GPU wu's hang at 99%

[cw6591]

Windows 7 here as well. I've not had a problem like this before, and things seem to be ok atm. Probably doomed this WU by saying that...

[Kent Irwin]

The problem continues here. Had to restart 2 clients today to get past stuck 8900 wu's. Logs just stop reporting progress is the only clue. Since cw6591 is running W7 it seems the only common denominator is an AMD gpu. It might be helpful if someone knowledgeable from Stanford were to comment.

[TheHiding]

Hey guys. New around here. Ran into the same issue here today.
The log stopped showing any updates from that GPU. No errors or warnings. No windows (8.1) warnings.
It's a AMD 7950, I also have a 6800 series in the same PC (14.2 Drivers). They started off running neck in neck with each other. While the 6800 seemed to stall out, running much slower then the 7900, it didn't stop posting to the log.

Something I haven't seen anyone else mention. While watching the GPU usage for the 7900 series it looked more like a heartbeat monitor. Spike here, spike there prior to reaching 95% after 95% the spikes got smaller, only reaching ~40% GPU usage. At 99.99% it flatlined.
(The above it how I found things in the morning, before I went to bed, both GPUs were running constant at ~99%)

I let it sit at 99.99% for 30 minutes. To try and jump start it back to life I switched Folding Power to OFF, the completion bar dropped to 30%, going back to any other setting it would jump back up to 99.99%. After rebooting, the GPU usage is back to ~99%.

Edit.. After a bit, the completion dropped back down to, well 36% when I noticed it. Also both GPU's are back to running near the same speed. With similar ETAs

[bruce]

None of you have reported that you tried reducing the clock rates or checking the GPU temperatures yet.

Did your system enter sleep/hibernate and then reawaken? Is there a GPU_RESET event in the Windows Event Viewer? Did a foreign system offer you a connection using Remote Desktop?

I believe this is only a problem with FahCore_17. Development is aware of the problem and working toward a fix.

[cvsi3]

I was having the issue for some time. My 7970 was clocked at 1240Mhz Core / 1600Mhz Memory.
Found that if I dropped it to 1225Mhz core, it didn't have an issue. For me heat was not an issue as the card only gets to about 46c.

I didn't start having an issue until AMD released the 13.12 drivers. Currently using 14.2 beta drivers without issue at the 1225mhz clock speed.

On my system when it was happening, it was the drivers crashing. I would get the bubble in the notification bar saying that the AMD driver has stopped responding.

[monkeyclaw]

I don't know if this is the same problem that my machine was running into, but I also had my GPU work units hang. It always happened at night though, so I figured out that the monitor going to sleep is what was causing it. After I set the monitor to never go into idle/sleep mode, it stopped hanging at 0% (which would then make the progress bar falsely read 99.99% when I got up and checked folding, but no work had been done).

If it was the same issue though, it should have been happening since the beginning with the R9 270 hardware (I only know it does it for R9 280x for sure), so it is strange that it started happening after a period time, so it might night be the same issue at all. Just thought I'd mention it if the problem has not yet been resolved/to see if it helps.