cuda_fermi running on non Fermi GPU


log in

Advanced search

Questions and Answers : GPU applications : cuda_fermi running on non Fermi GPU

Author Message
Profile stephenm
Send message
Joined: 14 May 99
Posts: 3
Credit: 7,942,144
RAC: 746
Canada
Message 1277931 - Posted: 31 Aug 2012, 5:17:56 UTC

I have a Geforce GTX 680 and GTX 275 installed on my system. Running Windows 7 x64 and BOINC 7.0.28. SETI 6.10 cuda_fermi WUs are assigned to the GTX 275 even though it isn't Fermi compatible hardware. Is this a BOINC or SETI bug? I figure it should assign WUs to the appropriate hardware. Currently have the 275 disabled so I don't make any more computing errors.

Profile Gatekeeper
Avatar
Send message
Joined: 14 Jul 04
Posts: 887
Credit: 176,479,616
RAC: 18
United States
Message 1277946 - Posted: 31 Aug 2012, 5:42:56 UTC - in response to Message 1277931.
Last modified: 31 Aug 2012, 5:45:09 UTC

I have a Geforce GTX 680 and GTX 275 installed on my system. Running Windows 7 x64 and BOINC 7.0.28. SETI 6.10 cuda_fermi WUs are assigned to the GTX 275 even though it isn't Fermi compatible hardware. Is this a BOINC or SETI bug? I figure it should assign WUs to the appropriate hardware. Currently have the 275 disabled so I don't make any more computing errors.


AFAIK, there is no difference among the stock apps 608, 609 and 610. There are probably a few regulars here who can explain the error messages on the 275 unit, but long story short usually a kepler (the 680) and a pre-fermi (the 275) are a bad combo in the same machine.

I'd recommend you check out the Lunatics optimized app (see thread in Number Crunching forum). That would allow you to do more work faster, and be ablee to run multiple units on the one 680 card.
____________

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2570
Credit: 5,879,088
RAC: 2,757
Bulgaria
Message 1278808 - Posted: 1 Sep 2012, 11:39:25 UTC - in response to Message 1277931.


It seems to me that all 'Error tasks for computer 6426047' are from the GeForce GTX 680
http://setiathome.berkeley.edu/results.php?hostid=6426047&offset=0&show_names=0&state=5&appid=

E.g.
http://setiathome.berkeley.edu/result.php?resultid=2582192637

BOINC will Not use by default the GTX 275 (in your mixed system with GeForce GTX 680 - BOINC uses only the most capable GPU) unless you use in cc_config.xml
<use_all_gpus>1</use_all_gpus>

http://boinc.berkeley.edu/wiki/Client_configuration

Are you sure the GTX 275 was in fact used?


____________



- ALF - "Find out what you don't do well ..... then don't do it!" :)

Profile stephenm
Send message
Joined: 14 May 99
Posts: 3
Credit: 7,942,144
RAC: 746
Canada
Message 1285072 - Posted: 18 Sep 2012, 5:28:44 UTC - in response to Message 1278808.

GTX275 is the one generating the errors. GPUs are numbered and listed in the event log. When this problem occurs, I am using cc_config.xml. 8800GT and GTX 275 worked marvelous together. GTX 275 and 680 not at all. =)
____________

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2570
Credit: 5,879,088
RAC: 2,757
Bulgaria
Message 1285084 - Posted: 18 Sep 2012, 7:00:07 UTC - in response to Message 1285072.
Last modified: 18 Sep 2012, 7:24:14 UTC

GTX275 is the one generating the errors.

Some of the latest (18 Sep 2012) errors are from GeForce GTX 275

But those from 14 Sep 2012 are from GeForce GTX 680 (as was the case when I checked on 1 Sep 2012, at that time all of them showed GTX 680):

setiathome_CUDA: CUDA Device 1 specified, checking...
Device 1: GeForce GTX 680 is okay
SETI@home using CUDA accelerated device GeForce GTX 680
.....

http://setiathome.berkeley.edu/result.php?resultid=2604714532
http://setiathome.berkeley.edu/result.php?resultid=2604694358
http://setiathome.berkeley.edu/result.php?resultid=2604694350
.....


These 2 are from 28 Aug 2012 and show only GTX 680
http://setiathome.berkeley.edu/result.php?resultid=2582100946
http://setiathome.berkeley.edu/result.php?resultid=2582100958

Stderr output <core_client_version>7.0.28</core_client_version> <![CDATA[ <message> Incorrect function. (0x1) - exit code 1 (0x1) </message> <stderr_txt> setiathome_CUDA: Found 1 CUDA device(s): Device 1 : GeForce GTX 680 totalGlobalMem = -2147483648 sharedMemPerBlock = 49152 regsPerBlock = 65536 warpSize = 32 memPitch = 2147483647 maxThreadsPerBlock = 1024 clockRate = 1058500 totalConstMem = 65536 major = 3 minor = 0 textureAlignment = 512 deviceOverlap = 1 multiProcessorCount = 8 setiathome_CUDA: CUDA Device 1 specified, checking... Device 1: GeForce GTX 680 is okay SETI@home using CUDA accelerated device GeForce GTX 680 setiathome_enhanced 6.09 Visual Studio/Microsoft C++ libboinc: 6.3.22 Work Unit Info: ............... WU true angle range is : 0.427204 Optimal function choices: ----------------------------------------------------- name ----------------------------------------------------- v_BaseLineSmooth (no other) v_GetPowerSpectrum 0.00018 0.00000 v_ChirpData 0.01435 0.00000 v_Transpose4 0.00645 0.00000 FPU opt folding 0.00328 0.00000 CUFFT error in file 'd:/Projects/SETI/seti_boinc/client/cuda/cudaAcc_fft.cu' in line 62. </stderr_txt> ]]>


*****

I don't say that all tasks are errors, you have 'Valid (139)'
I say that your confidence that only "GTX275 is the one generating the errors" is not correct.


____________



- ALF - "Find out what you don't do well ..... then don't do it!" :)

Questions and Answers : GPU applications : cuda_fermi running on non Fermi GPU

Copyright © 2014 University of California