Message boards :
Number crunching :
gpu stop work if cpu wu´s at high priority
Message board moderation
Author | Message |
---|---|
tpl Send message Joined: 12 Nov 03 Posts: 461 Credit: 243,368,408 RAC: 14 |
hi, at my i7 with 2 nvidia 570 running 8 wu´s cpu and 4 wu´s at gpu....but when all cpu wu´s running at high priority the ´gpu stops crunching. don´t know why :-( |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
hi, The technical term is 'CPU overcommit' and is complicated. Your GPU worker threads that feed the GPU's need to be higher than the CPU tasks, or they will never get any CPU to feed them. If you are using some priority changing software, like Fred's priority tool, or Process Lasso, I would suggest to keep CPU processes at 'Below Normal', and boost GPU host processes to 'Above Normal' [Edit:]Another option is to keep 1 real or virtual core free to feed them Jason "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
I think there was also a bug in certain versions of Boinc that would stop GPU crunching if CPU work was in high priority. But I am not sure what versions it was. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
tpl Send message Joined: 12 Nov 03 Posts: 461 Credit: 243,368,408 RAC: 14 |
thx for quick answer, i change from 7.0.23 to 7.0.25 and at the moment all running "normal" maybe a bug at bm 7.0.23 |
Cliff Harding Send message Joined: 18 Aug 99 Posts: 1432 Credit: 110,967,840 RAC: 67 |
hi, Regardless of which version of BOINC you are running you should leave at least one CPU core available for GPU processing. Both of my machines (i7/930 & i7/950) both have 2 physical GPU devices on board and are running 7 cores for CPU processing (90%). I don't buy computers, I build them!! |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
The technical term is 'CPU overcommit' and is complicated. No, the technical term is 'bug' ;-) This one was fixed by changeset [trac]changeset:25497[/trac]: client: fix typo that prevented GPU jobs from running if CPUs were filled with EDF jobs Part of the confusion is the multiple uses of the word 'priority': although I'm perfectly happy to talk about application, process, or thread priority in context, I think in the context of BOINC Manager it more commonly means the 'Earliest Deadline First' in David's checkin note. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
The technical term is 'CPU overcommit' and is complicated. Ah, 'bugs', I read about those... "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Wembley Send message Joined: 16 Sep 09 Posts: 429 Credit: 1,844,293 RAC: 0 |
Ah, 'bugs', I read about those... I use 'Raid' |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13732 Credit: 208,696,464 RAC: 304 |
Regardless of which version of BOINC you are running you should leave at least one CPU core available for GPU processing. Both my systems use all CPU cores for crunching as well as the GPUs. No problems. Maybe running a couple of dual GPU cards would make it worth it, but for just the one GPU it's not worth leaving a CPU core unused for CPU processing. Grant Darwin NT |
©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.