EXTREMELY slow GPU slot--happens after it basically crashes

If you think it might be a driver problem, see viewforum.php?f=79

Moderators: Site Moderators, FAHC Science Team

Post Reply
GPU timpster
Posts: 65
Joined: Mon Nov 02, 2015 2:57 am

EXTREMELY slow GPU slot--happens after it basically crashes

Post by GPU timpster »

Hey, whenever my overclock goes too far the core basically goes into slow mo, I'm talking, for example, 12 minutes TPF on core 18s, and currently "1 hours 10 mins" on core 18, project 10476.

I've deleted and re added the core SEVERAL TIMES, even deleted it, and rebooted, SEVERAL TIMES, and it's still stuck in the stupid slow mode. When the thing basically crashes, my GPU clocks revert to stock, so there is no overclock currently. How do I get back to the normal performance, and also why does this happen?

Also it seems no amount of deleting / re adding the core does anything, here's another example "2 hours 26 mins" Core 18, project 9140.
GPU timpster
Posts: 65
Joined: Mon Nov 02, 2015 2:57 am

Re: EXTREMELY slow GPU slot--happens after it basically cras

Post by GPU timpster »

You'll probably need some logs so here it is!
I'm guess I'll use code to put it in a box

Code: Select all

23:04:50:WU00:FS01:0x18:*********************** Log Started 2015-11-02T23:04:49Z ***********************
23:04:50:WU00:FS01:0x18:Project: 10472 (Run 0, Clone 158, Gen 216)
23:04:50:WU00:FS01:0x18:Unit: 0x0000011c538b3dbb53beb71ceda2ca87
23:04:50:WU00:FS01:0x18:CPU: 0x00000000000000000000000000000000
23:04:50:WU00:FS01:0x18:Machine: 1
23:04:50:WU00:FS01:0x18:Reading tar file state.xml
23:04:51:WU00:FS01:0x18:Reading tar file system.xml
23:04:51:WU00:FS01:0x18:Reading tar file integrator.xml
23:04:51:WU00:FS01:0x18:Reading tar file core.xml
23:04:51:WU00:FS01:0x18:Digital signatures verified
23:04:51:WU00:FS01:0x18:Folding@home GPU core18
23:04:51:WU00:FS01:0x18:Version 0.0.4
23:05:26:Removing old file 'configs/config-20151102-224542.xml'
23:05:26:Saving configuration to config.xml
23:05:26:<config>
23:05:26:  <!-- Network -->
23:05:26:  <proxy v=':8080'/>
23:05:26:
23:05:26:  <!-- Slot Control -->
23:05:26:  <power v='full'/>
23:05:26:
23:05:26:  <!-- User Information -->
23:05:26:  <passkey v='********************************'/>
23:05:26:  <team v='111065'/>
23:05:26:  <user v='Timpster-GPU_Nvidia_660Ti'/>
23:05:26:
23:05:26:  <!-- Folding Slots -->
23:05:26:  <slot id='0' type='CPU'>
23:05:26:    <cpus v='6'/>
23:05:26:  </slot>
23:05:26:  <slot id='1' type='GPU'>
23:05:26:    <gpu-index v='0'/>
23:05:26:  </slot>
23:05:26:</config>
23:05:26:WU00:FS01:0x18:Completed 0 out of 5000000 steps (0%)
23:05:26:WU00:FS01:0x18:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
davidcoton
Posts: 1102
Joined: Wed Nov 05, 2008 3:19 pm
Location: Cambridge, UK

Re: EXTREMELY slow GPU slot--happens after it basically cras

Post by davidcoton »

There is a known issue with a "go-slow" mode (possibly associated with excessive GPU memory use) which, on Core21 at least, will eventually result in a "Bad State" error and a return to a previous checkpoint. Normally a pause and resume will return to the last checkpoint and normal speed, until it strikes again. Too many occurrences on one WU will result in a Bad WU error and premature WU end. A possible cure is to reduce the overclock, sometimes even to below stock frequencies (this is for older nVidia cards, the Maxwell cards seem to need a reduction of Memory clock, but this is not yot conclusive).

I notice that the log you have posted does not show the go-slow, since the log ends just as the WU starts.
Image
GPU timpster
Posts: 65
Joined: Mon Nov 02, 2015 2:57 am

Re: EXTREMELY slow GPU slot--happens after it basically cras

Post by GPU timpster »

I'm experiencing the same issue after completely uninstalling and checking the data box, it's 43+ minutes EDIT, TPF, core 21 project 9635. What gives?

Alright, it's normal again, finally! I hate when that happens! I always takes me about 3 hours to get it going right again, always leads to me uninstalling and reinstalling.
Last edited by Joe_H on Tue Nov 03, 2015 3:16 am, edited 1 time in total.
Reason: edited for language
toTOW
Site Moderator
Posts: 6296
Joined: Sun Dec 02, 2007 10:38 am
Location: Bordeaux, France
Contact:

Re: EXTREMELY slow GPU slot--happens after it basically cras

Post by toTOW »

GPU timpster wrote:When the thing basically crashes, my GPU clocks revert to stock, so there is no overclock currently.
Your overclock might be unstable ... it might be a good idea to stick to stock clocks for a while and see how it goes ...
Image

Folding@Home beta tester since 2002. Folding Forum moderator since July 2008.
GPU timpster
Posts: 65
Joined: Mon Nov 02, 2015 2:57 am

Re: EXTREMELY slow GPU slot--happens after it basically cras

Post by GPU timpster »

OP: how long have you waited for the TPF to update--and what does the log say? You can check TPF times in the log, but give it 20 minutes or so to update, and it may go away. If it does not, just restart without pausing the stop that is affected, and that should fix it.
toTOW
Site Moderator
Posts: 6296
Joined: Sun Dec 02, 2007 10:38 am
Location: Bordeaux, France
Contact:

Re: EXTREMELY slow GPU slot--happens after it basically cras

Post by toTOW »

Do you realize that you are the OP ? :shock:
Image

Folding@Home beta tester since 2002. Folding Forum moderator since July 2008.
Post Reply