is there any performance indication from Pi to FAH?

I'm just wondering if there's any performance indication in between SuperPi and F@H?? I know there's many software (benchie or real world) doesn't really represent the whole picture. e.g. some processor is good at 3D, but slow in F@H, or some is excellent in F@H, but it's slow on other benchie. therefore, I'm just want to see IF SuperPi is a good indicator to F@H performance or it won't show anything at all in F@H world with a fast SuperPi time??

reason I'm asking this question is I want to see/predict the performance of Conroe/Merom/Yonah in F@H by their SuperPi time........ IF SuperPi has ANY indication of F@H performance.....

From what i understand of SuperPi (which is not much!), it is not going to exercise the memory very much, mostly cpu.

Also, not sure what its algorthim uses but i would think it uses interger instructions vs FAH using floating point heavily.

ok, so SuperPi time pretty much has nothing to do with F@H's performance then....

I would say not.
Folding projects require different things from the computer to run efficiently. QMD cores require mad bandwidth to fold effectively, while tinkers are all about floating point, while Dgromacs are all about SSE2, and the 147x-1481 were all about crazy fast cache. For example:

The dothan was a great superPi cruncher, yet average on tinkers (almost pure floating point) and about average in gromacs, and weak on QMD's (lack o bandwidth), while they rule all with the 147x-1481 projects (fast cache)

how about this...?? any better incidcation?? or still no...??

http://www.ocforums.com/showthread.php?t=452698

how 'bout Cinebench03?? I remember a 920 @ around 4ghz could score about the same in multi-cpu test as the yonah @ around 3ghz, ~ 840 score.....

or there's NO app that will reflect the performance in F@H at all??

How about Sandra CPU Media Benchmark? Does not FAH use floating point?

I guess one could take four benchies (one for sse, one for floating point, one that favors fast cache, and one for sse2), come up with an average of some sort, and perhaps have a general indication of folding performance. I wish i could give you a viable answer to gauge folding performace. I tried to make a core performace guide a while back which may have helped, but different projects using different cores produce different results on diffferent puter setups (yes ram quantaty and speeed does play a factor in some projects)


As far as what the yonah core will do as a folder, you can pretty much use dothan numbers to suggest how it will do, except in the matter of QMD's where mem bandwidth will help out tremendously, except if you run two clients then running out of bandwidth will hurt ya.

maybe this can help a little
link

I look at pi and Sandra. I would think that Sandra memory bench would be a good indicator. I run PI 1m for a quick stability test and 4m and 32m for stability. If Pi errors on a 1m test is propably ee out a work unit.

what about the bench that folding does when it starts

doesnt it perform some test or am i mistaken

It does bench with verbosity 9 ( i think). Everytime you make a change you'll need to restart the core and you'll have to wait and see for any errors.