Message boards :
Number crunching :
Is there a way to limit the number of cpus...
Message board moderation
Author | Message |
---|---|
Send message Joined: 13 May 11 Posts: 6 Credit: 7,002 RAC: 0 |
crunching milkyway wus? I have an AMD Phenom II X4 and milkyway uses all of the cores. But there are other projects running in the list, so I want them be running too. Looked at the configs in my account but there is nothing like this. Thanks for the tips. |
Send message Joined: 8 May 10 Posts: 576 Credit: 15,979,383 RAC: 0 |
crunching milkyway wus? I have an AMD Phenom II X4 and milkyway uses all of the cores. But there are other projects running in the list, so I want them be running too. Looked at the configs in my account but there is nothing like this. Thanks for the tips.See here: http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2403&nowrap=true#48735 However you can use app_info.xml to set the number of threads manually, by adding something like this: <avg_ncpus>2</avg_ncpus> <max_ncpus>2</max_ncpus> <cmdline>--nthreads 2</cmdline> |
Send message Joined: 16 Dec 10 Posts: 46 Credit: 205,697,511 RAC: 0 |
I think Levent means the project weight. You can set the project weight in each projects preferences. Default is 100. If you run 3 projects on your quadcore... set MW to 200 and the other two leave 100. Now MW uses 2 cores and the other two projects each one core.... If you change things like this... can take some time before it works correctly... give Boinc manager some time. |
Send message Joined: 1 Sep 08 Posts: 204 Credit: 219,354,537 RAC: 0 |
No, that will not give him what he's asking for. MW n-body would still take all 4 cores, if it runs. However, the Boinc manager should average project times out according to ressource share - so this [ressource share] is indeed what he should set. MrS Scanning for our furry friends since Jan 2002 |
Send message Joined: 27 Aug 08 Posts: 1 Credit: 100,380 RAC: 0 |
I think my issue is the essentially the same, so I will post it here. Moderator - please move if I posted in the wrong place. I have an i7 processor. When I receive multi-threaded tasks from milkyway, instead of my system all 8 cores at 100%, the processor utilization bounces around between 30-60% for all cores. If I suspend all other processing and let the milkyway task have the entire processor, the task completes in about 8 minutes. If I just let it run, without suspending the other tasks, the milkyway task takes hours to complete; and virtually nothing else is processed. Likely more time is spent swapping out tasks than actually crunching numbers... It would be nice if I could set something to let the milkyway task take over the system when the application switch occurs. Then it could run to completion, after which the other tasks in the queue would run normally. Are there any project setting that control how multi-threaded applications interact with single-threaded apps that would help me? I was told the BOINC client does not have anything in this regard... TIA for any assistance, Tom |
Send message Joined: 29 Sep 10 Posts: 54 Credit: 1,343,989 RAC: 14 |
Elliotts, Just go with the same solution that Matt posted. Limit MW tasks to at most 4 processors on your i7 (odd numbers are also valid). I've set BOINC & MW to only use 4 CPUs, and MW tasks finish in 8 minutes. |
Send message Joined: 13 May 11 Posts: 6 Credit: 7,002 RAC: 0 |
Thanks for the advices and information. Have a nice day. |
©2024 Astroinformatics Group