Message boards : Number crunching : CPU usage
Previous · 1 · 2 · 3 · 4 · Next
Author | Message |
---|---|
Mod.Sense Volunteer moderator Send message Joined: 22 Aug 06 Posts: 4018 Credit: 0 RAC: 0 |
Joshua, what BOINC version is involved? I see you have several hosts. Is this occurring on some hosts more often then others? Rosetta Moderator: Mod.Sense |
michaelmastro![]() Send message Joined: 11 Oct 05 Posts: 51 Credit: 1,530,918 RAC: 0 |
Joshua, what BOINC version is involved? I see you have several hosts. Is this occurring on some hosts more often then others? This is Michael - and it only happens on the Vista machine - Mac has no problems. |
![]() Send message Joined: 6 Oct 07 Posts: 5 Credit: 5,571,366 RAC: 0 |
Yes it is. The problem concerns the AMD-machines only. Settings and OS are the same. BOINC version is 6.6.31 (host 1082849) and 6.6.36 (host 902983). |
![]() Send message Joined: 6 Oct 07 Posts: 5 Credit: 5,571,366 RAC: 0 |
Good news: now the problem occurs also on the newly installed Intel-system (BOINC 6.6.36). Any suggestions? I can't believe this is only happening to me... :( |
mikey![]() Send message Joined: 5 Jan 06 Posts: 1895 Credit: 8,893,974 RAC: 711 ![]() |
Good news: now the problem occurs also on the newly installed Intel-system (BOINC 6.6.36). Any suggestions? I can't believe this is only happening to me... :( Downgrade to version 6.2.19 if you are not using CUDA or version 6.5.0 if you are. The 6.6.? versions all seem to have 'issues' and the powers that be are "working on them". All versions of Boinc can be gotten here: http://boincdl.ssl.berkeley.edu/dl/ ![]() |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2059 Credit: 40,463,815 RAC: 2,908 ![]() |
I say this totally respectfully, as I am addressing the programmers of Rosetta: it's because their programs have more errors in them than the 1985 Los Angeles Dodgers infield. I told you, shoddy code... of course nobody's going to know why you can't detach... not unless Rosetta will open their source code and present it to the community, preferably compiled -and- decompiled. Otherwise, you'll continue to have program cancer, which Rosetta ironically tries to treat :p You've already seen the metastasis in some areas... After a cursory look at some of your timed-out WUs that've been taken up and completed by others, who are also running Vista, they seem to get returned pretty quickly. That kind of implies it's not the Rosetta code (else it'd happen to everyone) and not Windows (ditto) but something to do with the machine and maybe the other programs running on it and the priority they run at. I know Rosetta runs at lowest priority - do other projects? Is that the difference? I use a Sidebar utility that shows all 4 cores running at 100% and only a couple of minutes (over a couple of hours) between CPU time and elapsed time. I accept that the display of elapsed time rather than CPU time is a distraction and I don't understand the motivation behind that change, but I'm just not seeing the problem described. I'd suggest your issue is closer to home. What do you think it is that makes your machines report this? ![]() ![]() |
![]() Send message Joined: 6 Oct 07 Posts: 5 Credit: 5,571,366 RAC: 0 |
Thank you. Installed BOINC 6.5.0, but the problem persists - only 50 % of CPU-utilization, but now time stops running. I think this is more a problem of the minirosetta task as all machines were running well for a long time and nothing was changed meanwhile. |
mikey![]() Send message Joined: 5 Jan 06 Posts: 1895 Credit: 8,893,974 RAC: 711 ![]() |
Go into the Boinc manager and click on advanced then preferences and then under processor usuage what does it say in the bottom 3 boxes? Mine is 60 minutes, 100% and 100%. This is a by machine setting not a global setting so be careful if you close the box by saying ok, if you make no changes you should probably click cancel instead. ![]() |
![]() Send message Joined: 6 Oct 07 Posts: 5 Credit: 5,571,366 RAC: 0 |
My settings are identical. I also checked the "allow computation when PC is used" (I translated it because I use the german version, maybe the wordtext is somewhat different), but this also had no effect. |
![]() Send message Joined: 16 Jun 08 Posts: 1227 Credit: 14,140,275 RAC: 477 ![]() |
I'm having this issue occasionally with Rosetta on my dual-core x64 Vista machine. Not frequently enough to warrant detaching, but enough to warrant leaving a message. An idea the developers should check: Are workunits compiled for 32 bits capable of running under x64 Vista at addresses beyond the first 4 GB of memory? If not, they need to ask for future versions of x64 BOINC to have two different sets of restrictions on how much memory can be used - one set for workunits compiled for 32 bits, and one set for all workunits, including those compiled for 64 bits. Also, in this case, the more memory-hungry BOINC projects need to accelerate their move to workunits compiled for 64 bits. Currently, my 64-bit Vista computer has problems getting the boincmgr.exe program to start; right-clicking on its icon fails to bring up a menu of next steps. The task manager programm will still run, and shows that only slightly more tham 50% of the 8 GB of memory are in use, but that could be enough to exhaust the usable memory space for any programs that can run only in the first 4 GB. My 64-bit Vista computer is currently set to participate in about 12 BOINC projects - easily enough to get enough workunits to fill half of its memory. I think my 64-bit computer is running BOINC 6.6.28, but currently don't see how to check without rebooting it, since I can't start boincmgr.exe to display it. I know it's using Vista SP2. Also, I suppose you know that the lockfile problem (mentioned in many threads) is a known cause of low CPU utilization; once started, it essentially shuts down the CPU usage in any workunits assigned to the same thread as the one that started the problem, until boinc.exe (not just boincmgr.exe) is shut down (not just suspended), and then restarted. I've found Rosetta@home and RALPH@home to be the two BOINC projects which seem to be most likely to start the lockfile problem, but a number of other BOINC projects are affected once the problem is started. |
![]() Send message Joined: 16 Jun 08 Posts: 1227 Credit: 14,140,275 RAC: 477 ![]() |
I finally was able to restart boinc.exe on my x64 Vista machine by killing the boincmgr.exe process, then restarting it manually, then using the normal suspend and restart process. It is using the 6.6.28 version of BOINC. I saw no sign that any of the workunits had encountered the lockfile problem. The restart dropped the amount of memory used to 2.09 GB, well under the limit BOINC is allowed to use on that machine. Not much CPU time was lost by the restart. I plan to wait for an answer to whether workunits compiled to run under 32-bit Vista can run in memory above 4 GB under x64 Vista before deciding whether to change BOINC versions. Also, my last message should have referred to workunits assigned to the same slot, not the same thread, as the one that caused the lockfile problem. |
![]() Send message Joined: 16 Jun 08 Posts: 1227 Credit: 14,140,275 RAC: 477 ![]() |
This is concerning Rosetta on my laptop, running Vista. When any other projects are running, my CPU usage is in the 95-100% range. Recently, when Rosetta is running the CPU usage drops dramatically - yesterday it was down to 45-50%, and today it's under 10%. All my preferences are set to use 100% of processors. Any ideas why Rosetta is only using 10% instead of 100%? Do the other projects happen to require less memory than Rosetta does? If so, they could continue running well when Rosetta starts spending most of its time swapping various parts of the program in and out of memory instead of using CPU time. Also, the lockfile problem is known to reduce CPU usage, but it reduces the usage for any CPU core trying to run a workunit in a slot with this problem nearly to 0%, and not just when a Rosetta workunit is running. Does rebooting the laptop make the problem disappear for a while? Also, you could use Task Manager to check whether some CPU cores are nearly stopped, but others still run at near full speed, the next time you see the problem. |
![]() ![]() Send message Joined: 1 Dec 08 Posts: 28 Credit: 225,332 RAC: 0 |
I'd suggest your issue is closer to home. What do you think it is that makes your machines report this? My machine (I run 1, with 2 separate CPU's, and a friend collaborates with me on my account) isn't the problem. Here's the thing: I've gone over to POEM. No problems; no hanging such that I must reset BOINC; no discarded, shredded, folded, spindled, mutilated, returned for full refund, or any other kind of program problem. My WU's come in, get computed, and go back out. They get reported in 24 hours unless I update first, per BOINC specs. I have no problems. Thus, whatever it is that makes BOINC, Rosetta, and my CPU's not get along, I'm not sure, and I have little reason to care *shrug*. I love Rosetta and what it does; however, POEM is very important, too. I have contributed to it in the past. It does a lot of what Rosetta does, plus adds a facet of cataloging. Works for me and many others. I can say this: Rosetta never damaged or messed up any other prog on my laptop, nor did it adversely affect the BOINC client per se (it didn't change its coding or cause it not to recognize other projects, e.g.). It just performed substandardly. I can also say: I'm not the only one presenting with these symptoms. My computer has an issue with x; someone else is complaining about y; and another has trouble with issue z. Read the thread back to yourself. The problems here aren't a faulty diode, clocking, or some scheduling or priority conflict. It's different WU's, different errors. The efficacy of the code is lacking, and depending on what the bugs come into contact with on a system, they're manifesting in a plethora of ways. This is 30 years of using computers talking; you feel it after a while, the crux of the problems. I am a user and user-end whiz kid (36, an old kid). I can't program too well, and I'm not good at reading machine or coding language. I leave it to someone who can do all that to work out the bugs. You ask me what else it could be in my system? I don't know. I might not know if the code jumped right out on my screen in a box saying "This is why Rosetta is not working." To guess at it might send those in the know on a fool's errand, and I would rather not have that... My best to the programmers... may they debug their project... :) The lovely lady you see isn't I, but Hayley Westenra, a classical crossover singer from Christchurch, NZ. There is no known voice as hers. Check her out- she's seraphic. |
![]() ![]() Send message Joined: 17 Sep 05 Posts: 68 Credit: 3,565,442 RAC: 0 |
This is concerning Rosetta on my laptop, running Vista. When any other projects are running, my CPU usage is in the 95-100% range. Recently, when Rosetta is running the CPU usage drops dramatically - yesterday it was down to 45-50%, and today it's under 10%. All my preferences are set to use 100% of processors. Any ideas why Rosetta is only using 10% instead of 100%? This is an interesting problem that I have experianced from time to time on my I7 windows vista machine running 6.6.36 BOINC. This machine has 3 gigs of memory and runs 8 rossetta mini 1.80 workunits at a time. Yes, it is running hyperthreaded. It will for what appears to be no reason at all throttle down the CPU from 90 to 50 percent usage. Acording to Task Manager. Causing the exact same problem you have on 4 of 8 threads. This machine will normally do 8 workunits in 3 hours. When this problem occurs 4 wu take anywhere fron 12 to 50 hours. What I have to do to correct this problem is to go to the ADVANCED dropdown menu in boinc manager. Click on Run CPU Benchmarks. Run CPU Benchmarks shuts down the wu computations, runs the benchmarks then restarts the 8 workunits at 100 percent cpu usage according to TASK Manager. The eight wu's then complete in 3 hours. I have another I7 which just started running rosseta having this problem on a daily basis. This machine is the exact same setup as the older machine except it has 6 gigs of ddr3 memory. ![]() |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2059 Credit: 40,463,815 RAC: 2,908 ![]() |
I'd suggest your issue is closer to home. What do you think it is that makes your machines report this? Really... After another glance at your old WUs and seeing that they've run quickly and successfully on XP, Linux and Vista for other people it's apparent to me the problem still remains closer to home. Having an extra dozen years on you my first port of call would be to upgrade to Vista SP2 and upgrade Boinc to 6.6.36 to ensure that your complaints hadn't already been fixed here. Even then, the timeline here since June 10th shows there's still a lot of unreleased updates to come. Then check your startup programs and adjust with MSConfig (taking online advice through Google) to ensure you don't have unnecessary programs running that may introduce conflicts. Then check your services with advice here. Even if that doesn't help your system will run pretty light and you can more confidently bat away criticisms that it's a quirk of your personal setup rather than your sweeping criticism of the project here. ![]() ![]() |
[AF>france>pas-de-calais]symaski62 Send message Joined: 19 Sep 05 Posts: 47 Credit: 33,871 RAC: 0 |
05/07/2009 14:34:00 rosetta@home update requested by user 05/07/2009 14:34:05 rosetta@home Sending scheduler request: Requested by user. 05/07/2009 14:34:05 rosetta@home Requesting new tasks for GPU 05/07/2009 14:34:10 rosetta@home Scheduler request completed: got 0 new tasks 05/07/2009 14:34:10 rosetta@home Message from server: Server error: can't attach shared memory BOINC 6.6.36 <= FAIL 1Go RAM |
darengosse Jean-Paul![]() Send message Joined: 9 Jun 06 Posts: 18 Credit: 259,459 RAC: 0 |
Hello. Perhaps here is not the good Forum, but my problem is that I receive permanently, a message in the BOINC 6.4.7, coming from the site of Rosetta, to download version 6.6.36. Me I do not want any for the moment. Indeed, with this new version it is not possible to configure it to use the maximum of the file of exchange which is on my Computer of 10317084 KB. Me hope for a solution with this problem which can be used for other users…. ![]() ![]() |
michaelmastro![]() Send message Joined: 11 Oct 05 Posts: 51 Credit: 1,530,918 RAC: 0 |
I detached from Rosetta; waited a few days and then reattached. Went out of town for three days and come back to find Rosetta is much worse than before - work showing 72 hours, 75% complete and 30 hours remaining. So I've detached from Rosetta, again. When they get the bugs worked out so I'm not wasting cpu time, I'll reattach. Somebody let me know when it's safe to go back in the water... Oh, and for the three days I was gone, Boinc was stuck on Rosetta the whole time and never switched projects. What a mess. ![]() |
mikey![]() Send message Joined: 5 Jan 06 Posts: 1895 Credit: 8,893,974 RAC: 711 ![]() |
I detached from Rosetta; waited a few days and then reattached. Went out of town for three days and come back to find Rosetta is much worse than before - work showing 72 hours, 75% complete and 30 hours remaining. So I've detached from Rosetta, again. When they get the bugs worked out so I'm not wasting cpu time, I'll reattach. Somebody let me know when it's safe to go back in the water... Your problem in Windows is that you are using one of the 6.6. ? Boinc versions, the 6.2.19 version is MUCH better, unless you use a CUDA device and then you should use version 6.5.0. Scheduling is really, REALLY bad in the new 6.6.? versions!!! They are 'working on it'! As for how long the units take, that seems to be a problem that is being seen by a lot of different people, although FAR from all, that may end up being a programming glitch. Only time will tell and I would guess that the programmers have looked at it and found nothing so far. Kinda like when your car makes this sound until the mechanic is there and then the car is silent!! ![]() |
mikey![]() Send message Joined: 5 Jan 06 Posts: 1895 Credit: 8,893,974 RAC: 711 ![]() |
Hello. Perhaps here is not the good Forum, but my problem is that I receive permanently, a message in the BOINC 6.4.7, coming from the site of Rosetta, to download version 6.6.36. Me I do not want any for the moment. Indeed, with this new version it is not possible to configure it to use the maximum of the file of exchange which is on my Computer of 10317084 KB. Me hope for a solution with this problem which can be used for other users…. This is a standard message because there were people that said they never heard a newer version was available. Use the version you like, if the project you like requires a higher version it will be made known on their boards that you will have to upgrade. ![]() |
Message boards :
Number crunching :
CPU usage
©2024 University of Washington
https://www.bakerlab.org