Now receiving older GPU2 units that hang my 670's

[Carbon_Rod]

So far, each of my 670's has received a 10504 GPU2 work unit that uses core 11. Apparently, neither of my Keplers like them because even though it appears to want to start folding, it never does and just hangs there. The troubling part is that I have to physically pause the folding, wait a minute or two for the message to appear that the core is shut down, and then restart it. But even after restarting, it looks like it's going to attempt to fold but it then errors out immediately with UNSTABLE_MACHINE and moves on to a different unit. I'd post logs, except that I don't have access to my folding boxen right now.

Anyone else getting these?

In order to try to stop getting them, I've set the CUDA index to 1 on the GPU slot... I dunno if that does anything, but it was the only setting I could see off hand that might do something. Anyone have any other ideas?

 

[Xavier Zepherious]

if you are running mixed cards you have to manual config of the gpu in v7
otherwise it gets mixed up with another card and you get the wrong core

if you have all 670's then it could be a server issue cropping up again
(which is why they took the servers down for maintenance)

 

[Carbon_Rod]

I can't say it's due to mixing of cards. Currently, each system that has received one of these units only has a single 670 installed in it. I still set the CUDA index to something other than the default of "-1" just to see if that will help. Other than that, all other slot options are at default.

I'm going to wait and see if this is a recurring problem, but I just found info on the "gpu-species" slot option which may just help fix things if it is.

 

[Nathan_P]

Its a server issue - a couple of threads have popped up over at FF