Reference the Beta forum post by uplinger explaining the mechanics for C4CW about a month ago and 'per-task' [perceived] credit disparity, I'm pleased to share following for a dual boot W7 / Linux quad Q6600 device running off a dozen on each.

Windows: Runtime average about 4.4 hours
Credit/Task Grant: About 78-79
Credit/Hr: 17.8

Linux: Runtime average about 4.77 hours
Credit/Task Grant: 88
Credit/Hr: 18.2

Currently for Clean Water, the Windows 7 platform has the upper hand by about 8% more throughput.

Personally, if that 8% could also be realized for the HCC (the 'par' speed of processing taken as a firm promise), I'd be over the moon.

-- Raker ;-)

PS, the variations on run time are minimal, but had one "on the road" passing through in between on the duo, that ran over 9 CPU hours instead of the clockwork regular 6.25. Never saw the outcome since the valid tasks are removed so quickly. BOINCTasks logged it as target03-0011005023_0 on Jan 5. Anyone else seeing extreme one-off outliers for this science/target?

PPS, noted that Clean Water is since about Dec 23 on full normal share at WCG (there was I recollect an upload bandwidth or intake volume constraint). C4CW even briefly peeked top position leading up to the new year, which was probably also driven by the run-time change... yes noted that if tasks are resized, the volume of tasks released to the feeder is not immediately adjusted ... saw this several times now in recent months, but it could be my creative thinking ;-)

PPPS: The research mean run time from old target02 to new [including size-up] target03, went from 4.14 CPU hours to 5.02 CPU hours, not even 25%. Before 6.14 went live and target03 was upsized under 6.13 the mean runtime briefly hit 6.95 hours, so the effective "all platform" average works out to be that we're doing 1.38 times the amount of work from before in the same unit of time. Pretty cool.