My short "expert" guess is that your Atom or whatever CPU is running with *hyperthreading* i.e. 2 threads per physical core. That makes jobs run near twice as long, but doing only little more real work over the same amount of time. Your client benchmarks to be near twice as fast and claims accordingly, but reality is that it probably is not.

Anyway, my not hyperthreaded laptop did 36.4 credit on the last HCC v 6.40 for 3.6 hours, the HCC jobs being very constant which is kind of in the ballpark what these should do per task, no matter how slow or fast.

Here my one-off ''expert'' guess ends.

--//--

Think the servers would not be able to distinct an Atom with HT on and off, so how would you pair and then you'd still have these time differences of 2 times and more. Skill pairing probably also adds substantial scheduler loads, though I'd not stop the techs from implementing, since then volunteers would learn how well their hosts do compared to the wingman. Likely the mechanism would take out allot of the claim disparities, but in that if a stock Q6600 on a 32 bit client on Linux 32 bit meets with a Q6600 on 64 bit host OCeed to the limit, there'd still be major differences, in time and claim, though there should not be... it's still per Cobblestone rules worth some 36 credit and not the Atoms 84.9, or?

HCC being that constant could as far as I'm concerned be given the same treatment as C4CW. The system is set to give a X amount of credit for a known amount of calculations and a small sway room for the slight variations, no matter what cycles/efficiency the CPU runs at, but that's another avenue.

2 Euro Cents worth of input I wanted to avoid to get into ;>)

--//--

Can't be done, anhhai, and wont be done either under the new policy. There's just too much variability in most projects... the infamous non-deterministic calculations. HCC is constant in that it passes through an image analysis with many different filters, the outcome in time barely varying, thus think it's bench-markable. If it's really doable and tech time can be allotted... don't know, but locking credits in within a small range and all knowing saves allot of debate... to include my time ;-)

--//--

I looked at my results logs and found my longest variance for HCC WU to be from 1.44hrs to 1.77hrs, so about 20% difference in time. However, most of them are between 1.44 hrs and 1.6hrs

Which is better?

10 credit/hour or 16 credit/hour