Einstein Ranks Lowest in Credit Granted
1 Mar 2006 2:55:00 UTC
Topic 190846
(moderation:
Bruce Allen and the powers that be ....Of all the projects in Boinc I participate in Einstein ranks lowest in credit granted per CPU time MY LIST-Einstein (lowest),LHC,Predictor,Burp,uFluids,Rosetta,CPDN,and Seti (highest especially with an optimized client). Now I know you were clamouring for more participation , but the crunchers who are credit hogs (me included) may shy away from your project unless you can adjust credit to be at least average. I know the science buffs will support you no matter what...but I am a mix of both science base and credit(look at my sig) and am now a seti whore due to credit acheived ,yet this is only temporary for many reasons. This IS a problem if you wish more output as I am sure many can correlate similar findings so I request you address this situation. Thank-you
Einstein Ranks Lowest in Credit Granted
)
I read you loud and clear. Hopefully we can convert our application to using the 'fixed credits per WU' scheme: that should address this.
Director, Einstein@Home
RE: I read you loud and
)
Thank-you Bruce for the quick response AND especially for addressing this situation...I Hope to be crunching more for you soon :)
[EDIT] Oh and I DO LIKE being rewarded (in Seti) For being able to use sse2,sse3,and hopefully sse4 (hint) [EDIT]
the x86 einstien app is
)
the x86 einstien app is universal. If you have SSEN it will use it automatically.
what's all the fuss about the
)
what's all the fuss about the low credit per CPU time....?
isn't the credit accreditation based on gflops of CPUs....
to the original poster, if you wish to get higher credits per CPU, get a PPC-based Mac, or just dedicate your CPU to one particular project that is optimised for your CPU, in your case, SETI@home.... no offence intended....
RE: the x86 einstien app is
)
Do you run seti? ...You get rewarded optimized in thoughput hence credit for sse3 vs sse2 etc,,
RE: what's all the fuss
)
Hello Elphideus .... Can"t you see? I AM Seti crunching duh look at my Seti rac in Seti on my sig ...and I am trying to help this project NOT hurt it by this post....More Crunchers would run Einstein if credit was equal....Bruce see's my point ....Why don't you?
RE: Hello Elphideus ....
)
that's the entire point....
SETI is optimised for it's SSE and SSE2 instructions on the Intel (if that's what you're using) and Einstein is optimised for Motorola G4/IBM G5 Altivec.... I've tried running Einstein on a P4 1.8Ghz non-HT, it sucks big time, not to mention completing a WU in 12 hours, but with a single core 1.9Ghz G5, I've got to complete them in under 3 hours.... so it's all about optimisation. So how can credits be equal..? Put this onto SETI, there ain't different between these two CPU times because the G5 was not optimised for, whereas it is optimised under Intel's SSE and SSE2, and the abundance of cache (correct me if I'm wrong).
and mind you, credit is accredited from an average of 3 to 4 results of different CPU crunchers (if you allow me), so RACs don't really play much of a role here....
oh btw, I only have one single-core G5 dedicated to Einstein, and please read my RAC....
RE: RE: Hello Elphideus
)
OK Eliphedius I will try 1 more time to get thru to you.... This post is about trying to get more crunchers working this project...NOT about your or mine rac or library functions....Do you not think more people would crunch Einstein if credits equalled Seti optimized or CPDN or Rosetta? That IS the real thing here not your ego....I am glad you are not of Project Staff as you can not see the Forest between the Trees ....:)-
RE: OK Eliphedius I will
)
I believe you've got everything turned upside down. It's not about my egoism that is in question here, it's more about your obsession with credits. Remember of what you're saying about your SETI RACs....? Psychological speaking, no pun intended...
Unfortunately, credit distribution differs from WU sizes to CPU integer and FP unit, and this differs all across the projects and platforms. If you were to asked about having one particular project to equalise/stardardise the distribution of WU credits, might as well direct your suggestions to the BOINC founders/programmers. After all, it's the BOINC client's decisive factor here. And why the hell would BOINC had our CPUs benchmarks internally once in a while if the credit distribution ought to be standardised as you've said. Might as well have our credit based on WU counts instead... get my point...?
So if you're so dissatisfied with the credit outcome, might as well dedicate all your CPU time to whatever project you think it's worthwhile of your system instead of messing the already messed up credit system.
To tell you the truth, I'd never like the idea of the BOINC credit system. I would have prefer the more traditional WU credit system employed by SETI Classic. The BOINC credit system is in place because of the irregularities in the WU sizes and the differential crunching time of individual CPUs based on the latters' abilities in terms of Integer and FP counts. And I've switched from SETI to Einstein because the G5 is more optimised for the latter.
And finally, stop getting my nick wrong. You've got it spelt wrongly in all your posts....
RE: isn't the credit
)
As far as I know, that will be part of the new system Bruce refered too.
Join the #1 Aussie Alliance on Einstein