Message boards :
Number crunching :
Much slower processing with SETI@home ver 4.5
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next
Author | Message |
---|---|
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
> Hi All, > I have been reading the post here and began wondering if it was not > intentional to make the pdb file to include the debugging code because of the > problems they have had with software, they now have the ability to track > crashes they can't watch or determine how happen. IMO I would rather see the > debugging in the program and our files to take longer, if it keeps the project > active without "hiccups". > Again I am not from a computer background and I am trying to understand. > If I understand correctly the pdb. file is longer because it includes the > debugging code, so wouldn't that make the WU's process more slowly because > they have additional code "Reviewing" the process. > > Thank You In Advance. > > Have A Great Day And A Better Tomorrow! > > Rocky Cudd BETA lived through stack walker releases of the code. These took 10 times as long as normal, not just a little longer. There are high angle and low angle WUs that take different amounts of time (I can't remember which is which at the moment). Perhaps we have some of the other type for a while. In any case the credit is granted based on CPU time, not on WU count. |
Misfit Send message Joined: 21 Jun 01 Posts: 21804 Credit: 2,815,091 RAC: 0 |
Well I still have about 6 days of setiathome 4.03 WU's cached so if this version is really crunching faster I'll wait until I run out before upgrading. |
eberndl Send message Joined: 12 Oct 01 Posts: 539 Credit: 619,111 RAC: 3 |
You only upgrade the BOINC core manually; all the clients update as needed the next time you contact the server to d/l units. So I'd say you'll be upgrading in 6 days or less =-) <br /> Feel free to take a look inside my brain |
gregh Send message Joined: 10 Jun 99 Posts: 220 Credit: 4,292,549 RAC: 0 |
> Has anybody else noticed -- my processing times per work unit have almost > doubled since the upgrade to SETI@home ver 4.5. Can anybody explain why? > > Which AV have you got? I know the latest Mcafee and lBoinc 4.10 and Boinc 4.11 dont get on. It may apply to other AV progs and latest Boincs too. No probs with Boinc 4.09 and Mcafee. Suggest you go back to Boinc 4.09 and check. If you are using that one anyway, check to see what is using the most processing time. It may be Seti and your AV are going off at each other. |
JAF Send message Joined: 9 Aug 00 Posts: 289 Credit: 168,721 RAC: 0 |
Crunching 30 - 40 percent longer per WU is fine, if it's for debugging or more intense computations. But crunching 30 - 40 percent longer because of a compile error or untested software is a waste of energy. I hope it's the former. |
Mike Send message Joined: 17 Feb 01 Posts: 34265 Credit: 79,922,639 RAC: 80 |
Hi AMD Athlon 2800+ nforce2 512MB dual Ram Win XP home. Seti 4.03 2.3 houres. Seti 4.05 4.1 houres. very heavy. greetz Mike |
mlcudd Send message Joined: 11 Apr 03 Posts: 782 Credit: 63,647 RAC: 0 |
> BETA lived through stack walker releases of the code. These took 10 times as > long as normal, not just a little longer. There are high angle and low angle > WUs that take different amounts of time (I can't remember which is which at > the moment). Perhaps we have some of the other type for a while. > >Hi John, I appreciate your response, and I had some additional questions. We know that credit is based upon CPU time, but wouldn't our credit basically be the same as it is now since everyone has the 4.05 core? I was just thinking that alot of people are recieving small amounts of credit for many workunits being done, but if my thinking is correct even though the units take longer to process,credit will still average out in the long run? Like a bigger piece of credit for fewer WU's. Thanks Again. Happy Crunching! Warm Regards, Rocky Cudd P.S. I forgot one thing, my slow box (Genuine Intel Pentium Microsoft Windows Millennium , (04.90.3000.00) was doing workunits at 12.25 hours each, they now run at 24.66 hours each.A tad more than double. My fastest box that runs the AMD Athlon, normally does a WU in 4.35 hours, now in 5.93, and finally my third box with a 2,8 Intel celeron was running at 5.37 hours each and now they run at 7.88 hours. Are the differences in the amount of time each one takes to process still due to how the processors caches handle the WU? |
Crunch3r Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0 |
> > BETA lived through stack walker releases of the code. These took 10 > times as > > long as normal, not just a little longer. There are high angle and low > angle > > WUs that take different amounts of time (I can't remember which is which > at > > the moment). Perhaps we have some of the other type for a while. > > > >Hi John, > I appreciate your response, and I had some additional questions. We know that > credit is based upon CPU time, but wouldn't our credit basically be the same > as it is now since everyone has the 4.05 core? I was just thinking that alot > of people are recieving small amounts of credit for many workunits being done, > but if my thinking is correct even though the units take longer to > process,credit will still average out in the long run? Like a bigger piece of > credit for fewer WU's. > > Thanks Again. > > Happy Crunching! > > Warm Regards, > > Rocky Cudd > > P.S. I forgot one thing, my slow box (Genuine Intel Pentium Microsoft Windows > Millennium , (04.90.3000.00) was doing workunits at 12.25 hours each, they now > run at 24.66 hours each.A tad more than double. My fastest box that runs the > AMD Athlon, normally does a WU in 4.35 hours, now in 5.93, and finally my > third box with a 2,8 Intel celeron was running at 5.37 hours each and now they > run at 7.88 hours. Are the differences in the amount of time each one takes to > process still due to how the processors caches handle the WU? > > src="http://boinc.mundayweb.com/cpdn/stats.php?userID=380&trans=off"> > There is seriously something going on in the wrong direction. I´ve got a lot of problems with boinc 4.09 wich stops processing and 4.10 even crashed several of my machines here. Processing a WU with seti 4.05 on my dual pIII 500 went from 11h to 21h that can´t be true. So i think somebody (Rom) should explain us what happend and why. |
Siran d'Vel'nahr Send message Joined: 23 May 99 Posts: 7379 Credit: 44,181,323 RAC: 238 |
> Has anybody else noticed -- my processing times per work unit have almost > doubled since the upgrade to SETI@home ver 4.5. Can anybody explain why? > > Hi Keith, This is rather reminiscent of what I noticed over at "Classic" SETI back in '99 or '00. They came out with an upgrade and my WU more than doubled in processing time. I don't remember if I was still running my 486 or if I was running the Pentium 166 at that time. But, I thought it was very weird. Fortunately, for now, I'm still running v4.03 SETI client with v4.09 core client. I still have quite a number of WUs needing v4.03. L8R.... --- Rick A. - BOINCing right along now.... It can only get better! "There is no fate except that which we create for ourselves." Live Long and Prosper.... |
RockyIII Send message Joined: 25 May 99 Posts: 28 Credit: 18,865,236 RAC: 0 |
With BOINC ver. 4.09 and setiathome 4.05, work units are processing in about 3 hrs. 10 mins. rather than about 2 hrs. 20 mins. as they did with previous versions. This is on a Pentium 4 3.0 GHz, 1 GB RAM, HT off, Windows XP. Rocky <a> [/url] |
Geek@Play Send message Joined: 31 Jul 01 Posts: 2467 Credit: 86,146,931 RAC: 0 |
If I remember correctly about the time Seti went live, back on the Beta project there was a success rate of about 80%. In other words about 20% of the work units failed with an error. They are probably the work units that complete in under 5 minutes. Now the debug code has been put back in along with the PDB file in order to further debug the failing work units. Hence the longer times to complete a work unit. After this is done and a satisfactory update issued I would say that the debug code would be removed and we would be back to normal crunch times. This seems like a good thing to me. Since everyone is crunching at the longer times no one person is penalized. Only difference is that it takes more time and therefore requests more credit than would be normal for the completed work units. Flame proof suit ready for your reply's. |
Prototype Send message Joined: 3 Apr 99 Posts: 67 Credit: 497,118 RAC: 0 |
> If I remember correctly about the time Seti went live, back on the Beta > project there was a success rate of about 80%. In other words about 20% of > the work units failed with an error. They are probably the work units that > complete in under 5 minutes. Now the debug code has been put back in along > with the PDB file in order to further debug the failing work units. Hence the > longer times to complete a work unit. After this is done and a satisfactory > update issued I would say that the debug code would be removed and we would be > back to normal crunch times. This seems like a good thing to me. Since > everyone is crunching at the longer times no one person is penalized. Only > difference is that it takes more time and therefore requests more credit than > would be normal for the completed work units. > So why there no official word on this (that I can see). Any person running a client that doesnt check here is likely to think something is wrong with their setup or system. A simple one line on the home page to inform users that debugging is back is all it takes. Lets not get back to the mushroom farm mentality (ie keep everyone in the dark) the SETI team was just starting to shake off. |
slavko.sk Send message Joined: 27 Jun 00 Posts: 346 Credit: 417,028 RAC: 0 |
I think that we are "cooking from water" ... We are comparing 2 different version of the worker client sofware and 2 differnet WU's. It is wrong, I think. If something changed in client worker from 4.03 to 4.05 AND something changed in WU's then the result time is different. And something changed in both, for sure, since 4.03 WU's MUST be proceed with 4.03 worker client. Crunch timing chnaged alreay a few times in SETI classic, somebody already mentioned it here. I don't think that something is wrong. CPDN takes 465 hours to complete. My box made 50 hours of CPDN till now and still only 7.33% completed. And one thing, reported time is reported by client, did anybody check it on watches Is that reported time right one? Was it reported right before? |
CyberGoyle Send message Joined: 2 Jun 99 Posts: 160 Credit: 3,622,756 RAC: 26 |
> My processing time on a pentium 4 3.2 went from 3.5 hours on the 4.3 units to > 5 hours on the first 4.5 unit I downloaded to 6 hours on the second 4.5 unit I > just crunched. Granted my processor is hyperthreaded and I am doing two wu's > at once but it slowed way down since the upgrade. Plus, the wu keeps going > when it says 100.000% done. > > Good to know someone else has the same specs as me - 3.2Ghz P4 HT, WU's took about 3.5 hours and now take 5 hours 20 minutes on Seti v4.05. Also, I have noticed that the WU's keep crunching for about 10 minutes after the status bar reaches 100% complete. Prolly just a minor bug in calculating completion time. I suspect that Seti v4.05 is performing more calculations on each WU than previous versions (better science). Rock on! <a> |
Harri Liljeroos Send message Joined: 29 May 99 Posts: 4191 Credit: 85,281,665 RAC: 126 |
I've had WU's (26ap04aa...) running on Seti 4.05 that have taken exactly the same time than with 4.03 (3:44 on P4 2.6GHz). The latest WU's I've got (04mr04aa...) are now taking about 4:44. So I think that the WU's are causing the longer calculation times, not the version of SETI@home. |
Anthony Brixey Send message Joined: 24 Jun 00 Posts: 102 Credit: 1,757,916 RAC: 0 |
It would be nice if there was a bench mark WU that is re-issued for each version of the Seti client that can be used to optimise systems and make sure they are working correctly, it would also remove the question of is the WU the cause of the longer completion times or is it the version. Anthony |
Petit Soleil Send message Joined: 17 Feb 03 Posts: 1497 Credit: 70,934 RAC: 0 |
Another simple solution would be SETI telling us what's going on ?!? |
Ulrich Metzner Send message Joined: 3 Jul 02 Posts: 1256 Credit: 13,565,513 RAC: 13 |
> Another simple solution would be SETI telling us what's going on ?!? > > What drugs are you on, considering such a simple solution ;) greetz, Uli |
WildWeasel Send message Joined: 2 Jun 99 Posts: 5 Credit: 485,315 RAC: 0 |
> It would be nice if there was a bench mark WU that is re-issued for each > version of the Seti client that can be used to optimise systems and make sure > they are working correctly, it would also remove the question of is the WU the > cause of the longer completion times or is it the version. > > Anthony > > <img> src="http://boinc.mundayweb.com/seti2/stats.php?userID=1580&trans=off"> > Try this : 1) Download the seti application source code archive (eg zip file). 2) In the \seti_boinc\client\test_workunits\ directory is a file called reference_work_unit.sah 3) Create two new directories called for example \seti403 and \seti405 4) Copy the reference work unit to each new directory calling it work_unit.sah 5) Copy the relevant seti 4.03 and 4.05 executables and .pdb files to each new directory 6) Ensuring that you run eactly the same test conditions eg hyperthreading off, no other programs running etc, run each executable in turn noting how long it takes. 7) Compare run times and stop whinging ! 8) Repeat for each new version of the seti client as it comes out... The Weasel |
Igor Klajo Send message Joined: 14 Oct 01 Posts: 12 Credit: 112,875 RAC: 0 |
Yup . . . slow like a snail It took me metween 2:20 and 2:38 to complete one WU and now it takes up to 4:20 IGOR(m Russian) Russian form of IVOR Pronounced: IE-vor, EE-vor Other forms: Ingvar (scandinavian); Ingvarr (norse)(origin form of the name) Ing was the Norse God of peace and fertility Derived from the Old Nors |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.