[Resolved] GT 650M driver: 31422: only error results for BRP4cuda32nv301

(retired account)
(retired account)
Joined: 28 Sep 11
Posts: 16
Credit: 7357648
RAC: 0
Topic 196939

With Nvidia driver 314.22 and BOINC 7.0.28 (x64) on Win7 SP1 x64 I am only producing erroneous Binary Radio Pulsar Search results with a GT 650M (mobile GPU, 28nm GK107 "Kepler").

Anyone have any suggestions?
Maybe running BRP4 successful on a similar card, if yes, what driver and BOINC version are you using?

Here the latest results:
http://einsteinathome.org/host/6392796/tasks&offset=0&show_names=0&state=0&appid=0

And this is the output from the BOINC manager messages:

03.05.2013 19:38:19 |  | NVIDIA GPU 0: GeForce GT 650M (driver version 314.22, CUDA version 5.0, compute capability 3.0, 2048MB, 8382398MB available, 730 GFLOPS peak)
03.05.2013 19:38:19 |  | OpenCL: NVIDIA GPU 0: GeForce GT 650M (driver version 314.22, device version OpenCL 1.1 CUDA, 2048MB, 8382398MB available)

I'd appreciate your remarks and suggestions. Thanks in advance and best regards.

EDIT: The exit code is - exit code -1073741515 (0xc0000135), just in case you read this after above linked results have been purged from the db.

Mark my words and remember me. - 11th Hour, Lamb of God

Horacio
Horacio
Joined: 3 Oct 11
Posts: 205
Credit: 80557243
RAC: 0

[Resolved] GT 650M driver: 31422: only error results for BRP4cud

There was a lot of this issues (with error 135) some time ago using Einstein, a Kepler GPU, BOINC 7.0.28 and Windows 7 64bits and it was no related to the drivers.
In that time, there was different fixes, sometimes a reinstall of BOINC as the right user fixed it because it was a permissions issue, but it doesnt worked always. Ussually the most general solution was to downgrade BOINC to version 6

Now, there is a new recomended BOINC version that fixes a lot of errors from 7.0.28... I think your should try first with the last recomended version (7.0.64)... Even if it doesn't fix your issue it will reduce the probability of other problems...

mountkidd
mountkidd
Joined: 14 Jun 12
Posts: 176
Credit: 12609622555
RAC: 8002583

With Boinc 7.0.28 a common

With Boinc 7.0.28 a common fix for error 135 was to uninstall OpenCL that Windows updates had provided. Boinc 7.0.64 allows OpenCL to remain in place.

Gord

(retired account)
(retired account)
Joined: 28 Sep 11
Posts: 16
Credit: 7357648
RAC: 0

Horacio, thanks for your

Horacio,

thanks for your quick reply. I installed BOINC 7.0.64 now as suggested. However, the errors reduced the daily quota for this host, so I have to wait a bit until new workunits come in. Will report results.

Thanks, mountkidd / Gord, for your reply, too.

Btw, note that the message related to GPUs has changed now and the bogus memory count is gone. :-)

Thanks for your help, regards

Quote:
03.05.2013 23:29:09 | | CUDA: NVIDIA GPU 0: GeForce GT 650M (driver version 314.22, CUDA version 5.0, compute capability 3.0, 2048MB, 1982MB available, 730 GFLOPS peak)
03.05.2013 23:29:09 | | OpenCL: NVIDIA GPU 0: GeForce GT 650M (driver version 314.22, device version OpenCL 1.1 CUDA, 2048MB, 1982MB available, 730 GFLOPS peak)
03.05.2013 23:29:09 | | OpenCL: Intel GPU 0: Intel(R) HD Graphics 4000 (driver version 9.17.10.2932, device version OpenCL 1.1, 1624MB, 1624MB available, 45 GFLOPS peak)
03.05.2013 23:29:09 | | Version change (7.0.28 -> 7.0.64)
(...)
03.05.2013 23:30:35 | Einstein@Home | Requesting new tasks for NVIDIA and intel_gpu
03.05.2013 23:30:38 | Einstein@Home | Scheduler request completed: got 0 new tasks
03.05.2013 23:30:38 | Einstein@Home | No work sent
03.05.2013 23:30:38 | Einstein@Home | (reached daily quota of 13 tasks)
03.05.2013 23:30:38 | Einstein@Home | Project has no jobs available

Mark my words and remember me. - 11th Hour, Lamb of God

(retired account)
(retired account)
Joined: 28 Sep 11
Posts: 16
Credit: 7357648
RAC: 0

Installing BOINC 7.0.64 was

Installing BOINC 7.0.64 was not the solution, unfortunately. Still the same error code.

Gord, I guess you mean the OpenCL 1.1. support for Intel cpus? Not sure if this was installed via Windows update or together with the Intel video driver. Anyway, I'll give it a try and deinstall it. Have to wait another day or so again until quota is sufficient...

Also added an ignore_intel_dev tag into cc_config now. Should not cause any harm at least. :)

Mark my words and remember me. - 11th Hour, Lamb of God

(retired account)
(retired account)
Joined: 28 Sep 11
Posts: 16
Credit: 7357648
RAC: 0

Yes, uninstalling OpenCL 1.1.

Yes, uninstalling OpenCL 1.1. for Intel did the trick. BRP4 are now calculated without an error after seconds. Thanks again to both of you, issue solved.

Mark my words and remember me. - 11th Hour, Lamb of God

(retired account)
(retired account)
Joined: 28 Sep 11
Posts: 16
Credit: 7357648
RAC: 0

to edit thread title

to edit thread title

Mark my words and remember me. - 11th Hour, Lamb of God

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.