2nd GPU not running tasks.
12 Jan 2022 18:51:46 UTC
Topic 226764
(moderation:
I have set <use_all_gpus> to 1 and had both gpus running until recently and once in a while the second gpu will kick in. I don't see anything in the log files that would prevent it from running. I turned on coproc_debug in Event Log options and I get the following messages and BOINC does recognize to assign a workunit to the gpu instance but thats it... Is this a BOINC issue or Einstein@home one? Here are the messages. Instance 0 seems to have addition messages but Instance 0. Just repeats on Assigning a WU to the instance.
1/12/2022 10:49:15 AM | Einstein@Home | [coproc] NVIDIA instance 0; 1.000000 pending for LATeah3011L06_692.0_0_0.0_19627524_0
1/12/2022 10:49:15 AM | Einstein@Home | [coproc] NVIDIA instance 0: confirming 1.000000 instance for LATeah3011L06_692.0_0_0.0_19627524_0
1/12/2022 10:49:15 AM | Einstein@Home | [coproc] Assigning NVIDIA instance 1 to LATeah3011L06_692.0_0_0.0_19648611_1
I have tried resetting the project and rebooting the computer
Regards,
BoincSpy
Hmm There is a hint in the
)
Hmm There is a hint in the cpu_sched_debug, message indicates that that the CPU is committed.. Message is:
1/12/2022 11:08:11 AM | Einstein@Home | [cpu_sched_debug] skipping GPU job LATeah3011L06_692.0_0_0.0_15274656_0; CPU committed
So need to figure out what CPU committed means now. I only let the have BOINC have only 75% of the CPUs.
First I would suggest that
)
First I would suggest that you make your computer(s) visible. Without this, we do not know what system you are running, what GPUs you are using, etc. You would get more help with the computer(s) visible.
Proud member of the Old Farts Association
what CPU project are you
)
what CPU project are you running?
_________________________________________________________________________
Hi GWGEORGE007, I have
)
Hi GWGEORGE007,
I have made the computers visible, computer name wks-ubuntu-005.
Regards
Digging a bit deeper
)
Digging a bit deeper their is a discussion about this on boinc forums.
GPU tasks skipped after scheduler overcommits CPU cores (berkeley.edu)
But appears there is no real solution to the problem.
BoincSpy
)
We cannot see your named computer "wks-ubuntu-005". Could you identify the computer in question by it's ID # ?
Proud member of the Old Farts Association
COMPUTER 12903076
)
COMPUTER 12903076
Does the second GPU start
)
Does the second GPU start working if you suspend CPU work? What CPU project are you crunching?
_________________________________________________________________________
Hmm - I was able to get the
)
Hmm - I was able to get the both GPUs going by deleting all of the non-GPU work units. Will see how long they continue to both work.
If you haven't already, you
)
If possible, could you copy/paste your cc_config.xml and your app_config.xml files here so we can see what you're up against?
Proud member of the Old Farts Association