Message boards :
Number crunching :
Just some early observations on V7 CUDA work.
Message board moderation
Previous · 1 · 2 · 3 · 4
Author | Message |
---|---|
Wiggo Send message Joined: 24 Jan 00 Posts: 34844 Credit: 261,360,520 RAC: 489 |
In fact it was because the 660's had switched over to backup projects for the 2nd time that I could be doubly be sure of this. A shame that I don't have my old 32bit Vista system otherwise that would certainly verify if it's just a 32bit thing or just something to do with my Q6600 rig but just running anything else doesn't produce this effect, only V7 CUDA work (plus I've been looking into this for the last few of weeks). BTW under V6 CUDA work this behaviour never happened and it is definitely cause by V7 CUDA work (none here now and no memory problems any more). Cheers. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
BTW under V6 CUDA work this behaviour never happened and it is definitely cause by V7 CUDA work (none here now and no memory problems any more). Hmmm, I'm not aware of extra potential for a memory leak in V7 (I re-use the same memory areas), nor any particular Cuda runtime bug that would result in such behaviour. Close inspection with Process explorer might be warranted to ensure the source. I suggest updating drivers with clean install advanced option, just to ensure nothing's gone awry either with the driver version itself having some sortof bug, or some damage. "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. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34844 Credit: 261,360,520 RAC: 489 |
BTW under V6 CUDA work this behaviour never happened and it is definitely cause by V7 CUDA work (none here now and no memory problems any more). I have tried re-installing the driver twice (I won't go higher presently as there could be more issues here and not just for BOINC/SETI) and everything else I can think of over the last 2wks and 1 thing that I havn't tried is just running 1 w/u at a time but that would really be counter productive here. But as I said, this shouldn't be a problem for me soon enough as this rig is getting a heart/lung transplant. But it's definitely the V7 CUDA though and running the same app under V6 did not do this. 1 thing I will try, when I get some more CUDA work, is to free up a 2nd CPU core and see if that might make a difference. Cheers. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34844 Credit: 261,360,520 RAC: 489 |
Well with no V7 CUDA work overnight the AP's running on the CPU instead of taking 18-20hrs have been done in 10hrs (with V6 CUDA tasks running AP's on CPU took around 12hrs) so I'll definitely free up a 2nd core and see what happens. Cheers. |
©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.