6.3.21
Message boards :
Graphics cards (GPUs) :
6.3.21
Message board moderation
Previous · 1 · 2 · 3 · Next
| Author | Message |
|---|---|
|
Send message Joined: 16 Aug 08 Posts: 87 Credit: 1,248,879,715 RAC: 0 Level  Scientific publications     
|
I switched to 6.3.21 yesterday and it seems to be having a bit of teething problems. I was also in the process of switching from 3x+1 as my primary CPU project to abc and simap happened to hit. Anyway, 6.3.21 downloaded 8 or 10, 3 hour WU's for ABC when configured for 6 hours max. It later downloaded more. I now have 52 of those 3 hour ABC wu's on a quad core machine. I expect 8 to 10. It also has one 3x+1 partially finished and 7 simap. In case it matters, I have PS3GRID set at 1000 resource, simap at 1000 resource, ABC at 500 resource and 3x+1 at 100 resource (and no new tasks) This may be one of those performance factors resolving things or one of those long term debt things, or it may be a new bug. |
|
Send message Joined: 28 Sep 08 Posts: 27 Credit: 6,201,632,872 RAC: 0 Level  Scientific publications 
|
Here're the messages when I last started the client right after I rolled back to .19 from .21: 01/11/2008 08:23:24||Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [x86 Family 6 Model 15 Stepping 11] 01/11/2008 08:23:24||Processor features: fpu tsc pae nx sse sse2 mmx 01/11/2008 08:23:24||OS: Microsoft Windows XP: Professional x86 Editon, Service Pack 3, (05.01.2600.00) 01/11/2008 08:23:24||Memory: 3.25 GB physical, 7.07 GB virtual 01/11/2008 08:23:24||Disk: 5.00 GB total, 1.30 GB free 01/11/2008 08:23:24||Local time is UTC -4 hours 01/11/2008 08:23:24||Not using a proxy 01/11/2008 08:23:24||CUDA devices found 01/11/2008 08:23:24||Coprocessor: GeForce 8800 GT (1) 01/11/2008 08:23:24||Version change (6.3.21 -> 6.3.19) 01/11/2008 08:23:24|climateprediction.net|URL: http://climateprediction.net/; Computer ID: 864551; location: (none); project prefs: default 01/11/2008 08:23:24|SETI@home|URL: http://setiathome.berkeley.edu/; Computer ID: 4348304; location: (none); project prefs: default 01/11/2008 08:23:24|GPUGRID|URL: http://www.ps3grid.net/; Computer ID: 11763; location: (none); project prefs: default 01/11/2008 08:23:24||General prefs: from climateprediction.net (last modified 19-Oct-2008 21:54:46) 01/11/2008 08:23:24||Host location: none 01/11/2008 08:23:24||General prefs: using your defaults 01/11/2008 08:23:24||Preferences limit memory usage when active to 1663.21MB 01/11/2008 08:23:24||Preferences limit memory usage when idle to 3326.42MB 01/11/2008 08:23:24||Preferences limit disk usage to 1.30GB My connect interval is 0 day, for 1 additional day. Since I don't recall exactly the max value, it may have coincided with (24 hours/day)*(3600 seconds/hour) * 16. This unknown factor of 16 is very much speculative though. On a slightly different topic, I haven't said anything about the disk usage limit before, but I've noticed ever since I started GPUGrid (and thus the development clients) that the disk usage limit is supposedly set to some strange number. I've got a 5GB partition for BOINC and BOINC alone. Usage limit is set to 5GB, leave 0.001GB free, use at most 100%, yet the startup message indicates that the disk limit is set at 1.3GB. However, on the disc usage tab BOINC reports that it is using 3.68GB, with 1.29GB free but unavailable to BOINC. |
|
Send message Joined: 28 Sep 08 Posts: 27 Credit: 6,201,632,872 RAC: 0 Level Scientific publications
|
Correction to my previous message: *16 should be roughly the amount of work BOINC wants to get, *4 was roughly where the max value was. |
 Krunchin-Keith [USA] Send message Joined: 17 May 07 Posts: 512 Credit: 111,288,061 RAC: 0 Level  Scientific publications 
|
Are you getting this error in 6.3.19 ? If only 6.3.21, I need the messages from that version. messages from another version do not help much. Do you get same values from both versions ? Check your preferences at CPdN, make sure they are correct. Try changing them slightly say to 4GB, then update project, and see how much the limit changes. Change them back and update again. Do the limits go back to previous value ? [edit] Sorry, I missed this line, see bold above. What does you o/s report is free, is it the same number ? Mine shows correct value in both o/s and 6.3.21. My memory limits also come out correct for amount of memory and percentages I have specified. I see nothing wrong with the cleint and it's limits. Your limit is set to 1.30GB, because that is all there is free. Something else is taking up the space. My o/s shows 151GB Free on drive C. 11/1/2008 12:20:58 PM||Memory: 2.00 GB physical, 4.85 GB virtual 11/1/2008 12:20:58 PM||Disk: 226.75 GB total, 151.15 GB free << BOINC says same. |
|
Krunchin-Keith [USA] Send message Joined: 17 May 07 Posts: 512 Credit: 111,288,061 RAC: 0 Level Scientific publications
|
I switched to 6.3.21 yesterday and it seems to be having a bit of teething problems. I was also in the process of switching from 3x+1 as my primary CPU project to abc and simap happened to hit. You would get something like this if last time you ran the project they had different length work units. Probably what has happened is the DCF is off. Look at the properties tab and see what the DCF (Duration Correction Factor) is for each project. Ideal is 1. What happens is the server uses that value to adjust amount of work. If it is real low from previous use, it thinks your computer is real fast and send you a lot of work. What will happen is as you run this work, it will self correct. When you run a project all the time you do not have this problem, as if project changes the length or factor they use for an estimate, you system would already have corrected. When not running for a period, you might miss this and then when you resume the numbers are out of whack. BOINC will correct for this and slowly settle back to normal. |
|
Send message Joined: 28 Sep 08 Posts: 27 Credit: 6,201,632,872 RAC: 0 Level Scientific publications
|
This are the message from 6.3.21: 01/11/2008 17:02:28||Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [x86 Family 6 Model 15 Stepping 11] 01/11/2008 17:02:28||Processor features: fpu tsc pae nx sse sse2 mmx 01/11/2008 17:02:28||OS: Microsoft Windows XP: Professional x86 Editon, Service Pack 3, (05.01.2600.00) 01/11/2008 17:02:28||Memory: 3.25 GB physical, 7.07 GB virtual 01/11/2008 17:02:28||Disk: 5.00 GB total, 1.29 GB free 01/11/2008 17:02:28||Local time is UTC -4 hours 01/11/2008 17:02:28||Not using a proxy 01/11/2008 17:02:28||CUDA devices found 01/11/2008 17:02:28||Coprocessor: GeForce 8800 GT (1) 01/11/2008 17:02:28||Version change (6.3.19 -> 6.3.21) 01/11/2008 17:02:28|climateprediction.net|URL: http://climateprediction.net/; Computer ID: 864551; location: (none); project prefs: default 01/11/2008 17:02:28|SETI@home|URL: http://setiathome.berkeley.edu/; Computer ID: 4348304; location: (none); project prefs: default 01/11/2008 17:02:28|GPUGRID|URL: http://www.ps3grid.net/; Computer ID: 11763; location: (none); project prefs: default 01/11/2008 17:02:28||General prefs: from climateprediction.net (last modified 19-Oct-2008 21:54:46) 01/11/2008 17:02:28||Host location: none 01/11/2008 17:02:28||General prefs: using your defaults 01/11/2008 17:02:28||Preferences limit memory usage when active to 1663.21MB 01/11/2008 17:02:28||Preferences limit memory usage when idle to 3326.42MB 01/11/2008 17:02:28||Preferences limit disk usage to 1.29GB 01/11/2008 17:02:28||Running CPU benchmarks 01/11/2008 17:02:28||Suspending computation - running CPU benchmarks 01/11/2008 17:02:59||Benchmark results: 01/11/2008 17:02:59|| Number of CPUs: 4 01/11/2008 17:02:59|| 3131 floating point MIPS (Whetstone) per CPU 01/11/2008 17:02:59|| 6679 integer MIPS (Dhrystone) per CPU 01/11/2008 17:03:00|climateprediction.net|Restarting task hadcm3ivolc_l3d2_2000_80_06001976_2 using hadcm3i version 602 01/11/2008 17:03:00|climateprediction.net|Restarting task hadcm3ivolc_l3c7_2000_80_06001971_1 using hadcm3i version 602 01/11/2008 17:03:00|GPUGRID|Restarting task JouG342-GPUTEST4-3-10-acemd_0 using acemd version 648 01/11/2008 17:03:00|SETI@home|Restarting task 04se08ab.30108.12751.14.8.98_0 using setiathome_enhanced version 603 01/11/2008 17:03:00|SETI@home|Restarting task 04se08ad.29656.206627.7.8.147_0 using setiathome_enhanced version 603 01/11/2008 17:03:40||[error] Proposed work request 11055335.522265 bigger than max 349059.490560 The OS reports 1.29GB free, just as BOINC is reporting in the message log. The BOINC files on this partition are occupying 3.68GB according to the OS, which agrees with the amount as reported in disc space tab on the client. Clearly, however, 3.68GB is larger than the 1.29GB that is supposedly the limit according the message log. Incidentally, 1.29GB is also the amount reported in the disc space tab as free space that is unavailable to BOINC. The max value of 349059s is rather close to 4 days (345600s), as I speculated in my previous message. The requested amount of work of 11055335s is roughly 128 days (11059200s). Sorry I got this wrong before. |
|
Krunchin-Keith [USA] Send message Joined: 17 May 07 Posts: 512 Credit: 111,288,061 RAC: 0 Level Scientific publications
|
OK, two things. #1 the [error] bug has been found, the developer hopes. Unfortunately you have to wait for next client since it is over requesting work. Best I can say is downgrade or try a smaller additional days value like 0.39 to 0.79. Thanks for your help on answering questions about this. #2 I'll pass along the disk limit and see what happens, maybe when the numbers get small enough something is transposed. I discovered the use at most is reduced by 7% (don't know why) for the limit. Since you have a 5GB partition and want to use all of it, I would raise this number, it won't hurt as the other limit is 100% of space and you want to use the whole partition. See if that helps or changes anything. [edit] now that I think on this more, I think boinc uses the correct value, the message just reports the wrong value. Otherwise your current usage of 3.68 would be over the limit and you would not be getting more work. |
|
Send message Joined: 17 Apr 08 Posts: 113 Credit: 1,656,514,857 RAC: 0 Level  Scientific publications
|
OK - so as we all understand - is 6.3.21 any better than any other client .... |
|
Send message Joined: 16 Aug 08 Posts: 87 Credit: 1,248,879,715 RAC: 0 Level Scientific publications
|
You would get something like this if last time you ran the project they had different length work units. Probably what has happened is the DCF is off. Look at the properties tab and see what the DCF (Duration Correction Factor) is for each project. Ideal is 1. A new button! Cool, I never saw the "properties" tab before. It is not on my 5.10.xx clients. And you are correct. My DCF for ABC is 5.2, and my client got about ... 5.2 times as many work units as it should. Thanks for teaching me a new trick. |
|
Krunchin-Keith [USA] Send message Joined: 17 May 07 Posts: 512 Credit: 111,288,061 RAC: 0 Level Scientific publications
|
OK - so as we all understand - is 6.3.21 any better than any other client .... Yes, It runs proper tasks. I've had zero failure since install to run correct amount of cpu and cuda tasks. It does have 2 minor bugs we have found. See rest of conversation. Which are in those clients also, they just may not show up unless certain circumstances exist. |
|
Krunchin-Keith [USA] Send message Joined: 17 May 07 Posts: 512 Credit: 111,288,061 RAC: 0 Level Scientific publications
|
.... This has been fixed. There were some places in the client where the disk amounts are being multiplied by 1e9 which is 1,000,000,000 and not the correct gigabyte amount of 2^20 which is 1,073,741,824. |
|
Krunchin-Keith [USA] Send message Joined: 17 May 07 Posts: 512 Credit: 111,288,061 RAC: 0 Level Scientific publications
|
You would get something like this if last time you ran the project they had different length work units. Probably what has happened is the DCF is off. Look at the properties tab and see what the DCF (Duration Correction Factor) is for each project. Ideal is 1. The properties button is new in the last few releases. Available on both the projects and tasks tab. Yes, that would do it. It will self correct over time or your can fix yourself if you know how to edit client_state.xml note: This has nothing to do with this specific client. It just happened at exact time the client was updated that projects were switched to one with a out of whack DCF |
|
Send message Joined: 28 Sep 08 Posts: 27 Credit: 6,201,632,872 RAC: 0 Level Scientific publications
|
Brilliant! Thanks for passing along the issues I've found, and I'm glad I'm contributing to the testing, however small a part it is :) |
|
Send message Joined: 18 Sep 08 Posts: 368 Credit: 4,174,624,885 RAC: 0 Level  Scientific publications
|
OK - so as we all understand - is 6.3.21 any better than any other client .... I agree, it does seem to run the Proper amount of Task's, but I did find 1 Box this morning running only 2 Regular Wu's & 1 GPU Wu on a Quad Core. I have 3 ncpu's set in the cc_config.xml file & until this morning all 5 of my GPU Capable Box's ran 3 Regular Wu's & 1 GPU Wu. I stop & re-started BOINC several times but it would go right back to 3 Reg & 1 GPU so I re-booted and then it stayed @ 3 Reg & 1 GPU Wu & so far has stayed there ... Running Win XP Pro 64-Bit on that Box |
Bender10 Send message Joined: 3 Dec 07 Posts: 167 Credit: 8,368,897 RAC: 0 Level  Scientific publications
|
PoorBoy, Did you try just using the 'Advanced, Preferences' menu , and setting % cpu used? I am also running a Win64 box, and this (6.3.21) seems to be working well. It is only an X2 box, but it's been running 1+1 fine (no re-starts) since I upgraded the Boinc.  Consciousness: That annoying time between naps...... Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it. |
|
Send message Joined: 18 Sep 08 Posts: 368 Credit: 4,174,624,885 RAC: 0 Level Scientific publications
|
Set it to what, it's set to 100% now, what else would I wnat it to be other than 100% usage ??? |
|
Bender10 Send message Joined: 3 Dec 07 Posts: 167 Credit: 8,368,897 RAC: 0 Level Scientific publications
|
If you want your GPU to have the most CPU cycles, set your Quad = 75%, your box should then run 3 cpu + 1 gpu tasks. a dual would be = 50% (1+1). If you look in Task Manager while running 4+1, you should see about 16% usage (Win64) on the GPU task. When you change to 75% in 'Preferences', that should change the GPU usage upwards (to maybe 35%). This may not be efficent in CPU utilization, but your GPU task will run faster. My Ubuntu (x64) box does not seem to have a CPU sharing issue. It only uses 1-2% cpu all the time (4+1 or 3+1). I'm not sure why this is. So it runs at 'Preferences = 100%' all the time. Maybe I have to run a newer driver.... And remove or edit out the 'nocpus' line in 'cc_config', or disable it or whatever you do to turn it off. Consciousness: That annoying time between naps...... Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it. |
|
Send message Joined: 18 Sep 08 Posts: 368 Credit: 4,174,624,885 RAC: 0 Level Scientific publications
|
Okay, I see now & will play with it a little & see what happens, Thanks Bender ... :) |
[SETI.USA]Tank_Master Send message Joined: 8 Jul 07 Posts: 85 Credit: 67,463,387 RAC: 0 Level  Scientific publications
|
I am now not getting work from 3x+1 I keep getting "not getting work, wont finish in time You comp is on 98.4% of the time and BOINC gets 97.8% of that." how could I not be able to finish a 2.5h wu in a month on a qx6700 running only 3x+1 and ps3grid? cache is set to connect every 0 days, buffer for 1 day boinc 6.3.21 x64 on server 2008 x64 |
 Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications |
Do you have the GPU-Grid ressource share massively higher than 3+1? MrS Scanning for our furry friends since Jan 2002 |
©2025 Universitat Pompeu Fabra