Run times are variable even within a batch, such are the traits of non deterministic calculations i.e. cant predict when the lowest/best energy docking is found.

Anyway, reviewing the BOINCview completed task history for my quad since Dec. 1 the run times have gone from 4 hours all the way up to 11+, down to 7 and the last few nearing 9 hours again. On a p4 that probably translates to 20 hours.

The project target is an average of 8 hours, but we seem to be running near 10 hours at the moment, which confirms your observation of a current upward trend.

[added: hi there Kawrence... 20 seconds only ;>)

greetings.
my pc is really powerful, the longest work unit I've had thus far took just over 8.12 hours, now thats fast!