Message boards :
Number crunching :
Nvidia driver 296.10 problem?
Message board moderation
Author | Message |
---|---|
Send message Joined: 9 Apr 08 Posts: 2 Credit: 17,234,986 RAC: 0 |
After installing the above driver I am getting computation errors for M@H WUs and Collatz Conjecture WUs on my nvidia GTI 560 Ti graphics card. In both cases the error appears after 1-2 seconds of computing. M@H is using opengl_nvidia and Collatz is Cuda23. I also run Einstein (Cuda32), GPUGrid (Cuda31) and Seti (Fermi) without any problems. Is the problem the nvidia driver or the W@H and Collatz applications? My current "solution" is to revert to the 285.62 driver, where all 5 applications compute without errors. Windows 7 Pro, Intel Core 2 Quad Q6600 processor |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
My guess would be the CUDA sleep bug that hit the 295.xx and 296.xx driver series. What happens is when the monitor goes to sleep it also puts the card to sleep and that does not allow CUDA processing. |
Send message Joined: 9 Apr 08 Posts: 2 Credit: 17,234,986 RAC: 0 |
I somehow don't see this a "sleep" problem as it only happens with two of the five Cuda applications I run. If it was a problem caused by the card going to sleep it would affect all the apps wouldn't it? I was actually wondering if it was a problem processing Cuda23 and opencl instructions. |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
I see the app falling back onto the CPU over at SETI as well, you have a -177 error there from the sleep bug. The errors are saying that they cannot find a CUDA device. |
©2024 Astroinformatics Group