Application in various aspects. The results will also help us tuning the parameters for the next, larger run (probably named S5R3).

With the amount of data to be analyzed, however, also grows the amount of data to be downloaded to your machines. To save bandwidth both on your and on our side, we developed a new scheme for data files. They are split into small files of about 3MB in size (that are somewhat re-combined in the App). When there is no more work available for the data files you already have on your machine, the scheduler will try to assign work to your host that minimizes the number of files you have to download, so you should in most cases get away with a new download of only two files (~6MB).

However, a complete set of files (that will be used for many tasks you get) can consist of up to 10 files, so the initial download can be more than 40MB (including the Application and other data files that need to be downloaded only once per run). This might be not suitable for dial-up / modem users anymore.

We are running out of S5RI work, and thus as usual your clients will have to download a new data file for only a few workunits anyway, so modem users: beware.

We expect to issue the first hundrets of S5R2 Workunits some time tomorrow.

At the beginning Apps will be available only for our major platforms (Windows-x86, Linux-x86, MacOS-PPC and MacOS-x86); other platforms will follow with time. I hope to get the Apps finished before the last results of S5RI are sent out.

BM

____________
BM

This sounds all very well to me, except that one part confuses me:
"At the beginning Apps will be available only for our major platforms (Windows-x86, Linux-x86, MacOS PPC and Intel); other platforms will follow with time."
Does this mean that AMD won't be supported from the beginning, that it won't be optimised for AMD, or is it something completely different you meant? Sorry if I am totally confused - I just woke up from a small nap to be honest.
____________

This sounds all very well to me, except that one part confuses me:
"At the beginning Apps will be available only for our major platforms (Windows-x86, Linux-x86, MacOS PPC and Intel); other platforms will follow with time."
Does this mean that AMD won't be supported from the beginning, that it won't be optimised for AMD, or is it something completely different you meant? Sorry if I am totally confused - I just woke up from a small nap to be honest.

Thank you for the information. Can't wait to get new work from S5R2!


regards


Constantinos

____________
Gravity increases significantly in Autumn, because apples fall in large numbers during that time!

We therefore will set up S5R2 as a short, experimental run that limits the search to parts of the parameter space where the overhead is well under control. During this short run we will improve the Application in various aspects. The results will also help us tuning the parameters for the next, larger run (probably named S5R3).

Thanks for the update Bernd.

All I can say is line them up, we'll knock them back down the church steps for you. ;-)

Also, it probably won't be quite so dead in the forums now, for a change. :-)

Alinator

But one question is left for me. Will S5R2 be as short as the actual run or even shorter?

Well, me i`m ready:) Bring `em on.

BTw one question: is any action necessary to switch for S5R2? Or will the transition happen automaticly?
____________

Also, it probably won't be quite so dead in the forums now, for a change. :-)

Alinator

This sounds very interesting, I can hardly wait for the new run, either. I hope the new app will be a bit nicer to Intel boxes ;-) since I got myself such a nice dual core. As for the chatting: Sounds good!
____________

should happen automatically unless you've installed a beta app and have an app_info.xml file. This can only happen deliberately so if you don't know what I'm talking about you're safe.
____________

Ready to crunch, Captain!

Keep us well informed about major changes to the time it takes to crunch/download new WUs, and you'll never get any complaints from those interested in the science.

____________
derekm
Team MacAddict

>> I have seen an e-mail stating that with the new application the amount of cobblestones awarded per WU will be reduced as it is believed to be granting too much credit per second compared to other projects.
If this is true then by how much will the reduction be?

I don't believe you to be granting too much myself and if you are doing this going on someone else's statistics then please do your own research before making a decision.
Also the amount granted really depends on the computer you have running. The faster the machine then of course the more credit per second is generated.

It could be the figure being distorted by all the new computing hardware used by partipating Universities (particuraly a few USA Universities and German Universities).
With all those new faster computers the volunteers with slower hardware are going to be penalised.

Thank you for your time and remember to keep smiling as it makes others wonder what you have been up to.
____________

<snip>> I don't believe you to be granting too much myself and if you are doing this going on someone else's statistics then please do your own research before making a decision.
<<snip>>

I don't know if einstien is going to change their credit rate or not. In the past most of why they've granted overly large credit levels is that instead of calibrating off of the x86-sse app that the overwelming majority of the crunchers are running they've been splitting the difference between it and the x86-nonSSE and other non x86 apps which are generally significantly slower. Meanwhile most other projects have been either calibrating to x86-sse and ignoring the handful older machines/odd balls crunching for them.
____________

Oh come on... please stop talking about credits already. The science run doesn't even have started yet, and I'm sure the project admins are going to make a sensible decision when it's needed. Let's please keep this thread about the science and the technology, that's way more interesting ;-) and also more productive.
One question to Bernd or any other dev/admin who is reading this: I've understood that dial-up users might get problems in the next run. Okay, so that doesn't mean me, I'm on fast broadband and will get sth even better soon. But what about slow computers? Will sth like a Pentium 3 or my old Celeron notebook still be able to participate in a sensible way? Or have the minimum computing power requirements also increased with the new WUs/application?
Thanks in advance
Annika
____________

Thank you for the update, Bernd. The updates are always appreciated.

I'm a physics student [BSc in physics more done than not] who has an extreme interest in general relativity. Gravitational wave astronomy gives me a huge science chubby and I can't wait for the first direct detection - this project is my way of directly contributing to the effort.
____________

Jowr, I understand how you feel, though in my case, it's more the technical side of things which I want to learn as much as possible about (naturally, since I study computer science) whereas astronomy and physics are more of a hobby for me and I don't have as much knowledge about that.
____________

Thank you for the update, Bernd. The updates are always appreciated.

I'm a physics student [BSc in physics more done than not] who has an extreme interest in general relativity. Gravitational wave astronomy gives me a huge science chubby and I can't wait for the first direct detection - this project is my way of directly contributing to the effort.

I'm not that much into assembler-programming yet, it's a bit too early (also I will try again to take more advantage of auto-vectorization this time, giving much more flexibility).

Anyway, the validator of S5R2 is one of the least tested parts, you might see credit values change or states flipping between valid and invalid of already finished tasks while we adjust it.

BTW: the first 1000 units have already been sent out, and we already discovered a problem (outdated value for estimated diskspace...).

BM

____________
BM

I'm not that much into assembler-programming yet

sorry Mike, always nice to chat with you, I hope we'll pick up later. some bugs will eat me here if I don't kill them now...

BM

____________
BM

I'm not that much into assembler-programming yet, it's a bit too early (also I will try again to take more advantage of auto-vectorization this time, giving much more flexibility).
BM

After sometime I terminated the first unit with S5R2.
I need a long time, about 12 hours (51.900 sec). So I needed about 5,4 times
more than normal. So I am a little bit disapointed that I get only 113 points.
The double of the normal units.

Well, I haven't had any luck so far, only "S5RI" WUs... hope to finish them soon and try again ;-)
____________

After sometime I terminated the first unit with S5R2.
I need a long time, about 12 hours (51.900 sec). So I needed about 5,4 times
more than normal. So I am a little bit disapointed that I get only 113 points.
The double of the normal units.

Well, I haven't had any luck so far, only "S5RI" WUs... hope to finish them soon and try again ;-)

Wow, 41,873.89 CPU seconds for a Core 2 E6600 to finish one result. Looks like some long crunch times ahead. The bright spot is that will amortize the network traffic of the big downloads.

Archae, are you serious? That will mean some gigantic crunching times on slower boxes... Luckily I've got two fast ones which should cope well. Maybe I'll switch the slower ones over to other projects and alter the resource shares on the racehorses a bit to make up for it.
____________

An update on our progress:

(1) We fixed a problem with the validator where some results where erroneously being marked as validate errors. I re-ran the fixed validator on these results so credit was granted.

(2) I've done some preliminary analysis of credit granted and determined that we were not awarding enough credit. For all S5R2 workunits WITHOUT a valid result (at the moment, all but 52) I have fixed this in the database. Credit awarded for the S5R2 WU will now be larger by a factor of 2.222. Note that in the past Einstein@Home has been giving out 30% too much credit compared with other projects (see BOINC cross-project credit comparison for details). The factor of 2.222 takes this into account.

Further updates will follow.

Note: we are still in the process of rsyncing (mirroring) our new S5 data to the mirror download servers. So the first S5R2 workunits all reference the Einstein@Home server in Wisconsin as the data source.

Cheers,
Bruce
____________

Archae, are you serious? That will mean some gigantic crunching times on slower boxes... Luckily I've got two fast ones which should cope well. Maybe I'll switch the slower ones over to other projects and alter the resource shares on the racehorses a bit to make up for it.

And what about the credits for this monster of WU?

Thanks to Bernd & Bruce for the heads-up.

Is Chilango's result (the only one I've seen linked so far) likely to be typical in terms of production run-time - i.e. 5.4 times S5RI run-time, in his estimation? (I stress production run-time, since it looks as if it still has a lot of debug output).

If so, you might like to consider whether 14 days is still the appropriate deadline: my Celeron, allowing for its current 50% share with SETI, would take about 17 days pro-rata. Not that the project should be run for the benefit of one superannuated cruncher, of course!

Thanks to Bernd & Bruce for the heads-up.

Is Chilango's result (the only one I've seen linked so far) likely to be typical in terms of production run-time - i.e. 5.4 times S5RI run-time, in his estimation? (I stress production run-time, since it looks as if it still has a lot of debug output).

If so, you might like to consider whether 14 days is still the appropriate deadline: my Celeron, allowing for its current 50% share with SETI, would take about 17 days pro-rata. Not that the project should be run for the benefit of one superannuated cruncher, of course!

Holy crap.

One of my machines takes ~150,000s to do the S5RI workunits. There are only 5 done at any one time on the status page. It tries real hard :(

Then again, it isn't as if it would be doing anything else were BOINC not on it.
____________

Our goal is run times in the range from 6 to 24 hours. Note that some optimization is expected in the future so the apps will get faster.

Cheers,
Bruce

Here's another finished one; 82931739.

7.5 hours on C2D at 3Ghz.

I've got no opinion on whether S5R2 should be made a beta, but if/when another beta project is planned, I hope E@H will consider using the method employed by malariacontrol.net; that is, a Yes/No entry in project preferences:

Run test applications?
This helps us develop applications, but may cause jobs to fail on your computer

.. I've got no opinion .... Very slick

I've got no opinion on whether S5R2 should be made a beta, but if/when another beta project is planned, I hope E@H will consider using the method employed by malariacontrol.net; that is, a Yes/No entry in project preferences:

Run test applications?
This helps us develop applications, but may cause jobs to fail on your computer

No need for app_info files or yet another project to attach to. When the test has work, you get some WUs from it, otherwise you continue to crunch the main version; user needs to do nothing else.

Very slick, but just my opinion...

Again, to Bruce, Bernd, Mike H., and anyone else who may be involved in / influential to the decision-making process...

Again, to Bruce, Bernd, Mike H., and anyone else who may be involved in / influential to the decision-making process...

Thanks for your comments. To clarify: I am a volunteer moderator, so I'm not in any decision loop outside that limited role - thank Heavens! :-)

However I'm sure the development crew ( when suitably rested!! ) will thoughtfully consider any and all feedback.

Cheers, Mike.

There was a specific reason why I said "influential to"... :-) I thought you might be influential... Don't let it go to your head... ;-)

Holy crap.

One of my machines takes ~150,000s to do the S5RI workunits. There are only 5 done at any one time on the status page. It tries real hard :(

Then again, it isn't as if it would be doing anything else were BOINC not on it.

Some quick calculations, based upon currently published results, have shown that S5R2 now needs 4,57 times more CPU time than S5R1. This means roughly above 17h on my 1.8ghz Barton and around 19h on 1.666ghz Duron. Fortunately, both are on broadbands, so there is no issue with bigger WU sizes.

Anyway, this also means that older machines are now basically severly impaired. I cannot of course, in any way compare my Barton to Cruncher`s elite machines, but it is still not that old. Still, there are plenty of >1ghz machines... and they may choke on new WU`s.

We must then all hope on some optimizations:)

I still haven`t got any R2 WU`s, then my calculations are theoretical.
____________

These apps aren't as fully optimized as the S5R1 app is so runtimes will improve somewhat. IIRC the previous app gained reduced crunch times by about a third, and that the optimizations did more for the higher end CPUs that the older ones which would narrow the spread somewhat.

Running multiple really low end PCs as crunchboxes is really inefficient. Each PC will draw something like $100+ worth of power a year @10c per kwh roughly double as much if you're forced to air condition the heat out of your house. My $700 c1D-173 laptop does ~15WU/day vs ~2 for alienators 4 antiques combined. That means that retiring his old machines and replacing them with 1 newer one would pay for itself moneywise in somehwere between 1 and 2 years with cheap electricity while doing almost 8x as much work. An otherwise bare bones DIY dualcore desktop can be had for less than half the price of my laptop and would pay for itself that much faster. About the only thing machines that old are still economically viable crunchers for is as space heaters during the winder if you have electric heat since in that case the little bit of crunching you get from them is essentially free.
____________

Except for one thing, my old timers were retired from daily use as workstations and now perform specialized custom tasks and experiments for which they are perfectly adequate, cost me nothing to aquire, and I would be hard pressed to justify spending the money to buy a 'modern' replacement.

IOW, none of my machines are running just to crunch a BOINC project. They would be running whether they are crunching or not, but still have enough horsepower to help out a project. Especially one that's smart enough to realize not everyone is going to want run right out and buy a 4 banger (or greater) with 16 GB of RAM and/or set their requirement bar so high you have to play.

In any event, when it gets to the point where I feel it's costing me more than the enjoyment I get out of it is worth, the power switch will be shut off.

Alinator

Except for one thing, my old timers were retired from daily use as workstations and now perform specialized custom tasks and experiments for which they are perfectly adequate, cost me nothing to aquire, and I would be hard pressed to justify spending the money to buy a 'modern' replacement.

I am running Linux on a 400 MHz Pentium II. Since replacing the cathode tube terminal with an LCD my electricity consumption dropped from 7 kWh/day to 4 kWh/day. I am running Einstein, SETI and QMC 24/7.
Tullio
____________

I am running Linux on a 400 MHz Pentium II. Since replacing the cathode tube terminal with an LCD my electricity consumption dropped from 7 kWh/day to 4 kWh/day. I am running Einstein, SETI and QMC 24/7.
Tullio

Our goal is run times in the range from 6 to 24 hours. Note that some optimization is expected in the future so the apps will get faster.

I've got two new WUs running on a 2.66 Pent D (runs XP Pro/1Gig RAM). They're looking like they'll take about 25-26 hours.

One is at ~22% after 5.5 hours and the other one is at ~15% after 4 hours.
____________
Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.

Got 1 WU om my AMD 3500+.
~6.1% after 1hour giving around 16+ hours for total time.
Have too wait a little while see the correct time.

Here are some preliminary S5R2 numbers from my own machines, for anyone who's interested.

Pentium III-450 Katmai
Elapased time so far: 30 hours 21 minutes
Progress so far: 42.792%

Duron 950
Elapsed time so far: 16 hours 5 minutes
Progress so far: 16.038%

Pentium III-733 Coppermine
Elapsed time so far: 14 hours 44 minutes
Progress so far: 34.894%

I don't have any R2's that have fully completed.

The only "modern" machine that I have is my Athlon64 3500+, but it's currently running a QMC unit. So, I don't have R2 numbers for it as yet.

I hope this helps.


____________

I have one unit that will be completed at 10.2 Hours on one core of my Duo 2 Core T7200. The one on the second core is slightly longer 13.1 hours. I guess the new WU's are not all identical in size. Something that might be taken into consideration is extending deadlines for people with slower computers. I was knocking about 12 WU's a day on average and I have a nice laptop... this is going to cut that to about 4 now.
____________
Windows Vista Home Premium
Intel Core 2 Duo T7200 @ 2.00 GHz 4MB L2 Cache
2048 Dual Channel DDR2 553 MHZ
ATi Mobile Radeon x1400 256 MB DDR2 x16 PCI-E

If your p3-450 can do a WU in a little over a day I don't think there's a big problem. Any machine a half dozen times slower is essentially worthless as I argued above and colelctively machines that old represents a very small fraction of the total CPU power on the project.
____________

My Pentium 3 Coppermine 667 estimates about 60 hours for one of the new WUs. As it has only just started, those figure is probably not very accurate, but compared to what Donald said it sounds quite plausible. Fyi: My 3500+ assumes about 15 hours for a WU, which also sounds logical since it's about 4 times as fast as the Coppermine in the benchmarks, but this may also be a very rough guesstimate atm.
____________

My 450 p3 estimates 96 hours.

BAM!
____________

So which one of my little shrubbery is the first to pull a S5R2? Yup, the 400 MHz Celeron. Initial estimate (pre-crunch, still in cache): 128 hours.

I'll keep y'all posted.

[quote

Our goal is run times in the range from 6 to 24 hours. Note that some optimization is expected in the future so the apps will get faster.

Cheers,
Bruce[/quote]

That 6 to 24 hour estimate sure has thrown my systems for a loop. I had 2 that I got yesterday on my AMD 64 X2 4800+ that said 5 hours and are finishing at 13. I'm now working on 2 that look like they are going to be 29 hours on this system. I haven't gotten any units on either the AMD 64 3700+ nor the AMD 2200+.
My biggest concerns are;
- 4800+ used to do a unit in about 2 1/2 hours - now 29 hours
- 3700+ used to do a unit in about 3 hours - now days ?????
- 2200+ used to do a unit in about 4 hours - now days ?????
- While I'm not much into the credit aspect, I'm curious how the extended times will affect whatever my average daily credit was.
- Also with the smaller units it takes sometimes 3 or 4 weeks or more to finally get credit for a unit I had finished in a day or 2 because it had timed out on someone else's system(s). With larger units I would expect the number of these to grow expodentially.

All in all I'm going to take a wait and see attitude before I decide if my 2 lower end systems are up to the task of crunching. I guess my wife's and the server systems could use the break if I stop them.

____________

How much variation in the size of s5r2 units is there?

I'm looking at two estimated times on my three machines between R1 and R2 they are:

2:27 13:59
2:25 9:10
3:08 8:29

A spread of less than 3:1 to more than 5:1, the first 2 machines are A64s, the last a C1D.
____________

Is there going to be any limitation regarding BOINC client ? I still have some clients running Boinc 4.x.
____________

Arion - the credit has increased a great deal. Its roughly 4 times the normal amount. Seems fitting for me since its taking about twice as long for me to complete one WU.

My first WU took roughly 11.3 Hours with 204.9 Credits.

Knocking out my third WU in 10 minutes and halfway completed on the fourth.
____________
Windows Vista Home Premium
Intel Core 2 Duo T7200 @ 2.00 GHz 4MB L2 Cache
2048 Dual Channel DDR2 553 MHZ
ATi Mobile Radeon x1400 256 MB DDR2 x16 PCI-E

Feel free to have a look on this host.
It's a dual P4 2800 Xeon System (with hyperthreading on, so 4 CPU's are available) with 3 CPUs dedicated to E@H.

S5RI times: 18.000 - 19.000 sec. and 54 Cr.
S5R2 times: 89.000 - 90.000 sec. and 204 Cr.

seems fair to me...
____________
Udo

Feel free to have a look on this host.
It's a dual P4 2800 Xeon System (with hyperthreading on, so 4 CPU's are available) with 3 CPUs dedicated to E@H.

S5RI times: 18.000 - 19.000 sec. and 54 Cr.
S5R2 times: 89.000 - 90.000 sec. and 204 Cr.

seems fair to me...

Is there going to be any limitation regarding BOINC client ? I still have some clients running Boinc 4.x.

Is there going to be any limitation regarding BOINC client ? I still have some clients running Boinc 4.x.

Hi Bernd,

is it possible that the Linux application is much faster than the Win-App?

I got the following results about the credit-reduction, which should be 30%:

Athlon64 4400+ (Win XP): ~ -40%
Athlon XP 2600+ (Linux): ~ -15.4%
Athlon XP 3000+ (Linux): ~ -12,8%

All results are compared to at least 3 different groups of "old" WUs.
And both Linux hosts get more credits/h than the faster Win-host(per core)!

cu,
Michael

____________

I should test that... dual-booting can be useful ;-)
____________

OK, I received two early S5R2 units (I have a dual processor). They came across with expected runtimes of roughly 12 and 14 hours. They ran for 42 and 44 hours, respectively (it's really depressing when the "time to completion" is going up instead of down). In S5R2 "faster hierarchical search", i was getting somewhere around 53-54 credits per WU, and they were taking 7-8 hours. For the two forty-plus hour runs, I'm getting under 100 credits each -- total of about 160 for the two. So, S5R1, I would get (conservatively) 500-550 credits for 86 hours of CPU work -- and now I'm getting 160.

I'm back to S5R1 work now, but I've downloaded another S5R2 workunit for later. This one shows as 149 hours (that's more than six days). I'm dreading what will happen when it starts.

Bernd and company, you might want to rethink the size of these WUs. I have a pretty big fast machine; most people won't be able to finish the WUs within the deadlines. And the credit is REALLY off -- not just a little!
____________

is it possible that the Linux application is much faster than the Win-App

OK, I received two early S5R2 units (I have a dual processor). They came across with expected runtimes of roughly 12 and 14 hours. They ran for 42 and 44 hours, respectively (it's really depressing when the "time to completion" is going up instead of down). In S5R2 "faster hierarchical search", i was getting somewhere around 53-54 credits per WU, and they were taking 7-8 hours. For the two forty-plus hour runs, I'm getting under 100 credits each -- total of about 160 for the two. So, S5R1, I would get (conservatively) 500-550 credits for 86 hours of CPU work -- and now I'm getting 160.

I'm back to S5R1 work now, but I've downloaded another S5R2 workunit for later. This one shows as 149 hours (that's more than six days). I'm dreading what will happen when it starts.

I am also worried about the run times of the newer wu's. Typically on the machine that has it I get 3-4 hour per wu, this one has run for 8½ and is saying it is only 36% finished.
____________
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

Argh, I don't believe it, my X2@2,63 GHz is crunching a 182.7 Hz WU, and after 5h it has already finished 27.5%!

How much time will slower hosts need for these WUs?

Is it possible that we are using an app that is a kind of AMD unfriendly? ;-)
Some wrong or missing compiler switches?

This would explain, why the Linux app is so dramatically faster than the Win app, at least on AMD.

My longest WU(210.3 Hz) crunched so far took 56,850 sec for 260,21 credits with an AMD XP 2600+@2.2GHz(Linux). This is really ok! But the time my X2(Win) will need, is totally unsatisfying!

Tonight, I will run the X2 under Linux to compare the results. Maybe that tracks down the problem.

cu,
Micha
____________

For anyone who's interested, here's the latest skinny on the S5R2 units on my machines.

The only machine that's finished one so far is my Pentium III 733 Coppermine machine. It took 41 hours, two minutes and ten seconds for an S5R2 with a claimed credit of 99.99. (Of course, I understand that that credit amount may be adjusted upwards.) My quorum partner isn't finished yet, but considering that he's running a fancy 2.8 GHz P-4, I would hope that he'll be finished soon.

For my other machines. . .

Pentium III 450 Katmai
Elapsed time so far: 52 hours 40 minutes
Progress so far: 77.477%

Duron 950
Elapsed time so far: 31 hours 39 minutes
Progress so far: 92.761%

Athlon64 3500+
Elapsed time so far: 3 hours 56 minutes
Progress so far: 18.536%

I hope this helps!
____________

Well I've completed my first of these.
http://einstein.phys.uwm.edu/workunit.php?wuid=33125698

It took 6 times the length of a normal unit, 12 hrs versus 2 hrs, but gave only twice the credit.


Edit: and only now of course do I spot the comment about the wrong credit above.
____________

Intel coreDuo T5500, 1GB RAM, WinXP
My first S5R2 unit: crunch time almost 18hrs, credit: 258 (14.4 per hour)
average S5R1 unit: crunch time ~3hrs, credit ~54 (18.1 per hour)

Well I've completed my first of these.
http://einstein.phys.uwm.edu/workunit.php?wuid=33125698

It took 6 times the length of a normal unit, 12 hrs versus 2 hrs, but gave only twice the credit.


Edit: and only now of course do I spot the comment about the wrong credit above.

Here's what I've got so far

Intel Core 2 CPU X6800 @ 2.93GHz, old work ~1:30, new ~9:20
AMD Athlon 64 FX-60 Dual Core, old ~2:15, new ~14:30
Intel Xeon X5355 Quad Core @ 2.66GHz, old ~1:45, new ~10:35
Intel Xeon 3.40GHz (with HT), old ~4:50, New 31:30
Intel Xeon 3.06GHz (with HT), old ~5:10, New 20:00

The AMD seems like it's haveing problems with the new work units
____________

With the excessive times it takes to run on my machines vs using S5R1,I might just have to quit Einstein@home and go back to doing dnetc(I really like einstein@home,but this just might break me away for good.)
____________

hello all R2 crunchers,

take care of the results from the second run of these s5r2. i see in the results, that the analysis is still in a debug mode, something like

4555, 4556, 4557, 4558, 4559, 4560, 4561, 4562, 4563, 4564, 4565, 4566, 4567, 2007-04-18 11:39:23.2812 [debug]: set_checkpt(): bytes: 884442, file: 884442

in the former s5r1 wasnt any stuff like that. i guess the coders are working hard, to solve the debug problem, it is an "hierarchic search" of the pulsars, i imagine this like a tree (top->down) and the sorting and filtering of interesting signals would be dramaticly false, if some attribute is wrong, so they need some test results for this. but why didnt they do this on bruce´s cluster with different os ?

the result time is now at a point, where my old machines (w/o 24/7) needs much time, but i guess it is possible to return it in 2 weeks. the credit system from now on is more fair to other boinc.projects, so i can understand the correction so far. our rac falls down from hour to hour, but anyhow - it is science, not warcraft :D

cu crunching

I just tried something. This is on an IBM x360, four 1.9Ghz HT MP Xeons 512 L2 and 1024 L3 cache, 2GB of ECC DDR2100 ChipKill.

Turned off HT, watched the time to completion rise as it had been with HT on. I then set processor affinity for each WU and noticed that the rate of climb started to fall off, by 12sec per 1 minute of run time. These 4 WU's are between 4% and 15% done.

Anyone with an SMP box give it a try and see if you come up with the same kind of results.
____________

I just tried something. This is on an IBM x360, four 1.9Ghz HT MP Xeons 512 L2 and 1024 L3 cache, 2GB of ECC DDR2100 ChipKill.

Turned off HT, watched the time to completion rise as it had been with HT on. I then set processor affinity for each WU and noticed that the rate of climb started to fall off, by 12sec per 1 minute of run time. These 4 WU's are between 4% and 15% done.

Anyone with an SMP box give it a try and see if you come up with the same kind of results.

They do go up and down, a little. But, since setting affinity the trend in down, not up as it had been. One that I restarted after shutting HT off is at about 11%, and it's time is dropping. Also, the time to completion on the queued WU's is dropping, not increasing.
____________

Ok, my result ran for 18 hours and a bit. So not bad on a P4 2.8GHz.
____________
Jord

-The BOINC FAQ Service
-CUDA/CAL Stream FAQ

Well, my first R2 wu finished on my 2.8GHz P4 in 82400 seconds, (22.8 hours), and has claimed 259 credit, (not granted yet). This gives 11.32 credits per hour, whereas this machine normally earns 15.3 credits per hour with R1 wu's.

From a purely runtime perspective, at least one of my machines will have to stop crunching Einstein.
____________
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

First 3 done, 1 granted 135.47 for 20.82 hours. The other 2 are pending. They claimed the same credit as the first one.

This machine was doing 18-18.5 WU's per day with HT on, and 17-17.5 with HT off. RAC was in the 950-980 range on R1's. If the R2's are all like the first 4, and I know they will not be, you are looking at about 625 a day for 4 WU's per day, plus a little on the next 4.
____________

From a purely runtime perspective, at least one of my machines will have to stop crunching Einstein.

From a purely runtime perspective, at least one of my machines will have to stop crunching Einstein.

As long as the deadlines remain roughly proportional to the crunching time, what difference does it make how long the individual tasks take? Taking arbitrary numbers out of the air, suppose an old and sporadically available system could only be sure of completing one ten-hour WU in two weeks. Shouldn’t it be equally capable of finishing a forty-hour run in two months?

Thanks to Bernd & Bruce for the heads-up.

Is Chilango's result (the only one I've seen linked so far) likely to be typical in terms of production run-time - i.e. 5.4 times S5RI run-time, in his estimation? (I stress production run-time, since it looks as if it still has a lot of debug output).

If so, you might like to consider whether 14 days is still the appropriate deadline: my Celeron, allowing for its current 50% share with SETI, would take about 17 days pro-rata. Not that the project should be run for the benefit of one superannuated cruncher, of course!

Our goal is run times in the range from 6 to 24 hours. Note that some optimization is expected in the future so the apps will get faster.

Cheers,
Bruce

I just tried something. This is on an IBM x360, four 1.9Ghz HT MP Xeons 512 L2 and 1024 L3 cache, 2GB of ECC DDR2100 ChipKill.

Turned off HT, watched the time to completion rise as it had been with HT on. I then set processor affinity for each WU and noticed that the rate of climb started to fall off, by 12sec per 1 minute of run time. These 4 WU's are between 4% and 15% done.

Anyone with an SMP box give it a try and see if you come up with the same kind of results.

Deadlines aren't changing. I'm now getting a 149 hour item, still two weeks.

My Celeron 433 got an extra long unit(243.10 Hz) that is supposed to take 9d 8h 47m 19s to complete. *lol*
And yes, I know there are much shorter units. But don't forget the people who crunch for more than one project. They may face serious problems with the current deadline.

cu,
Michael
____________

Actually, that's not too surprising. HyperThreading can help speed up server applications, but it can slow down most other applications. So, for desktop use, you're probably better off with HyperThreading turned off.

Looks like it will take around 26 hours on a Mac G5 2.0GHz.

____________

My Celeron 433 got an extra long unit(243.10 Hz) that is supposed to take 9d 8h 47m 19s to complete. *lol*
And yes, I know there are much shorter units. But don't forget the people who crunch for more than one project. They may face serious problems with the current deadline.

If you with last series get 53 kredits for 9000 seconds why now for 630000 seconds dont give 371 credit? You only give 204 credit!

All of us making Enstein project live works for something. And credits are just number and for project developer cost is zero. You can give 1000 credits: and nothing heapen: the world stil rotate around sun.
____________

@all AMD/Windows users:

The current science application seems to have a problem on these hosts, running about 25% slower than it should. Probably a compiler problem.
There are no problems on AMD/Linux, Intel/Linux and Intel/Win.
I'm shure Bernd Machenschalk will fix this problem as soon as possibe, so just sit down and hold the flag. ;-)

The time, the boinc client shows for the units is mostly far away from reality, so just wait till the first units are finished.

What I've seen so far, given credits are between ~130 and more than 300, so the time they take will be appropriate different.

On the three 'working' platforms the credit reduction is below 20%. We all have to face this and this is really ok.

cu,
Michael

[edit]spelling
____________

> S5R1 took on average 9400 seconds on an AMD X2 4800+, this gives 20.9 average cobblestones/hour approximately.
S5R2 (1 done) took 53270 seconds for 191 claim which gives 12.9 cobblestones/hour.

For this machine this works out at 38.3% reduction per hour.
30% would be around 14.5 cobblestones/hour (for this computer) which still puts it below most other projects that I work for including Rosetta.

Increase in WU length by 5.8 times. Increase in amount granted per WU is 3.5 times.

The 30% granted cobblestones above others was on a per second basis (does this then scale up to per hour basis as well?).

Waiting to see what my Opterons do, first estimate is over 18 hours (64800s).

____________

Am now crunching my first S5R2 workunit and the time to completion has gone up from sbout 16 hours to an estimated 50!!! Sheesh there must be a lot more in these new ones than the old!
____________

____________
Mrs Miggins - A lady of uncommon refinement

Hi ziegenmelker,

AMD/Windows users:

The current science application seems to have a problem on these hosts, running about 25% slower than it should. Probably a compiler problem.

Hi ziegenmelker,

AMD/Windows users:

The current science application seems to have a problem on these hosts, running about 25% slower than it should. Probably a compiler problem.

The current S5R2 code control the FPU in ugly and crazy ways.
It's coming from the compiler. The AMD CPUs dislike it.

akosf

Hi ziegenmelker,

AMD/Windows users:

The current science application seems to have a problem on these hosts, running about 25% slower than it should. Probably a compiler problem.

The current S5R2 code control the FPU in ugly and crazy ways.
It's coming from the compiler. The AMD CPUs dislike it.

akosf

Hi Akos,

nice to read you again! :-)

Hi ziegenmelker,

AMD/Windows users:

The current science application seems to have a problem on these hosts, running about 25% slower than it should. Probably a compiler problem.

The current S5R2 code control the FPU in ugly and crazy ways.
It's coming from the compiler. The AMD CPUs dislike it.

akosf

Hi Brian,

Hello Akos! Please take a look at this host. ...

Thanks for the heads-up about AMDs having problems under Windows, I'll draw my conclusions from that. Don't worry, you're not going to lose crunching time- thank goodness I dual boot!
____________

Thanks for the heads-up about AMDs having problems under Windows, I'll draw my conclusions from that. Don't worry, you're not going to lose crunching time- thank goodness I dual boot!

Hi Brian,

Hello Akos! Please take a look at this host. ...

You should try to contact with Bernd Machenschalk to find the cause of this fault. I think your OS cause it, but i'm not sure.

akosf

Hi Brian,

Hello Akos! Please take a look at this host. ...

You should try to contact with Bernd Machenschalk to find the cause of this fault. I think your OS cause it, but i'm not sure.

akosf

Akos,

I'm not the owner of that host. I don't have an S5R2 problem, because I stopped doing work for this project when S5R1 ran out. :-( With my background in software development, what was being said about R2 prior to its' release was enough for me to recognize that it should not be in production.

Since I have an AMD, and an overclocked one at that (so in theory more vulnerable to whatever problems there may be), I'll try out 1 or 2 S5R2 units. Perhaps they will help.

Brian

acording to earlier posts by either Bruce or Bernd there's roughly a 4x spread in WU size in s5r2. At present there doesn't seem to be any control over what hosts recieve what WUs, given the frantic nature of getting this app out before s5ri completed I'd imagine a WU/cruncher partition like with s5r1's slowere hosts only getting small WUs is on the TBD list but hasn't been implemented yet.
____________

acording to earlier posts by either Bruce or Bernd there's roughly a 4x spread in WU size in s5r2. At present there doesn't seem to be any control over what hosts recieve what WUs, given the frantic nature of getting this app out before s5ri completed I'd imagine a WU/cruncher partition like with s5r1's slowere hosts only getting small WUs is on the TBD list but hasn't been implemented yet.

Maximum daily WU quota per CPU 4/day

This user's Core 2 system running Linux is also crashing

S5R2 errored out
____________

I compared the Windows and the Linux app on one host(AMD X2):

Original host OS Win XP Pro:
workunit

63,516.48 sec / 247.51 credits

OpenSuSE 10.2 as guest in VM:
workunit

56,009.63 sec / 337.99 credits

I think there is no further comment about this needed.

cu,
Michael
____________

And in VMWare, too. One would think that costs a few flops extra. Really impressive, good reason to run my Debian more over the next time- and tip off my mate who runs all those AMD hosts...
____________

I've seen the lower 'efficiency' in the S5R2 work units as well (less credit per cpu cycle). Someone mentioned pushing Einstein behind Seti in their work share, well, at this point, perhaps some other project would make more sense. SETI is doing a fair amount of hardware shift out and for the next few weeks, you'll find it takes a lot more effort to get thru to Seti (both upload and download).

For me, I have pushed *both* SETI and Einstein back in the work share priority. I've got six active projects (though BBC Climate is a small runner for me), and so what I've done is reduct the work share for both SETI and Einstein back to 1/3 what I had been running. This results in a couple of my 'younger projects' (World Grid and Rosetta) picking up the increased share.

I suspect that the SETI issues will be resolved in a couple of weeks, then I'll move it back up the list. As to Einstein, well, here is hoping that they either tweak the code so it returns to its previous efficiency, or alternative, re-jigger the awards to get back in sync. At that point, I'll move Einstein back up the charts as it were. The thing is, there are a bunch of good projects out there, so it simply is a matter of allocating the CPU cycles between them for me.


____________

For me, I have pushed *both* SETI and Einstein back in the work share priority.

Hi!

oh oh...


It seems my first S5R2 work unit just crashed at the very moment I pushed the "Update project" ("Aktualisieren" in german, anyway) button :-(. Is this a general problem? Or was the task deliberately terminated by request from the server???

Host is a Pentium M (Banias) 1.5 GHz under Linux, kernel 2.6.17, BOINC version 5.8.15.

The work unit to be reported was still an S5RI one, the crashed WU was about 62 % finished.



Mo 23 Apr 2007 00:17:23 CEST|Einstein@Home|Sending scheduler request: Requested by user
Mo 23 Apr 2007 00:17:23 CEST|Einstein@Home|Reporting 1 tasks
Mo 23 Apr 2007 00:17:28 CEST|Einstein@Home|Scheduler RPC succeeded [server version 509]
Mo 23 Apr 2007 00:17:29 CEST|Einstein@Home|Deferring communication for 1 min 0 sec
Mo 23 Apr 2007 00:17:29 CEST|Einstein@Home|Reason: requested by project
Mo 23 Apr 2007 00:17:30 CEST|Einstein@Home|Deferring communication for 1 min 0 sec
Mo 23 Apr 2007 00:17:30 CEST|Einstein@Home|Reason: Unrecoverable error for result h1_0198.80_S5R2__48_S5R2c_1 (process got signal 11)
Mo 23 Apr 2007 00:17:30 CEST|Einstein@Home|Computation for task h1_0198.80_S5R2__48_S5R2c_1 finished
Mo 23 Apr 2007 00:17:30 CEST|Einstein@Home|Output file h1_0198.80_S5R2__48_S5R2c_1_0 for task h1_0198.80_S5R2__48_S5R2c_1 absent
Mo 23 Apr 2007 00:17:30 CEST|Einstein@Home|Starting h1_0198.80_S5R2__45_S5R2c_0
Mo 23 Apr 2007 00:17:30 CEST|Einstein@Home|Starting task h1_0198.80_S5R2__45_S5R2c_0 using einstein_S5R2 version 414

Let me say that even if the S5R2 run may have its problems at the beginning, I think it is worthwile to contribute to it, because this is the only way to help improve the software. There are so many different OSes, different hardware and different BOINC versions out there that you cannot expect everything to work without any glitch from the very beginning. Even if this was kind of experimental software, I'd still be happy to participate. And the credits....well... I couldn't care less :-)

CU

BRM

Yeah, I agree with that (well I do care a bit about credits, but only to see if I'm getting better and getting the most out of my boxes, and to compete a bit with some friends)- as long as we can help the project it's okay for me, even if it doesn't run that smoothly atm. And don't forget it's also quite informative to have a "live" view into the development process, especially with people like Bruce, Bernd and Akos taking there time to give us so much information about what's being done. Personally, I see this as an interesting challenge much rather than being annoyed- and hell, it DID wake up these message boards :-D
____________

Personally, I see this as an interesting challenge

It worked for me.

Estimated time after downloaded was 22 hrs plus a few minutes.

Crunch time was 24 hrs 46 so only a little off.

Using WinXP2 SP2, Athlon XP2800+, no overclocking.

Unit validated successfully and received credit.

I've been watching my systems pretty close over the weekend. On my dual core AMD 4800+ units were downloading with expected completion time to start with of 5 hours. Its has increased through 14, 24, 41 and 48 hours. For a 24 hour completed workunit to only receive 310 credits is either a serious flaw in the calibration or an insult to my computers spare cycles. I was averaging 18 to 20 workunits with an average credit of 53 to 54 each for a total daily average of around 1000 credits for this system. To have it cut to 1/3 the credits it was EARNING with the S5R1 units is NOT an adjustment.
Both of my other systems are AMD's, one a single core 64 3700+ and the slowest an Athlon XP 2200. For most of the weekend they were both still getting the old workunits. They both have received units with expected 13 and 16 hour completion times with the new S5R2 units. I'm going to allow those 2 systems to finish them, but I've set all my system to "No New Tasks" until this mess is straightened out. I aborted the 48 hour workunit on my 4800+ system since I can't see wasting my computer's time on an obvious flawed unit.
When all this mess is straightened out I will gladly resume working on new units. I'm sorry but I beta test enough things now that beta testing this is overload for me.

____________

I have finished 5 units so far.

Only one on my slower P4 2.5 that took about 32hrs (336.68 claimed credits)

And 4 on the dual P4 3.2 that took about 17hrs (each @ 172.70 claimed credits)

I haven't tried the AMD 3200 yet but may have to do that now.



____________

I just completed my first S5R2 using a 2.07 3200+ 64 bit Athlon and the points that I received on a per hour basis were considerably less than I was recieving from the S5R1 jobs. For example, on average running a S5R1 job I would recieve approx. 54 points for approx. 10,900 seconds of work. By comparison, I received only 204 points for 66,700 seconds of work running an S5R2. A considerable drop in points received/time.

If this is the same for everyone, then there is not problem, but I remember reading another post a couple of days ago that this result is typical only for users of AMD chips, and Intels were getting points at about the same rate as they did for S5R1 jobs.

Has anyone else noticed a drop in the number of points recieved over a specific time running the S5R2 jobs with AMD chips?

Gary


____________
In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.....Douglas Adams

and Intels were getting points at about the same rate as they did for S5R1 jobs.

I have two a64's and a c1d machine. Both A64's are getting about 60% the credit they did under s5r1, the c1d is getting 75%. The developers are waiting until they're sure all the bugs are worked out before replacing the c++ hot loops with assembly code which should even out the difference. Akos has described the compiler output as being very ugly and aparently intels on chip otimizer is managing to do a better job of sorting it out.
____________

I have two a64's and a c1d machine. Both A64's are getting about 60% the credit they did under s5r1, the c1d is getting 75%. The developers are waiting until they're sure all the bugs are worked out before replacing the c++ hot loops with assembly code which should even out the difference. Akos has described the compiler output as being very ugly and aparently intels on chip otimizer is managing to do a better job of sorting it out.

BTW, What is the URL that allows you to view your stats in that small rectangular box in your post? I can find all of that data at
http://www.boincstats.com/search/all_projects.php?cpid=e13c95468dd08fb861911e0947bc99e2 but have yet to come up with my stats in the above form that can be added to a post.

Gary

Hi....

Seems like I got lost in the shuffle on this (I don't have a lot of spare time to invest, just spare computing power).

I didn't realize there was a changeover happening...last task was taking over 30 hours and I thought something was wrong so I aborted the task. Since then I've tried to get more work but it isn't happening.


I'm running 5.8.8 on XP 64, athlon processor. Is this part of my problem? Can somebody give me a simple answer what I need to do to get back on board with this?

Thanks, Dennis

I have a WU crashed too.... In my case, it crashed when it was as a screensaver. When I went back to the machine, and touch a key, it crashed...

Hey, last time we had a sort of a beta E@H program, it was shutdown with somewhat bad manners because they said that Akosf efforts and all the betas he made where hurting the program and compromised it... which was completely wrong... now they do a beta run for the whole community!!!!
____________

Hi....

Seems like I got lost in the shuffle on this (I don't have a lot of spare time to invest, just spare computing power).

I didn't realize there was a changeover happening...last task was taking over 30 hours and I thought something was wrong so I aborted the task. Since then I've tried to get more work but it isn't happening.


I'm running 5.8.8 on XP 64, athlon processor. Is this part of my problem? Can somebody give me a simple answer what I need to do to get back on board with this?

Thanks, Dennis

[/quote]
Nope, not your problem. But using an AMD chip may very well be. I have 5 of them, but only one is a 64bit, but from reading several of the posts in this thread as well as others, running these jobs may be significantly faster with an Intel. Oh well, they have to get something right every now and then.

Gary[/quote]

Well, I do this because I like being part of something bigger than myself but Intel aint gonna happen here. Ill just periodically keep trying to connect until this straightens out then.

Dennis

The developers are waiting until they're sure all the bugs are worked out before replacing the c++ hot loops with assembly code which should even out the difference.

The bugs that are currently on display should have caused a waiting period before replacing S5RI with S5R2. S5R2 in its' current state is clearly a beta product.

After sometime I terminated the first unit with S5R2.
I need a long time, about 12 hours (51.900 sec). So I needed about 5,4 times
more than normal. So I am a little bit disapointed that I get only 113 points.
The double of the normal units.

Please read this complete thread...

As Dr. Bruce Allen has said here:

Our goal is run times in the range from 6 to 24 hours. Note that some optimization is expected in the future so the apps will get faster.

Cheers,
Bruce

No one is forcing you to run S5R2.

Then again, there are no S5R1 and S5RI results anymore, but for a couple of resents, so what do you want? To have the big group of Einstein's volunteer crunchers to wait around for that other smaller group of volunteer crunchers to Beta test S5R2? ;-)

After sometime I terminated the first unit with S5R2.
I need a long time, about 12 hours (51.900 sec). So I needed about 5,4 times
more than normal. So I am a little bit disapointed that I get only 113 points.
The double of the normal units.

I've just started my first S5R2 WU. It's behaving like Rosetta, i.e. the "time remaining" increments along with the CPU time. Is this to be expected. The estimated time for the WU was around 12 hours. It's done 3.33 hours (16%) but the "to completion" time is showing 11:17:43...all OK? I think not!!

No one is forcing you to run S5R2.

Correct. I went to no new work just prior to it because I had a feeling that something like this was going to happen.

Then again, there are no S5R1 and S5RI results anymore, but for a couple of resents, so what do you want? To have the big group of Einstein's volunteer crunchers to wait around for that other smaller group of volunteer crunchers to Beta test S5R2? ;-)

In my opinion, yes. Alternatively the project could do something like what SETI Classic did and simply reissue some of the S5RI work. S5R2 has what was being defined as "brand new coding" (paraphrase). Just that simple statement in and of itself should've called for an open beta. They have done so in the past for other releases.

S5R2 is a beta product. The project team apparently made the decision to go anyway, apparently based on something that I sense that you're echoing; that "some work being done is better than having a couple of weeks with no work".

Jord, seriously, look around. There are immediate crashes. A project developer has plainly stated that things look bad right now. Prior to release another project developer stated that they were taking care of "showstoppers" in the 3-5 day window before the app was sent out to the public. S5R2 is beta. If you would like to explain how I am incorrect, then feel free to do so, but please, don't just try to use the "no one is forcing you" line again.

Thanks,

Brian

[/quote]No one is forcing you to run S5R2..... ;-)[/quote]
Jord is right...no one is forcing anyone to process S5R2. Our AMD machines will crunch elsewhere until all this is resolved...CU later(hopefully)....Cheers, Rog.
____________

At any rate, perhaps they alternatively could have said something like "the whole project is going to Beta for the next few weeks, we are looking for........ If you do not wish to participate, please come back an see us in a few weeks!"

That would at least have given people a "heads up" to know what to expect.

At any rate, just my 2 cents.

We therefore will set up S5R2 as a short, experimental run that limits the search to parts of the parameter space where the overhead is well under control. During this short run we will improve the Application in various aspects. The results will also help us tuning the parameters for the next, larger run (probably named S5R3).

Einstein@Home has a beta test page from the beginning of the project.
As far as i know it's open for everybody. S5R2 code was tested on it.

-from the OP at the start of this very thread. Sounds like a pretty good "heads up" to me.

Einstein@Home has a beta test page from the beginning of the project.
As far as i know it's open for everybody. S5R2 code was tested on it.

S5R2 is a beta product. The project team apparently made the decision to go anyway, apparently based on something that I sense that you're echoing; that "some work being done is better than having a couple of weeks with no work".

I haven't seen S5R2 being in test, to be honest, so yes, your anaolgy could be correct. Then again, the same thing happened for S5RI ... :-)

And the credit thing is explained: EAH was asking too much.
Since credits are now set server side, it doesn't matter much. Can you even compare it to S5R1 (and RI)?

The RI was a name change only. It was done to purge the last unoffficial akos s5 clients from early in the run.
____________

Einstein@Home has a beta test page from the beginning of the project.
As far as i know it's open for everybody. S5R2 code was tested on it.

I never once saw it listed Akos... If you are correct, I will revise my strong opinion, but I can say that I personally never saw it...

Has anyone heard if there is going to be an adjustment in the S5R2 credits? Like a lot of people posting here I have seen a big drop with the S5R2, my T2250 went from about 17/HR down to 13 hour. My P2 2.4 going from about 12.5/Hr to about 9/Hr.
Ray
____________

Try the Pizza@Home project, good crunching.

Has anyone heard if there is going to be an adjustment in the S5R2 credits? Like a lot of people posting here I have seen a big drop with the S5R2

You can find a cross-project credit comparison here. You will see that almost all projects are in good agreement with SAH. EAH was out of line. We're now fixing that.

(The credit comparison chart was created at my request by James Drews. It provides a solid basis for cross-project credit comparisons.)

Cheers,
Bruce

[quote
Credits here have been server-side for quite some time now...including RI. Bruce made adjustments to R1 and/or RI during the run.[/quote]

I read that Bruce had made adjustments to the earned credits and while I might not agree with his reason for doing it, I think that the "adjustment" is out of whack. I have 3 machines that were working.

#1 - AMD 64 X2 4800+ went from 2 1/2 hrs 54 credit to 48 hours 320 credit
#2 - AMD 64 3700+ went from 3+ hours 54 credits to 13 hrs 171 credits
#3 - AMD XP 2200 went from 3 1/2 hours 54 credit to 16 hours 170+ credits.

I'm not that big on credits because I volunteer my time and equipment to the 2 projects because of the science and interest. However, I do look at my clients stats on a regular basis for comparison as how each machine does to my overall effect. After seeing my graphs take a nose dive I came on the board to find out what happened. I've read Bruce's explaination and decided to see how it went (specially with his warning to dial up users). I've watched the results on my systems since last week when I started getting the new S5R2 units and have since watched in amazement as my Boinc overall project starts have nose dived. I honestly would have expected to see some adjustment, but I certainly didn't expect to see the same amount of work on a daily basis send me crashing downward. Added to that with smaller units, I was having to wait sometimes 3 or 4 weeks or more for credit to be applied because someone had not finished/ passed the deadline. I can only imagine at this point that those number of units are going to vastly increase since the longer the units take the more likely that the new users getting them are going to stop working and there's another 2 week delay before its sent out again.
I've been with Seti@home since 1999 and Einstein@home since 3/2005. I've weathered through a lot of problems (mostly with Seti) and even here recently when we had all those server problems around the first of the year WITHOUT shutting any of my systems off a project. I came to Einstein when Seti suggested that people had a backup project in case there were problems with one and instead of leaving our systems idle we could be using them for the benefit of other worthwhile projects. It didn't take me long to see that Einstein was the most reliable/stable project and I changed all my systems over to Einstein as my primary and left Seti as a backup. As of last night however, I have reluctantly aborted all work on Einstein and set my clients to "No New Work" until this "mess" is resolved.
I honestly don't mind my systems taking a bit longer to process the work, specially as Bruce mentioned that the new units were going to include a lot more science. But I think the times are EXCESSIVE when my systems on an average take 10 or more times as long to complete the work and I only receive at best 1/2 the credit. I understand there are problems and that adjustments to optimilization will occur over time. As I stated in a previous message I beta test for a number of companies and I therefore am familiar with what we go through trying to work the bugs out. IMHO, S5R2 hasn't even reached the beta stage with all the problems of credits, times, efficent use of various CPUs, etc. I have neither the time nor the patience to throw one more beta program on my systems. When Einstein gets the majority of these problems resolved I will gladly resume my work with the project.

[edited for spelling]
____________

but also because a lot of us would had gladly participated in a beta testing, had they offered one.

IMHO, S5R2 hasn't even reached the beta stage with all the problems of credits, times, efficent use of various CPUs, etc.

Arion: the 40% drops on your 2nd and 3rd machine are about what I saw with my two A64x2 machines. The much larger drop on your first machine makes me think Bruce didn't fix all of the first round of WUs which were issued which much lower credit amounts. One of the moderators probably should send a ping about this his way.

Einstien was giving ~1/3rd too much credit. The s5r2 app is dropping results by about 25% on intel, 40% on AMD. The spread is due to the app using compiled code instead of hand tweaked asm for the hot loops and intels on chip optimizer doing a better job with the output. Eventually the s5r2 app should be optimized the same way as the s5r1 app was and the CPU type difference closed. There are occasional reports of people not having the new app work yet, tracking down the handful of rare bugs is a higher priority.
____________

I understand what you're saying, but I disagree on a number of things in that statement. I do feel that S5R2 is at a "beta" level. I cannot say that it is "alpha" / "pre-beta". The issue with the credit level needs to be taken out of the discussion as well. That isn't determined by the application that you and I have. Credits are a distinctly seperate, although related, issue.

After sometime I terminated the first unit with S5R2.
I need a long time, about 12 hours (51.900 sec). So I needed about 5,4 times
more than normal. So I am a little bit disapointed that I get only 113 points.
The double of the normal units.

I've just started my first S5R2 WU. It's behaving like Rosetta, i.e. the "time remaining" increments along with the CPU time. Is this to be expected. The estimated time for the WU was around 12 hours. It's done 3.33 hours (16%) but the "to completion" time is showing 11:17:43...all OK? I think not!!

On all 3 of my systems the DCF has gone up between s5r1 and s5r2. That is to say that at the transition point your ETA for completing a WU is too low. Boinc appears to be adjusting this value during the running of a single WU and is adjusting the ETAs as a result. MY PCs have all stabilized around the new correct value after a half dozenish WUs.

Einstien was giving ~1/3rd too much credit.

tracking down the handful of rare bugs is a higher priority.

Arion: the 40% drops on your 2nd and 3rd machine are about what I saw with my two A64x2 machines. The much larger drop on your first machine makes me think Bruce didn't fix all of the first round of WUs which were issued which much lower credit amounts. One of the moderators probably should send a ping about this his way.

Einstien was giving ~1/3rd too much credit. The s5r2 app is dropping results by about 25% on intel, 40% on AMD. The spread is due to the app using compiled code instead of hand tweaked asm for the hot loops and intels on chip optimizer doing a better job with the output. Eventually the s5r2 app should be optimized the same way as the s5r1 app was and the CPU type difference closed. There are occasional reports of people not having the new app work yet, tracking down the handful of rare bugs is a higher priority.

Einstien was giving ~1/3rd too much credit.

That's the "half-empty" view. The "half-full" view is that the other projects were not giving enough credit...

I agree it would probably be better to run a seperate Beta project, but that would entail a significant increase in backend resources which just may not be available.

Agreed, and I'm always in favor of more input from the team.

At the risk of being accused of being an EAH 'cheerleader' though, I offer as a mitigating factor in the current situation that Dr. Allen moving to Germany so close to this particular transition must have thrown a wrench or ten into works. ;-)

Alinator

BTW, What is the URL that allows you to view your stats in that small rectangular box in your post? I can find all of that data at
http://www.boincstats.com/search/all_projects.php?cpid=e13c95468dd08fb861911e0947bc99e2 but have yet to come up with my stats in the above form that can be added to a post.

Gary

If you follow through to the detailled pages on BOINCStats you'll find:



You can see the code format by clicking 'reply' to this message: but if you plan to use it permanently, please put the code into the 'signature' section of your forum preferences (it's much kinder for dial-up users: they - or anyone else - can filter the images out so the pages load quicker).

I'm still trying to decide what I'm going to do with the 3 S5R2 units that I suspended... I need to do something with them in fairness to the other person who received a download...and in fairness to the project...

And I'm desperately trying to make my Debian cooperate with the cheap trashy modem my new ISP provided... as that host is an AMD, I'd really prefer running Linux, but it simply won't let me go online and get BOINC and a few WUs... Murphy's law, sth like this happens in the worst possible moment in the worst possible way (if the host in question was my notebook, I could just take it to Uni with me and upload there, but no, the notebook is an Intel and it's the desktop which has problems...). I'm still trying not to let Einstein lose crunching time but unless I manage to go online via PPPOE and without DHCP really soon (which I've never done before as my former ISP provided a proper router, which unfortunately doesn't work together with the new VoIp solution) I'll have a hard time keeping that promise.
As for the current run's "beta" state... I don't know more than anyone else about the internal decision-making, of course, but I prefer the current state of things to not getting any Einstein work at all. This is simply my favorite project (by far) and has always been reliable, so a few issues won't be enough to drive me away ;-) Of course, I'd gladly participate in an organized beta test, too, if there was one for future runs. But as this run was started in a hurry and the team was probably delayed by all the server trouble earlier this year, I think it's quite understandable that the app couldn't be tested as much as would have been ideal...
____________

The floating point speed of the whole project is going down.... Now is the lowest in 8 weeks and sinking fast. I am worried that it means that a lot of people is quitting the project. I suggest that it will be wise for the developers to say something before it becomes a permanent trend... I think if they talk honestly people will understand and collaborate on a beta test, or at least come back a week or two later, when everything will be working right.
____________

Maybe it's going down because computers aren't connecting as much since the work units are so big?

I mean, my computers used to connect daily for work. Now, some go several days without needing to connect.
____________

And I'm desperately trying to make my Debian cooperate with the cheap trashy modem my new ISP provided... as that host is an AMD, I'd really prefer running Linux, but it simply won't let me go online and get BOINC and a few WUs...

Well, if it's an option for you and you have 256kB memory left, you could just run Windows and install VMWare Player, like I did. You could do a small Linux installation, get 35% more credits and 35% more benefit for the science. :-)

Well, if it's an option for you and you have 256kB memory left, you could just run Windows and install VMWare Player, like I did. You could do a small Linux installation, get 35% more credits and 35% more benefit for the science. :-)

What hardware platform are you running on? IIRC someone else tested on his dualboot and only found a few % difference between windows and his linux distro.

I have to add that my router, a Celeron 433 ist much faster with th new app:

old: 1.49 cr/h
new: 2.82 cr/h (extrapolation)

:-))

This clearly proves, that there is no SSE used in the new app.

cu,
Michael
____________

Os Win XP pro:
old app: 23.12 credits/h
new app: 14.11 credits/h !!!

Guest: OpenSuSE VM 1,2:
new app: 21,72 and 21,75 credits/h :-)

I have to add that my router, a Celeron 433 ist much faster with th new app:

old: 1.49 cr/h
new: 2.82 cr/h (extrapolation)

:-))

This clearly proves, that there is no SSE used in the new app.

I have to add that my router, a Celeron 433 ist much faster with th new app:

old: 1.49 cr/h
new: 2.82 cr/h (extrapolation)

:-))

This clearly proves, that there is no SSE used in the new app.

Not really. All it proves is that the margin between the two is much narrower. IIRC one of the changes made for this build was that the compiler was set with autovectorization turned on which generates SSE code without the developer having to do as much setup work in advance. Since the einstien apps have been designed to automatically switch between x87 and SSE code depending on CPU capabilities you'd need to actually decomile the app to see if it does or doesn't.

Whoa. Definately going to have to setup some VMs myself at some point in the near future.

I have Rosetta already loaded, and if the AMD is an inferior CPU running E@H THAN I"LL JUST SWITCH

Why don't you give the developers a chance to correct the mistakes? It's not a problem of AMD, it's a problem of the MS compiler they used, respectively the compiler swiches. The Linux app on AMD cpus runs as fast as it is currently possible, so that proves, where the problems are.

Now E@H runs 100% priority, as i`d like to get a few of new WU`s done as soon as possible. Then i`ll be able to decide, either to continue, or maybe wait, doing only S@H, till the app gets debugged. I`m not krunching for credits, but i do not want my cpu`s cycles being wasted... or maybe - not being used effectively.

That`s why i`m running S@H apart from Einstein. LHC has very little work, also getting processed very fast.

I`m patient... i`m not throwing everything overboard, nor panicking...

But this compiler issue is affecting over half of OS`es and almost half of CPU`s... this is not an obscure, hard-to-spot mistake. It should be spotted a bit earlier, than several days after official R2 start. Anyway, it`s too late by now. I just hope some conclusions would be drawn from this accident.
____________

Well I'm probably going to choke on crow with this but I Took all my systems down from Einstein yesterday and went back to running Seti as my primary project. I only run standarized clients on both projects. Based ONLY on the results on my X2 64 4800+ system I'm beginning to get a clearer picture of the parity of the credits granted between Seti and Einstein. I understand that each project grants credits on a different scale. (client side as opposed to server side) And I've found workunits that compare in the amount of time processed and the results were.

Seti - 2 1/2 hours aprox. process time and 27 credits
Einstein S5R1 - 2 1/2 hours aprox. process time and 54 credits

For the S5R2 comparison I had to add 3 seti units to get an equivilant amount of processing time and therefore I added the credits to get a comparison. The times and credits are an aprox. figure as I rounded off for simplicity sake.

Seti - 17,000+17,000+14,000 Seconds = 61+61+50 (172) credits
Einstein S5R2 - 48,000 seconds - 172 credits

Earlier tonight I had decided that I'd been running einstein non stop for 2 years (with exception for when the project was down and we couldn't get workunits)<--- a RARITY as opposed to Seti being up and down numerous times. For a large part of that I had 2 of my systems running both projects with occasional times when I would run Einstein exclusively on them. First when Seti was down and couldn't get work and the past few months devoting all my systems to Einstein as it was short workunits, and the projects servers stablilized after the Jan. problems.

With this in mind and taking into consideration of all the other pluses with Einstein I brought both of my slower systems back on to Einstein figuring that I'd weathered a few rough times in the past and that this was just another small bump on a lengthy journey with Einstein@home. Even though my 2 systems do a miniscule amount of work compared to all the others maybe my workunits can in some small part help the developers figure out what's wrong with the AMD clients and help them get a fix in place. In the meantime, I'm letting my 4800+ system run Seti exclusively so that I can work on bringing my total credits with Seti to the 100k mark. When I reach that goal or Einstein gets the AMD problems worked out (whatever is first) I'll bring it back online with Einstein exclusively.

Back to the credits granted for work - I've never really paid much attention to what I was getting on each project before now. I've mostly looked at the graphs in the clients and watched what my daily averages were on both total credits and individual hosts. There is a lot to be said for looking at the stats and seeing each system spend about the same amount of time on a regular basis and getting about the same amount of credit as well. With Seti the workunits are all over the place and the amount of credits follows suit. With Einstein I could figure that x system runs 18-20 units a day and gets x number of credits. (same applies to both slower systems)
Based on my rudimentary calculations above it appears that Seti and Einstein are granting the same amount of credit for the work done. At the very least it doesn't look like I'm wasting my time and resources for next to nothing in credits granted. Not knowing the difference in granted credit between both projects before the adjustment I thought it was fair. Is it fair now? I honestly don't know. I do know this for a fact though, I devote my time and effort to Einstein for a lot of other reason besides credits. To keep the credit allocation the way it is now or bring it back to what it used to be is not the priority that bothers me as much as bringing the workunit sizes down to a manageable time frame, both in the amount of time it takes to process and for being able to receive the granted credit in a timely manner without having to wait weeks for someone else to return their results. This I leave to the project folks. In the meantime I'm just going to sit back and relax with a fresh cup of coffee.

Arion
____________

As we are now seeing a reduction in credits/time compared to S5R1. could it be that the complaints we are seeing be self inflicted.
There are quite a few users that chase credits, so as S5R1 gave the most/hr they migrated here and therefore enabled S5R1 to be completed much faster than otherwise. This possibly had a knock on effect to S5R2 in that in had to be introduced earlier, without the normal testing, and some are now feeling the effect of that.

Just finished my first S5R2 workunit @ 49 hours. Pages and pages of printout about checkpoints etc. I'm accepting the time, but is this stuff in the results to be expected?

____________
Mrs Miggins - A lady of uncommon refinement

Well I'm probably going to choke on crow with this but I Took all my systems down from Einstein yesterday and went back to running Seti as my primary project. I only run standardized clients on both projects.

And I'm desperately trying to make my Debian cooperate with the cheap trashy modem my new ISP provided... as that host is an AMD, I'd really prefer running Linux, but it simply won't let me go online and get BOINC and a few WUs...

Well, if it's an option for you and you have 256kB memory left, you could just run Windows and install VMWare Player, like I did. You could do a small Linux installation, get 35% more credits and 35% more benefit for the science. :-)
And the best, if one knows how to do it, it takes just about an hour.

If you like to do so and you have any questions, then come over to our team forum - it's a lot easier for me to write in my mother language. ;-)

cu,
Michael

Who cares about the other projects points etc.
I do run only einstein and i want the old units/points.
S5R2 as they made it its a bullshit, nobody will stay to the project with units that needs one day to be finish and take 200 points.
If einsten@home bosses want to stay alone and crunch with their computers only, continue like this.
Its the end of the project.
Be serius and bring back the old unit system, otherwise bye bye.
____________
----------------------------
http://www.boincsynergy.com/stats/teams.php?team=1290&project=eah
----------------------------

Who cares about the other projects points etc.
I do run only einstein and i want the old units/points.
S5R2 as they made it its a bullshit, nobody will stay to the project with units that needs one day to be finish and take 200 points.
If einsten@home bosses want to stay alone and crunch with their computers only, continue like this.
Its the end of the project.
Be serius and bring back the old unit system, otherwise bye bye.

Who cares about the other projects points etc.
I do run only einstein and i want the old units/points.
S5R2 as they made it its a bullshit, nobody will stay to the project with units that needs one day to be finish and take 200 points.
If einsten@home bosses want to stay alone and crunch with their computers only, continue like this.
Its the end of the project.
Be serius and bring back the old unit system, otherwise bye bye.

People hate change, I find it a new challenge. I am happy crunching Einstein, even if the credit changes. It's a worth while project, and deserves to get done. I have even added more resources to it.

Lots of the new ones have been granted credit.

I do not see what all the hype is about. Things change, you either live with change, or live a very boring life.

____________

Personally, I absolutely agree with you there. You just have to keep in mind that not everyone running BOINC has to be IT savvy, and for someone who wants to just "attach and forget" a project, the current situation is certainly difficult. How should someone who barely gets by with Windows switch his AMD hosts to Linux in order to crunch efficiently, for example? Plus, there are people who have the skill, but simply not the same motivation as you and I. So the current situation is certainly not what one would call ideal, but you're right, it's not as bad as some people make it look. As for adding extra resources- so have I. Someone has to make up for the people having problems or leaving (temporarily) for other projects ;-)
____________

Hello.
The claimed credit for this WU does not show up in my pending-list.
The initial replication is also 3 instead of the standard 2 whereas the minimum quorum remained at 2.
Any one else experienced such a thing?
What is going on?
____________

I personnaly don't think skill with computers is the issue here.

1.) I have mentioned it before, you just need to change the deadlines for less sophisticated computers.

2.) For Program Developers: Who cares about credits honestly? Its not going to get me a new car or groceries for that matter. People honestly are kinda superficial and will want some kind of reward (i.e. credits) for helping out. Here is a suggestion that might make everyone happy. Why don't credits go based on the amount of time spent on WU per user? It would be a very complicated implementation but if one guy has a P3 450 MHz processor and is only producing one unit every 9 1/2 days, why not give him 1900 credits? Thats roughly 200 credits a day that he didn't earn compared to his Pentium and AMD bretheren capable of multiple WU's a day. Maybe you guys talk and make a standard of how many credits a CPU Second is, or a CPU Hour, or a CPU Day. Maybe get bonus credits for continuous labor? Now, I personally don't pay much attention to credits, but I like to collect results myself and browse through them, especially from SETI and Einstein (Sky programs rock!) The problem is you will lose a lot of crunch time from the users that feel they aren't not making the best of their time (don't ask me) with their CPU.

3.) For Other Users: If you really want credits and are credit hungry... get a new computer. Thats the only point of obtaining credits isn't it? So you can compete with everyone else? Compete with the best and own the best. It takes me 8-12 hours per wu depending on which one I get. And that is on one core of my dual core running Vista. I knock out about 4 E@H WU's a day or 2 Einstein@Home WU's and about 8 Seti@Home WU's.

4.) Enough with the "Linux is the way" crap. Quit bashing people who use Windows. If Linux was so much better why did they have develop VirtualPC? A lot of people donate the computer they use (yes people don't just own computers for distribution) in order to help out. Who cares if you get a few extra CPU cycles or less cycles for that matter? I know I don't, but lately I will watch a dvd on my ps2 instead of my laptop just to make sure I am not wasting the cpu cycles for BOINC.

5.) Seti@Home currently does credits in the 49-65 per WU right now. Switch to that project if you want faster credits.

6.) IF you run an old computer left to just be on and "distribute" why the hell do you care if takes 9 days? (If this is about deadlines see #1).
____________
Windows Vista Home Premium
Intel Core 2 Duo T7200 @ 2.00 GHz 4MB L2 Cache
2048 Dual Channel DDR2 553 MHZ
ATi Mobile Radeon x1400 256 MB DDR2 x16 PCI-E

Hello.
The claimed credit for this WU does not show up in my pending-list.
The initial replication is also 3 instead of the standard 2 whereas the minimum quorum remained at 2.
Any one else experienced such a thing?
What is going on?

Hello.
The claimed credit for this WU does not show up in my pending-list.
The initial replication is also 3 instead of the standard 2 whereas the minimum quorum remained at 2.
Any one else experienced such a thing?
What is going on?

Who cares about the other projects points etc.
I do run only einstein and i want the old units/points.
S5R2 as they made it its a bullshit, nobody will stay to the project with units that needs one day to be finish and take 200 points.
If einsten@home bosses want to stay alone and crunch with their computers only, continue like this.
Its the end of the project.
Be serius and bring back the old unit system, otherwise bye bye.

I care about other projects points. It has been annoying me recently that E@H has moved into the top spot on my personal stats. It has never had more access to my hosts than SETI, so it should not have been able to pass SETI.

@FearCain: What was said here about having to run boxes under Linux was imo not "bashing Windows users". While I personally like using Linux, I also use Windows quite a lot for some things, and I can perfectly well understand if someone prefers using Windows. I'm not a fangirl or sth. Nor was this what the others here were about. I mean, I know some people think everyone using Windows is stupid or so, but what was said in this thread was simply that atm, the science app has problems under Windows, mostly on AMD CPUs, and that therefore, if you want to get the maximum crunching power, it is recommended to run Linux if you can and want. Nothing else ;-)
____________

Thanks a lot, Michael :-) I might do that, but atm my favorite idea (which came to me in an extremely boring maths class) is to simply boot my desktop from a Linux live cd (they are made to cooperate with all sorts of strange setups, so maybe that'll get along okay with the modem, and if not... you know Backtrack und my nooblike neighbors? ;-) ), I mean, the Linux BOINC version is a shell script, so I wouldn't have to install anything, simply mount a partition of my hard disk to store the client and WUs on. Maybe that'll work. I'll try it out as soon as I get home from uni and language class.
If I can't get one of the live cds to run and go online in a reasonable amount of time I'll probably try the VMWare approach, and if I have problems with that I'll gladly accept your offer. It's about time for me to finally register at the Heise boards instead of just reading them anyway ;-) Btw, I have two Gigs of RAM (recently upgraded for a uni project involving Rainbow Tables) so I should be able to run a VM okay.
Somehow, I'll get that box to run under Linux at least on a temporary basis. Anything for the science :-)

Well I'm probably going to choke on crow with this but I Took all my systems down from Einstein yesterday and went back to running Seti as my primary project. I only run standardized clients on both projects.

Arion,

this is your problem. While the E@H was a highly optimized client with little potential for optimization left, the official SETI@Home client is not optimized at all. There are optimized SETI clients out there and these crunch a workunit almost twice as quickly as the standard client. If you put these numbers into your calculation you see that (considering the same level of optimization) both projects grant the same credit/time.
Now the first release of the E@H client is not as optimized as the one before, but this will change with time.

Well I'm probably going to choke on crow with this but I Took all my systems down from Einstein yesterday and went back to running Seti as my primary project. I only run standardized clients on both projects.

Correct me if I'm wrong, but is the primary issue regarding S5R2 an AMD/Intel issue, or is there something else I'm missing. Although all of the the S5R2 jobs I've completed are still pending, there is without a doubt a decrease in credits per time with S5R2, which is fine if everyone is playing by the same rules so to speak. I have been trying to find anyone who has yet to received a single credit for any of their completed work. I realize they have to have several completed results to verify the accuracy, but this is about as long as I have had the same finished job