I run 2 xeon quads clocked to 3.5 I bought a 9650 . and the bench mark is 2/3 the others and its at 3.0 is the gpu in the other units adding to thee rating ?

more from an 'outside' observer 2 cents comments:
- accordingly rosetta@home doesn't use the gpu (yet), perhaps someone may like to correct that if things has changed :)
- accordingly boinc use the whetstone benchmark, if you are referring to that boinc points/credits which is based on whetstone benchmark, my guess is that these points/credits do not necessarily measure the prowess of crunching rosetta@home. it may literally be doing r@h faster, but the points/credits are more of the whetstone benchmark extrapolated based on perhaps the time taken to run the jobs. but 'underneath' during r@h runs, it (the xeon cpu) may have processed more (models) compared to say another cpu that gives a better whetstone score but in fact did less models compared to your xeon cpu

the notion is that 'other' cpus rate the r@h job being 'worth' x whetstone credits, while your cpu deemed it is worth y whetstone credits. it may just be the same amount of r@h 'work', in the end the 'average' of the ratings is awarded e.g. a higher whetstone credits/points.

just 2 cents, i'm not sure if it is correct or if it even made sense |)

imho rosetta@home is very much a *benchmark* on its own, perhaps measured in terms of how many 'models' a cpu can process in a unit time. just that for now it'd be a little difficult to say which 'model' is most representative of r@h jobs & for which version of optimised r@h code it applies. & of all things this r@h 'benchmark' is perhaps a 'real world' benchmark that involve solving problems of 'real world' complexity with all the necessary 'compromises', certainly very different compared to synthetic benchmarks such as whetstone or linpack which probably does not reflect 'real world' scenarios lol

-

the CPU was crashing so I had to trash it . but anything I run only uses 2 cores [ the gpu runs is core ] leaving a full core idle .