GPU Upgrade from GT240 to GT430 results in 100% Compute Errors


log in

Advanced search

Questions and Answers : Windows : GPU Upgrade from GT240 to GT430 results in 100% Compute Errors

Author Message
Profile woodyrox
Volunteer tester
Send message
Joined: 7 Apr 01
Posts: 34
Credit: 16,069,169
RAC: 0
United States
Message 1142213 - Posted: 18 Aug 2011, 19:09:30 UTC

The error I get is:

Cuda error 'cudaMemcpyToSymbol(cudaAcc_GaussFit_settings, (void*) &settings, sizeof(settings))' in file 'c:/sw/gpgpu/seti/seti_boinc/client/cuda/cudaAcc_gaussfit.cu' in line 454 : invalid device symbol.


The link to my computer with errors is here:

http://setiathome.berkeley.edu/results.php?hostid=6109088&offset=0&show_names=0&state=5&appid=

I'm running the most recent 280 nVidia drivers. While I wait for insight from others, I'll roll back to 266 drivers that I've used with a GT430 before.
____________

Profile woodyrox
Volunteer tester
Send message
Joined: 7 Apr 01
Posts: 34
Credit: 16,069,169
RAC: 0
United States
Message 1142248 - Posted: 18 Aug 2011, 19:59:27 UTC - in response to Message 1142213.

It looks like the GT 430 will only run cuda_fermi tasks. My computer only had cuda and cuda_23 tasks. I aborted those & was given fermi tasks and they seem to be running ok.
____________

Profile Lint trapProject donor
Send message
Joined: 30 May 03
Posts: 871
Credit: 27,841,029
RAC: 6,775
United States
Message 1142713 - Posted: 19 Aug 2011, 19:34:44 UTC



As you have discovered, a FERMI architecture Nvidia card does not 'crunch' well unless you have a FERMI compatible CUDA app installed. The video driver makes no difference (as long as it meets the requirements for the video card).

I see you have returned and validated a few cuda_fermi workunits. You're on the right track now!

I don't believe you needed to abort the previous workunits, but it's done now.

Keep crunching!

Lt

Profile woodyrox
Volunteer tester
Send message
Joined: 7 Apr 01
Posts: 34
Credit: 16,069,169
RAC: 0
United States
Message 1155247 - Posted: 23 Sep 2011, 14:43:43 UTC - in response to Message 1142713.
Last modified: 23 Sep 2011, 14:46:07 UTC

So I'm getting ready to downgrade a GPU in one of my computers from a 9800 GTX+ (1GB RAM) to a 9600 GSO (768 MB RAM). If the work units don't run or if they error out, how do you suggest this downgrade be performed -- other than aborting all of the work units?

I'd like to avoid aborting the 654 work units waiting, if at all possible.
____________

Profile arkaynProject donor
Volunteer tester
Avatar
Send message
Joined: 14 May 99
Posts: 3728
Credit: 48,768,260
RAC: 1,737
United States
Message 1155279 - Posted: 23 Sep 2011, 15:47:04 UTC

That will not be a problem as they are compatible cards and apps.
____________

Profile woodyrox
Volunteer tester
Send message
Joined: 7 Apr 01
Posts: 34
Credit: 16,069,169
RAC: 0
United States
Message 1155301 - Posted: 23 Sep 2011, 16:20:07 UTC - in response to Message 1155279.

Thanks for your reply. Looks like you are absolutely right. I did the downgrade and the partially completed work unit picked up right where it left off. I'll monitor for errors and stuck work units, but so far, so good!
____________

Questions and Answers : Windows : GPU Upgrade from GT240 to GT430 results in 100% Compute Errors

Copyright © 2014 University of California