On Thu, 2010-07-01 at 14:04 +0300, Dimitris Zilaskos wrote:
> στις 1/7/2010 1:58 μμ, O/H Stephen Burke έγραψε:
> > LHC Computer Grid - Rollout
> >> [mailto:[log in to unmask]] On Behalf Of Dimitris Zilaskos
> > said:
> >> Ok, but isn't there a conversion factor between SPEC.org numbers and
> >> HEP-SPEC?
> >
> > That's a different question, it's talking about the various SPEC 06
> > benchmarks. SI00 is SPECINT 2000 and spec.org don't calculate that any
> > more. In glue 1.3 we only have SI00 (and SF00) so we have to do the best
> > we can with that. In glue 2 there is a Benchmark object with a type and
> > a value and you'll be able to publish as many as you like.
> >
> > Stephen
> >
>
> Ok thank you for taking the time to explain. The reason I am dealing
> with this is that because if I use the conversion in either direction, I
> get numbers that are way far from the actual results obtained from
> benchmark tools, for clusters with stable composition for quiet some time.
>
> Cheers,
>
>
Hi, I have a slightly offtopic question:
Are we expected to buy the SPEC benchmark suite, or is it enough to use
results provided by other people who ran it on similar hardware, or even
"extrapolate" starting from values scored by other CPU models close
enough to ours(same core, relatively close frequency)?
Regards,
Cristian
--
Cristi Măgherușan, alumnus System/Network Engineer
Technical University of Cluj-Napoca, Romania
http://cc.utcluj.ro +40264 401247
|