Message boards : BOINC Manager : Limiting GPU load
Message board moderation
Author | Message |
---|---|
Send message Joined: 4 May 10 Posts: 2 |
Hello Is there any way to limit GPU load like you could limit CPU load? i would like to allow my GPU to calculate but if it is always 99%, my comp starts making too much noice |
Send message Joined: 29 Aug 05 Posts: 15560 |
In BOINC, limiting the CPU load will also limit the GPU load (since the CPU is in use by the actual program that feeds the GPU). But if you're looking for an external program that works with BOINC, check out TThrottle. |
Send message Joined: 4 May 10 Posts: 2 |
Hey Do you mean if i set CPU load to 50% then GPU load limit should also be 50%? Tryed it, but GPU keeps still load @ 99% Throttle makes the whole load weird... before the load was stable, after messing with throttle GPU ja CPU load keeps jumping... Also is there any way to suspend work if the computer is used for gaming? Its ok if the project is running if im using just internet. |
Send message Joined: 29 Aug 05 Posts: 15560 |
Hey The throttle function in BOINC will suspend and resume the science application depending on what value you used. If 50%, it'll suspend for a second, resume for a second, etc. Since the GPU applications do not run on the GPU, but on the CPU (you can see it in Windows Task Manager, so it runs on the CPU), any throttling effect you give to the CPU will also slow down the throughput of data to the GPU. The GPU will however run the data at full whack. It can't do it any slower, it's either on or off. Why no GPU throttle then? It's not available in the GPU manufacturers' drivers releases, so the BOINC developers can't develop something that's platform independent and works across all of them (Windows, Linux, Mac). Also is there any way to suspend work if the computer is used for gaming? Yes, see my CUDA & CAL FAQ, the "Ways around it?" section. You can use the <exclusive_app> tags (suspends the whole of BOINC) and <exclusive_gpu_app> tags (Suspends GPU crunching only, CPU crunching continues). |
Copyright © 2024 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.