Issue getting multiple GPUs working

Evil-I · 9 Apr 2017 at 16:35

Hi ALL,

Been doing some distributed computing on Einstein, Rosetta and Seti all for Team OCUK but had taken a break for a while.

Just fired up the main rig aiming to run and only one of my three gtx970 gpu's is being used? I've tried both with and without Sli and the log seems to suggest boinc is seeing all three.

After a bit of googling i found a post about puuting a cc_setup.xml in the Boinc program data folder telling it o use all GPUs, but that doesn't seem to have fixed it either.

Any suggestions?

Thanks,

E-I

Evil-I · 17 Apr 2017 at 11:31

Thanks all, sorry for the slow reply, easter and family stuff been getting in the way.

Yes, have now got all three gpu's working, just have some instability problems on this rig.

I'm currently running two rigs for Boinc projects, my test bench which has a 10 core xeon (ES sample similar to E5-2690V2 but 2.1Ghz stock and up to 2.9Ghz turbo) and a gtx960 which is going great guns, but my main rig (as per sig) restarts when doing gpu compute after about an hour.

I've downclocked everything to stock and it still does it, reset projects, updated all drivers but still not stable for gpu compute tasks (seems rock solid in everything else though). There's some stuff going on where I have an old cuda dev sdk which won't uninstall and some other oddities. I did do a complete Nvidia clean and re-install but still can't get rid of the Cuda stuff.

Long story short, think I need to do a system re-install but know its going to take the best part of a day :-(

Any suggestions as to how I can fault find this GPU Compute instability issue? Currently running Seti, Einstein, Grid and Rosetta.

Thanks,

E-I

Evil-I · 17 Apr 2017 at 18:48

Good thought Andy!

Gone a slightly different route in that I make system backups with Reflect every now and again anyway, so just did a new one to my server and wiped to start again. Got it pretty much back up together and running boinc and it already seems better. Running for about an hour now, so let's see.

Evil-I · 17 Apr 2017 at 18:49

I have left my GPUs at stock for now. They all run modded firmware anyway so boost higher than normal just not to the level it can go under water, if it's stable I'll put the overclock back on and see if it's still stable.