Message boards : Questions and problems : What correct way to accomplish such config?
Message board moderation
Author | Message |
---|---|
Send message Joined: 9 Apr 06 Posts: 302 |
Again, lack of project pairing ability.... I have 1 CPU based app, 1 almost totally GPU-based app and 1 that uses GPU only in part of computations and use CPU for other parts. The problem is: how to keep both CPU and GPU busy in such situation? ( cases when some project have no work should be considered too) |
![]() Send message Joined: 14 Mar 09 Posts: 215 ![]() |
what version are you running of boinc, os, graphics card? |
Send message Joined: 9 Apr 06 Posts: 302 |
OS? BOINC scheduling OS-dependend? Well, Vista, 6.10 branch (need ATI GPU support), ATI GPU. |
![]() Send message Joined: 14 Mar 09 Posts: 215 ![]() |
what is your cache set on? it may help to run a few more cpu projects. |
Send message Joined: 9 Apr 06 Posts: 302 |
currently 0.1+0.28 days (to get total cache less than 3 days and keep MW from going into high priority mode). + 1" backup" CPU-only project possible, with much low share than main CPU-based project. |
Send message Joined: 9 Apr 06 Posts: 302 |
No ideas? What I need is to limit number of app instances that BOINC will launch while keeping ncpus and coproc values low enough to allow BOINC to run it together with other consumers of those hardware resources. What I see now - BOINC will preempt let say GPU tasks and launch many instances of my app. But there is another resource that will limit useful instances number - GPU memory amount. How can I account for this in app_info file ? |
Copyright © 2025 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.