Advanced search

Message boards : Number crunching : Persistent Gerard Failures on a Modest Rig

Author Message
tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41477 - Posted: 4 Jul 2015 | 16:00:25 UTC

My daughter runs GPUGrid longs on her Dell 9100: Win7, 4G RAM, 2-core processor and ASUS GTX 660.

Noelias run fine, but more than half the Gerards she gets fail, here.

Any thoughts?

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2048
Credit: 14,826,285,069
RAC: 2,412,335
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41479 - Posted: 4 Jul 2015 | 20:06:34 UTC - in response to Message 41477.
Last modified: 4 Jul 2015 | 20:12:20 UTC

The GPU temperature looks fine.
Try to reduce the GPU clock by 50-100 MHz, if it doesn't help then try to reduce the GPU's memory clock by 100-200 MHz.
GERARDs push the GPU a bit harder than NOELIAs.
Even the successful GERARD workunits are full of "simulation became unstable" messages on this host.
It could be a power supply problem also, check all power connectors for burn marks.

tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41482 - Posted: 5 Jul 2015 | 5:40:01 UTC - in response to Message 41479.

The GPU temperature looks fine.
Try to reduce the GPU clock by 50-100 MHz, if it doesn't help then try to reduce the GPU's memory clock by 100-200 MHz.
GERARDs push the GPU a bit harder than NOELIAs.
Even the successful GERARD workunits are full of "simulation became unstable" messages on this host.
It could be a power supply problem also, check all power connectors for burn marks.

Thanks, Retvari. I'm on the job!

Jim1348
Send message
Joined: 28 Jul 12
Posts: 695
Credit: 1,371,992,468
RAC: 3
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 41483 - Posted: 5 Jul 2015 | 6:22:49 UTC
Last modified: 5 Jul 2015 | 6:30:50 UTC

I have had to reduce at least the GPU clock on every GTX 660 that I have to make it stable on the hardest work units. In one case, even going down to the chip default frequency (980 MHz as I recall) was not enough, I had to reduce it to 967 MHz, and even reduce the memory clock as well as RZ mentions. Apparently Nvidia specified the operating limits of that particular chip right at the limits of what it would reliably do, though some cards are better than others due to heat-sinking or maybe power regulator capabilities. If your Dell is the laptop, it would have even less of a margin. Good luck.

Post to thread

Message boards : Number crunching : Persistent Gerard Failures on a Modest Rig