Strange CUDA calculation

[Sutaru Tsureku]

I looked to the task list of my GPU cruncher and found three strange outputs:

http://setiathome.berkeley.edu/result.php?resultid=1396515334
http://setiathome.berkeley.edu/result.php?resultid=1396515282
http://setiathome.berkeley.edu/result.php?resultid=1396515175

For sure, no reboot.

I don't think because of EDF.
3 day WU cache.

BOINC V6.6.37 , with nVIDIA_driver_190.38, CUDA_V12_app and CUDA_V2.3 .

Strange is, that nowhere is written that a restart was happened.

I have in the preferences 'write to disk every 60 sec.'.
IIRC with BOINC V6.6.x it's x4? So every 4 min. checkpoint?


Any ideas?

Thanks!

[Josef W. Segur]

I looked to the task list of my GPU cruncher and found three strange outputs:

http://setiathome.berkeley.edu/result.php?resultid=1396515334
http://setiathome.berkeley.edu/result.php?resultid=1396515282
http://setiathome.berkeley.edu/result.php?resultid=1396515175

For sure, no reboot.

I don't think because of EDF.
3 day WU cache.

BOINC V6.6.37 , with nVIDIA_driver_190.38, CUDA_V12_app and CUDA_V2.3 .

Strange is, that nowhere is written that a restart was happened.

I have in the preferences 'write to disk every 60 sec.'.
IIRC with BOINC V6.6.x it's x4? So every 4 min. checkpoint?


Any ideas?

Thanks!

A restart before the first checkpoint is always a fresh start since there's no saved data to start from. Only the middle one of the three you linked has a BOINC run time which is more than 240 seconds plus the "Wall-clock time elapsed since last restart:", and it's only ~27 seconds longer than that.

All 3 were reported together, so my guess is that BOINC shut them down to do its benchmarks. If you look at the message logs they'll probably tell what happened more reliably.
                                                                                     Joe