GTX 680 Folding

[sswilson]

Bah.... pulled a bloody 7640. Guess I turn off GPU folding on my 680 for a couple of days. :(

[Mathus]

I'm back after a bit of a break from the forums. Has there been any updates to the V7 client to increase folding performance on AMD 7900 series cards? Given that the compute power is so much higher than the 680's I would expect better results.

[lowfat]

Originally Posted by Mathus

I'm back after a bit of a break from the forums. Has there been any updates to the V7 client to increase folding performance on AMD 7900 series cards? Given that the compute power is so much higher than the 680's I would expect better results.

I wouldn't expect more folding@home performance from AMD cards any time soon. As in years.

[sswilson]

Bah...... thought my PPD was low... turns out the "new" 8008 WUs don't run on the 680... same story as previous failed WUs in that it pegs power @ 30% and GPU usage @ 50%, but never logs any progress. I attempted to pull a different WU by deleting the slot & all files in the work folder, but it just pulled the exact same WU when I recreated the slot. Guess I'll leave the GPU idle for a few days and then check back later in the week to see if it pulls a WU that'll function on my card. :(

[draemn]

Well, here goes nothing... trying to get my 680 folding after a HUGE break from folding. Will let you know what happens once I find out.

[sswilson]

Originally Posted by draemn

Well, here goes nothing... trying to get my 680 folding after a HUGE break from folding. Will let you know what happens once I find out.

Let us know for sure. I've stopped GPU folding on my main rig until such time as they release a proper non-beta client for kepler.

[draemn]

I don't get this, is there a NDA on instructions to get F@H working on a 680? I can't find any of my old posts describing (with images) how to set up the flags for the slot to make it work... and of course I can't get it to work. And google gives no useful posts... wtf.

Update: saw a few posts hinting that people started get get kepler working v.6 clients, but I don't have the patience to try that. As for now:

GPU device id=0
20:38:58:WU00:FS00:0x15:Working on GRowing Old MAkes el Chrono Sweat
20:38:58:WU00:FS00:0x15:Client config unavailable.
20:38:58:WU00:FS00:0x15:Starting GUI Server
20:40:00:WU00:FS00:0x15:Setting checkpoint frequency: 250000
20:40:00:WU00:FS00:0x15:Completed 3 out of 25000000 steps (0%).

that is what I've got in my log file and MSI afterburner shows 99% GPU usage on a 8018 WU (had to remove the first WU it downloaded as that was a bust).

Updated 2: 2% complete!!! I will update more tomorrow to see if it keeps getting WUs that work.. It appears that after you add the GPU.txt (and I added the "client-type : beta" flag to be safe), you need to restart but also delete any WU that was downloaded. Will be interesting to see if babysitting is required to remove bade WUs... tomorrow I will play around and see what happens without setting the slot to beta status as it is recognized (might just get unusable WUs).

[sswilson]

It's not supported outside of the Beta team and they're getting a little tired of saying that. :)

You need the client type beta switch made

[draemn]

First 8018 was great and my second unit is another 8018! :D 18-20k PPD if I didn't distract the GPU with anything else. Can't wait to see what it's like tomorrow.

[draemn]

20k PPD on the 8018 WUs. So far I only have 8018 and they behave extremely well :)