After testing here are my R@h and WCG results (taken from the Projects tab of BOINC...which express all the totals as Host Average Credit which are pretty much on target with
http://stats.free-dc.org, which lags about 12 hours behind).
All are identical servers at the same physical location, with roughly the same uptimes in their respective groupings. They aren't at their maximum RAC but have run for 8 days in the first grouping and 5 days in the second grouping which should minimize the statistical variation.
First Grouping:
68 6431 WCG
70 7257 R@h
72 7642 R@h
73 6424 WCG
Second Grouping:
74 5605 R@h
75 5115 WCG
This means the average First Grouping R@h is 7449, and WCG is 6427, meaning a 15% edge to R@h, the second grouping gives a 10% edge to R@h.
I just started one that is 50% WCG and 50% R@h which would have been a better test subject from the get go but won't have meaningful data anytime soon.
Based on that data which I feel is statistically significant, our 1.5 multiplier for WCG is too high despite what my previous research had implied, which is a shame as I personally like that project better. My feeling is we need to reduce the multiplier as soon as feasible to between 10% (1.1 multiplier) and 15% (1.15 multiplier) to equalize the difficulty with the 15% (1.15) equalization seeming at this point to be the more equitable factor.
Hmmm - there is still something preventing us from reconciling this situation though.
I committed two identical servers to this ( 6 core amds with the same single SSD in each one, they are about 4 years old) and both have win7, and both are only running boinc.
My RAH server has settled down with 4500 RAC, my WCG server has settled down with 3200 RAC (this is a factor of 1.40 for WCG at this point) - close to our 1.50 prod setting.
So I wanted to point out, when I started this test, I deliberately went into my WCG Web settings, and selected All projects (I checked every box). First question: Are you running all WCG project or just the defaults? (Right now its running Mapping Cancer markers and Fight Aids for example).
