GPUGrid and BOINC points

[123bob]

As I understand it right now, the BOINC points we produce are somewhat leveled across all BOINC projects by some mechanism. Is it G Flops or something like that? I'm wondering why my 260-216 produces WUs worth a rough average of 9500 PPD BOINC points vs a good OC quad, which produces on the order of 2500 PPD.

Is there true equivalence here? Anyone know?

Sorry for the dumbazz question if it's some obvious answer. This will certainly not be my last dumb question....

Regards,
Bob

[[XC] riptide]

Bob.... back in Rosetta there was the issue brought up about cross project scoring and how it didn't exist. Unless this has changed, I will assume that BOINC points across diff projects are NOT the same. But if you can prove me wrong cool. I see seen some of your posts recently that seem to compare the Q6600 with a gfx on boinc points alone.... but like I said I don't think you can compare cross projects.

[fallwind]

http://boinc.berkeley.edu/wiki/Computation_credit So, 100 BOINC credits per GigaFLOP is the standard. How they apply this to GPU's I don't know since CPU's are double-precision and GPU's are single.

[123bob]

Thx for the info guys. The article at the Berkley page is informative. I'll have to study it a bit.

I was hoping to have a mechanism to directly compare CPU crunchinng to GPU crunching. It appears to be not so simple.

It may help if I asked at WCG and GPUGrid how the claimed credit is derived. If they both follow the Berkley standard, a direct comparison should be possible using claimed credit. However, it says on the page that some projects do their own method of calculating claimed credit. In this case some backwards extrapolation would be required to make it apples-to-apples again.

Regards,
Bob

[STEvil]

Wish I knew how it was scored.

Put my 8800GTS/512 (800c/2000s/1050m) and 8600GT (702c/1728s/777m) on and they scored 9200 points on the first day of work so far. XP64 with 8 cores of E5450 @ 3.6ghz behind them... also doing WCG at the same time ;)