| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Windows 7_64 Uitimate and BOINC |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 58
|
|
| Author |
|
|
Richard Mitnick
Veteran Cruncher USA Joined: Feb 28, 2007 Post Count: 583 Status: Offline Project Badges: 
|
4:30AM 11/24/09
----------------------------------------Here are messages from the Einstein report onto now, out of work: 11/23/2009 8:27:50 PM Einstein@Home Reporting 3 completed tasks, not requesting new tasks 11/23/2009 8:27:56 PM Einstein@Home Scheduler request completed 11/23/2009 10:46:38 PM Leiden Classical Computation for task wu_164284800_1257325319_40733_1 finished 11/23/2009 10:46:40 PM Leiden Classical Started upload of wu_164284800_1257325319_40733_1_0 11/23/2009 10:46:40 PM Leiden Classical Started upload of wu_164284800_1257325319_40733_1_1 11/23/2009 10:46:41 PM Leiden Classical Finished upload of wu_164284800_1257325319_40733_1_1 11/23/2009 10:46:43 PM Leiden Classical Finished upload of wu_164284800_1257325319_40733_1_0 11/23/2009 10:47:40 PM Leiden Classical Computation for task wu_164284800_1257325319_40743_1 finished 11/23/2009 10:47:42 PM Leiden Classical Started upload of wu_164284800_1257325319_40743_1_0 11/23/2009 10:47:42 PM Leiden Classical Started upload of wu_164284800_1257325319_40743_1_1 11/23/2009 10:47:44 PM Leiden Classical Finished upload of wu_164284800_1257325319_40743_1_0 11/23/2009 10:47:44 PM Leiden Classical Finished upload of wu_164284800_1257325319_40743_1_1 11/23/2009 10:48:12 PM Leiden Classical Computation for task wu_164284800_1257325319_40729_0 finished 11/23/2009 10:48:14 PM Leiden Classical Started upload of wu_164284800_1257325319_40729_0_0 11/23/2009 10:48:14 PM Leiden Classical Started upload of wu_164284800_1257325319_40729_0_1 11/23/2009 10:48:16 PM Leiden Classical Finished upload of wu_164284800_1257325319_40729_0_1 11/23/2009 10:48:17 PM Leiden Classical Finished upload of wu_164284800_1257325319_40729_0_0 11/23/2009 10:49:42 PM Leiden Classical Computation for task wu_164284800_1257325319_40739_1 finished 11/23/2009 10:49:44 PM Leiden Classical Started upload of wu_164284800_1257325319_40739_1_0 11/23/2009 10:49:44 PM Leiden Classical Started upload of wu_164284800_1257325319_40739_1_1 11/23/2009 10:49:46 PM Leiden Classical Finished upload of wu_164284800_1257325319_40739_1_1 11/23/2009 10:49:48 PM Leiden Classical Finished upload of wu_164284800_1257325319_40739_1_0 11/23/2009 10:52:05 PM Leiden Classical Computation for task wu_164284800_1257325319_40737_1 finished 11/23/2009 10:52:07 PM Leiden Classical Started upload of wu_164284800_1257325319_40737_1_0 11/23/2009 10:52:07 PM Leiden Classical Started upload of wu_164284800_1257325319_40737_1_1 11/23/2009 10:52:08 PM Leiden Classical Finished upload of wu_164284800_1257325319_40737_1_1 11/23/2009 10:52:08 PM Leiden Classical Computation for task wu_164284800_1257325319_40731_0 finished 11/23/2009 10:52:09 PM Leiden Classical Finished upload of wu_164284800_1257325319_40737_1_0 11/23/2009 10:52:10 PM Leiden Classical Started upload of wu_164284800_1257325319_40731_0_0 11/23/2009 10:52:10 PM Leiden Classical Started upload of wu_164284800_1257325319_40731_0_1 11/23/2009 10:52:11 PM Leiden Classical Finished upload of wu_164284800_1257325319_40731_0_1 11/23/2009 10:52:12 PM Leiden Classical Finished upload of wu_164284800_1257325319_40731_0_0 11/23/2009 10:52:45 PM Leiden Classical Computation for task wu_164284800_1257325319_40741_1 finished 11/23/2009 10:52:47 PM Leiden Classical Started upload of wu_164284800_1257325319_40741_1_0 11/23/2009 10:52:47 PM Leiden Classical Started upload of wu_164284800_1257325319_40741_1_1 11/23/2009 10:52:49 PM Leiden Classical Finished upload of wu_164284800_1257325319_40741_1_1 11/23/2009 10:52:50 PM Leiden Classical Finished upload of wu_164284800_1257325319_40741_1_0 11/23/2009 10:53:38 PM Leiden Classical Computation for task wu_164284800_1257325319_40735_1 finished 11/23/2009 10:53:40 PM Leiden Classical Started upload of wu_164284800_1257325319_40735_1_0 11/23/2009 10:53:40 PM Leiden Classical Started upload of wu_164284800_1257325319_40735_1_1 11/23/2009 10:53:41 PM Leiden Classical Finished upload of wu_164284800_1257325319_40735_1_1 11/23/2009 10:53:42 PM Leiden Classical Finished upload of wu_164284800_1257325319_40735_1_0 |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
mitrichr,
----------------------------------------re the line 11/22/2009 9:45:27 AM NVIDIA GPU 0: GeForce GT 220 (driver version 18627, CUDA version 2020, compute capability 1.2, 1024MB, 128 GFLOPS peak) if you're not doing GPU crunching presently, switch that off with the appropriate cc_config.xml line <no_gpus>1</no_gpus> in the <options> section. Whilst GPU and CPU have separate scheduling in 6.10, there still appears to be some cross impacting in assigning the CPU's. Also, when you look next time in the project properties annotate the DCF's and look in the client_state.xml what the <time_stats> are. With hyperthreading I'm expecting the DCF to be quite a bit above 1.0. Just curious how that works out as.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Richard Mitnick
Veteran Cruncher USA Joined: Feb 28, 2007 Post Count: 583 Status: Offline Project Badges:
|
First, nothing is going to happen now for a week, as I am away. I can check the GPU status when I get home.
---------------------------------------->>RSM ---------------------------------------- [Edit 1 times, last edit by mitrichr at Nov 25, 2009 4:51:23 AM] |
||
|
|
Richard Mitnick
Veteran Cruncher USA Joined: Feb 28, 2007 Post Count: 583 Status: Offline Project Badges:
|
Sek, I found in
----------------------------------------http://boincfaq.mundayweb.com/index.php?language=1&view=471 about CUDA and GPU the instruction how to turn GPU off in the projects preferences. I found the spot in one project where one can check to use or un-check to turn off. If I go to that line in each project and un-check the box, will that do the trick? It certainly would be nice to be able to use this feature, what with having the right chip. Do, you think that this problem will be addressed in a later build of the BOINC agent? |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
I don't know. It makes zero sense why with 6.10.18 you're having idling cores. Is your Activity menu set to run always and network always on? Try 6.10.19.
----------------------------------------Did you look for the DCF's? I still speculate that jobs assigned to hyperthreaded cores cause hyperinflation of the DCF value and still 6.10.18 should fetch work for idle cores... that is, if back-off counters are not already running for a project?????? I emulated HT unwillingly on my duo by severe throttling and WCG was listed as having a DCF of 6.5. Since, even with a cache set to 1.0 it's refusing to fetch more than the current running jobs and only calls last minute. As for disabling GPU, if there are projects that allow you to chose CPU or GPU projects and work calls, great, but the cc_config.xml line will kill it all (see that FAQ's "can I turn the darn thing off" section. Also, I understand that for instance WCG now tells GPU work callers automatically to back off for 7 days before asking again even though it has as yet not got an option on the device profiles.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Richard Mitnick
Veteran Cruncher USA Joined: Feb 28, 2007 Post Count: 583 Status: Offline Project Badges:
|
"...Did you look for the DCF's? I still speculate that jobs assigned to hyperthreaded cores cause hyperinflation of the DCF value and still 6.10.18 should fetch work for idle cores...."
----------------------------------------I will be looking at DCF when I get home Sunday. So, then, if I report the values, will there be some direction from them as to what to do next, e.g., it sounds like high numbers are not good, so how to reduce them, etc. "...if there are projects that allow you to chose CPU or GPU projects and work calls, great, but the cc_config.xml line will kill it all (see that FAQ's "can I turn the darn thing off" section..." So, what does "kill it all" mean, and, do you imply that turning off GPU in each projects preferences is not sufficient? I have saved that BOIN FAQ page regarding CUDA and GPU as a bookmark so that I can get to it quickly. Lovely weather in California. >>RSM |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Let's talk further when you're back at the machine and got your bearings sorted behind the square tube looking into your computer.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
Did you look for the DCF's? I still speculate that jobs assigned to hyperthreaded cores cause hyperinflation of the DCF value The DCF in WCG isn't a good indicator, since it can be heavily influenced by whatever sub-project you're running, and if you're example running children or parents and so on... But anyway, in WCG: computer A has... DCF 1.5176 after a string of MW. computer B has... DCF 1.4030 after a string of MX. computer C has... DCF 0.9146 after a string of MW. In SIMAP: A has 0.3324 B has 0.4874 C has 0.3559 In SETI@home: A has 0.3668 B has 0.1486 C has 0.1409 Hmm, actually, it didn't become much clearer by looking on a couple other projects either... So, can you spot the i7-HT in this list over DCF? It's B.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
If you want to play silly games and pose silly question, you can get silly answers too... let people answer the inquiries posed to them instead and NOT to you.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges:
|
I will be looking at DCF when I get home Sunday. So, then, if I report the values, will there be some direction from them as to what to do next, e.g., it sounds like high numbers are not good, so how to reduce them, etc. Basically, the DCF is mainly an indication of how "wrong" the project-estimates are... but can also in some instances show computers having problems. But, you'll need to compare with other computers in the same project to be sure there's any problems or not. Example, in WCG, any DCF from 0.8 - 1.6 is completely normal, and doesn't signify any problems. Also, it's not really anything you should do to reduce them. In other projects the "normal" numbers is differrent, and in instances a projects gets the estimates completely wrong, having a DCF of 100+ can happen, and it still doesn't indicate a problem with users setup. "...if there are projects that allow you to chose CPU or GPU projects and work calls, great, but the cc_config.xml line will kill it all (see that FAQ's "can I turn the darn thing off" section..." So, what does "kill it all" mean, and, do you imply that turning off GPU in each projects preferences is not sufficient? If you're not attached to any GPU-capable projects, or, doesn't want to use the GPU in any of the attached projects, it's better to completely disable it, since it will often give unneccessary extra connections, and is more likely to give unexpected problems than if you're only running CPU. To completely disable the GPU, the "best" method when you're running Vista or Win7 is to just install BOINC as a service. But of course, if you wants to use the GPU, and it's fast enough, you can't disable it.  BTW, for when you returns, there's atleast one problem with GPU that should have been fixed in v6.10.19, so is one of the reasons for upgrade. How this somehow has influenced cpu-scheduling is on the other hand a mystery... There's also a fix with multi-threaded applications, but since you're not running Aqua you shouldn't have this problem. In v6.10.18 you can be stuck with a single core running, while the Aqua-task waits on it's turn... But, while multi-threaded application-scheduling is improved, there's still problems with it in v6.10.19, so this part needs further changes. Also, interestingly enough, the AFAIK only other reported problem with someone running-out of work is not with the cpu-part, but the CUDA-part, and is on an Amd-computer running Linux... "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
|