Why do I get cuda workunits?

Message boards : Number crunching : Why do I get cuda workunits?
Message board moderation

To post messages, you must log in.

AuthorMessage
Claes Holmerup

Send message
Joined: 25 Jul 99
Posts: 7
Credit: 9,446,728
RAC: 25
Sweden
Message 1223510 - Posted: 26 Apr 2012, 8:54:39 UTC

Since I have a low-end graphics card, I've shut off GPU processing in the preferences - but lately, I've noticed that I get cuda WU's even so. When I run out of regular WU's, the cuda WU's are just sitting there with "ready to start" status, but no tasks are processed until I get new regular ones.
I updated BOINC to the latest version, in case it was some kind of bug - but it didn't help.
Is there something I can do about it besides aborting the tasks manually?
ID: 1223510 · Report as offensive
Profile Khangollo
Avatar

Send message
Joined: 1 Aug 00
Posts: 245
Credit: 36,410,524
RAC: 0
Slovenia
Message 1223514 - Posted: 26 Apr 2012, 9:04:15 UTC - in response to Message 1223510.  
Last modified: 26 Apr 2012, 9:13:52 UTC

You can use BoincRescheduler to reschedule existing GPU tasks to CPU.

To prevent boinc from ever doing GPU work, regardless of preferences, you can create a file cc_config.xml, put it inside BOINC's data directory (not program files). On XP, it's usually C:\Documents and Settings\All Users\Application Data\BOINC
Contents:
<cc_config>
  <options>
    <no_gpus>1</no_gpus>
  </options>
</cc_config>

(and restart BOINC).
ID: 1223514 · Report as offensive
Claes Holmerup

Send message
Joined: 25 Jul 99
Posts: 7
Credit: 9,446,728
RAC: 25
Sweden
Message 1223525 - Posted: 26 Apr 2012, 9:53:48 UTC

Thanks a lot for the help! :)
After I added the xml-file, the cuda WU's which were in the list, said "GPU missing, Ready to start".
To make it easier for myself, I just aborted them rather than converting. It looked like a little too much work for so few WU's...
ID: 1223525 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14654
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1223526 - Posted: 26 Apr 2012, 9:56:03 UTC - in response to Message 1223510.  

You may have found an interesting server bug.

Firstly, both of the workrounds that Khangollo has suggested are absolutely fine, and sensible under the circumstances.

But I wonder if you wouldn't mind checking something for me?

I've been looking at your task 2413411004, which your computer received at 26 Apr 2012 | 8:50:04 UTC - about an hour ago.

Would you mind looking in the Event Log (press Ctrl+Shift+E) for around that time, and seeing if you see a bunch of messages about "Resent lost task"?

I'm asking, because the report deadline for that task is 13 days, 17 hours and 53 minutes after you received it, whereas it would normally be 13 days, 18 hours and 16 minutes for that sort of task - a 'VHAR', as we call them.

If you do see any messages about lost tasks, it may be that the server isn't checking (or honouring) your preferences when resending tasks - which would be more than a little irritating for you, I can imagine.
ID: 1223526 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51469
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1223570 - Posted: 26 Apr 2012, 14:02:55 UTC

Another possible option would be to run the Lunatics installer and not check the boxes for GPU processing. I think the scheduler should recognize that there is no app for GPU and not send WUs for it.

Or, you may wish to try running the opti app on your slower GPU and see what it can do. Unless it causes system availability issues.

If it does not work out, you can always run the installer again with the boxes unchecked and it will remove the GPU app from the app-info file.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1223570 · Report as offensive

Message boards : Number crunching : Why do I get cuda workunits?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.