You claim about 11 per hour, so why credit 20? The 107/67 condition has been submitted under an enquiry request. What type of CPU and OS and Client version are you running?

At WCG, each Work Unit (WU) is normally sent out to 2 of us crunchers, and the number of Points Credited is normally the average of the Points Claimed by each of the 2 crunchers.

In the WCG Results Status pages, if you click on the names of WUs that have been declared Valid, you will see the CPU time and Points Claimed for both you and the other cruncher. The ratio between your CPU Time and Points claimed should be fairly constant if you only have 1 machine, but the Points Claimed and Times will be different for the other crunchers.

That the other guys' times are different should be obvious - their machines have different speeds to yours. Speed accounts for most of the differences in Points Claimed per hour, but there are also other fudge factors and I don't know much about these. Machines with more resources - memory, disc space, etc, are valued more than the little guys, for example. The bullies win yet again - there's no reward for the struggling underdog at WCG.

I think that the CPU speed factor in Points Claimed is measured by BOINC's CPU Benchmarks. I have evidence that these ars not an accurate way to estimate the true throughput of WUs. I too have an AMD64 X2, and it's overclocked from 2.2GHZ to 2.76GHz. However, to reach this CPU speed, I've had to reduce the the ratio of memory speed to CPU speed, so the increase in my FAAH crunching speed IS NOT in proportion to the increase in clock speed. However, the increase in BOINC CPU Benchmarks speed IS in proportion to theincrease in CPU speed. I believe that the difference is because when the Benchmarks are running, only a small area of RAM is accessed, and the CPU can use its cache memory rather than the much slower external RAM. However, running Autodock to crunch FAAH WUs requires many accesses to external RAM so I lose some of my overclocking gain. This effect is exaggerated on the A64, as it's a bit short of cache (aren't we all? ) with only 512Mb/core compared with, say, the Intel Q9450's 3Mb. I have observed this behaviour directly, using a freely-downloadable program called perfmonitor.exe . Running BOINC Benchmarks, the L2 Cache Success Rate is over 99%, while it's only around 86% running Autodock.

I also tested my theory with WCG Results Status results. I added my Points Claimed, and the other guys' Points Claimed, for 10 WUs crunched with my slow RAM setting and calculated the Me:Them ratio. I repeated this with another 10 WUs crunched with the standard RAM speed. The ratio seems to be about 5% higher when I'm running with slow RAM, though more results need to be compared to give a more accurate estimate.

That test shows the effects of changing RAM behaviour on a single machine, but it does not give info on the effect of the amount of cache available. Gefore I overclocked it, my A64 was about 15% optimistic in its claims. This could be due in part to it having less cache RAM, more main memory or more disc space, etc, than than the average. There may also be differences in the relative performances of different CPUs when running the mix of CPU instructions in the bechmarks compared to the instructions in the real thing.

(Now off-topic): Although it's inefficient, I'm leaving my machine at its slow RAM setting because I think it crunches more WUs than with other settings. It also inflates my Points Awarded by a considerable amount, and has an incremental contribution to world-wide inflation. If WCG don't like it, they can fix their benchmarking system to be fairer!

PS: I've also ordered the bits for a 2nd computer, partly because it will become my TV set ("HTPC") after my old one died, but mainly to give WCG quite a few more MIPs. Q9450, Asus P5K3 Deluxe, etc. I intend to track the Claimed/Awarded ratios with both machines, and to keep you posted.