Quad Core plus 2 cuda uses only 3 cpu cores.

Questions and Answers : GPU applications : Quad Core plus 2 cuda uses only 3 cpu cores.
Message board moderation

To post messages, you must log in.


Send message
Joined: 29 Jul 99
Posts: 12
Credit: 13,474,762
RAC: 0
Message 1041996 - Posted: 14 Oct 2010, 7:58:52 UTC

All of a sudden this quad core with 2 cuda cards, started utilizing only 3 of the 4 cpu cores.

I've read that this un-bottlenecks gpu processing, but if i watch cpu utilization, it floats around 80%. I would like to test this, but i can't seem to get the 4th cpu core back online.

Any advice appreciated.
ID: 1041996 · Report as offensive
Profile Gundolf Jahn

Send message
Joined: 19 Sep 00
Posts: 3184
Credit: 413,080
RAC: 85
Message 1042009 - Posted: 14 Oct 2010, 11:01:55 UTC - in response to Message 1041996.  

If the values for "1 CUDA + 0.x CPU" in the Status column of the Tasks tab sum up to more than 1 cpu; BOINC will use that count of CPUs to feed the GPUs.

There's no easy way around that when you are using stock applications.

One way would be using the anonymous platform mechanism. Then you yourself can set the fraction of CPUs used in the app_info.xml file.

Another way might be telling BOINC that you have 5 cores instead of the really existing 4 ones, using the cc_config.xml file, but that's only speculation on my side.

If the links I gave you don't provide enough information, you should ask in the Number crunching subforum for help.

Computer sind nicht alles im Leben. (Kleiner Scherz)

SETI@home classic workunits 3,758
SETI@home classic CPU time 66,520 hours
ID: 1042009 · Report as offensive

Send message
Joined: 29 Jul 99
Posts: 12
Credit: 13,474,762
RAC: 0
Message 1042244 - Posted: 15 Oct 2010, 10:12:25 UTC - in response to Message 1042009.  
Last modified: 15 Oct 2010, 10:13:29 UTC

ah, yes. i see now. each cuda card (2) says 0.53 cpu, so that would add up to 1 core. if i recall, they _used_ to read 0.12 each. i wonder why that changed. it doesn't appear accurate according to task manager. 0.12 would however.

anyhow, settting the config file to max 5 cpu has me back to 4 cpu and 2 gpu tasks..so far, with no noticable loss in performance on either.

thanks a bunch. this will help me catch my bud in unit counts..the ten year battle resumes!!!! eat this iceman!! lol
ID: 1042244 · Report as offensive
Robert P. Herbst
Volunteer tester

Send message
Joined: 10 Jun 03
Posts: 45
Credit: 62,347,173
RAC: 3,349
United States
Message 1042697 - Posted: 16 Oct 2010, 15:20:48 UTC

I have 6 computers going here, all with quad core and three GPU cards in three card SLI configuration in each of them. I use the NIVIDIA 9800GTX+ cards.
Only two of the computers are doing CUDA projects but still have CUDA work to do in the TASK list. Several computers don't even get CUDA work.
Why is this and who is it that decides who gets what work. With all that's being said about needing help, one would think SETI would use every inch of usable computing space.
I'm getting a bit tired of spending big bucks to make my computers ready yo do SETI work, only to have SETI withhold the work and my computers sit idle.
I have tried running other projects only to find they don't get CUDA work either.
Anyone with any ideas?
Please Visit Mount Perry, Florida
Home to Florida's Only Snow Capped Mountain
ID: 1042697 · Report as offensive
Profile BilBg
Volunteer tester

Send message
Joined: 27 May 07
Posts: 3717
Credit: 9,264,565
RAC: 311
Message 1042765 - Posted: 16 Oct 2010, 18:47:53 UTC - in response to Message 1042697.  

1) Try to switch SLI Off
2) Look in Messages tab for a clue why the GPUs are not used
(in BOINC 6.12.2 I think it is hidden in Advanced menu)

- ALF - "Find out what you don't do well ..... then don't do it!" :)
ID: 1042765 · Report as offensive

Questions and Answers : GPU applications : Quad Core plus 2 cuda uses only 3 cpu cores.

©2018 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.