Computation error due to 'exceeded elapsed time limit'
22 Nov 2011 15:38:18 UTC
Topic 84784
(moderation:
Albert@Home BRP3cuda32 WUs occasionally seem to abort after ~45 minutes with a message like the following:
Quote:
22/11/2011 16:18:58 | Albert@Home | Aborting task p2030.20100913.G48.73+01.03.S.b4s0g0.00000_3232_2: exceeded elapsed time limit 2612.18 (2800000.00G/1071.90G)
I'm not sure what to make of this, since 45 minutes hardly seems like a long time and other units appear to finish without any problems. The host in question is here.
Computation error due to 'exceeded elapsed time limit'
)
To follow up on that, I just checked and I've gotten a lot of these computation errors. They all error out around 43:32 minutes, which corresponds exactly with the elapsed time limit of 2612.18 seconds from the error message.
Ironically, going by the progress rate, the tasks should finish in a little under 48 minutes on my card. So it gets pretty close ... but not close enough. I just watched as a task was aborted around 91%. I'm suspending GPU work for now.
For what it's worth, this
)
For what it's worth, this seems to be either mostly or completely fixed now. Yesterday I still had about 10 WUs with computation errors queued up for reporting; today I had none.