With the new BoincTasks with numbered checkpoints it is less boring to monitor a CEP2 WU.
So I have just watched the end of a job #2 in a CEP2 WU running alone of its kind together with 3 HCC ones.

And I could perfectly see about 6-7 minutes of CPU time vanishing from the total CPU time when this job completed!

In fact I think that the new total CPU time was actually the CPU time of job #2.

And until the total CPU time reaches what it was at the end of job 2 BoincTasks was showing a beautiful red line with 0 % CPU activity, which can be considered as normal if its author is computing CPU activity by comparing consecutive measurements.

If this is happening a few times during the whole WU that explains much (or all?) of the differences between elapsed time and CPU time at the end.

Note that from Ubuntu's system monitor I have not seen any pause between the consecutive processes, so things seem to be normal in this area.

After I initially noticed Ubuntu reporting a truer CPU time, I didn't go back to do some spot checks. Reading this thread reminded me and yes I've noticed the CEP2 jobs on my Ubuntu box on average are about 45 minutes CPU time short as compared to Elapsed time. But, the CPU time is closer to Elapased at least in Ubuntu than the other three major distros I tried. Those other three were missing over 1/2 to 2/3 the actual CPU time.

i think the same problem exists on MacOS 10.6.3 ...

i dont even remember when the CPU time was close to Elapsed time

and i've never got a 2-days run time log on this 2-core laptop with 24/7 crunching WCG.

screenshots attached...

current WU



device info log (took biggest 24/7 period as example)