Nathalie,
Using ratios would make more sense, wouldn't it? :) Thanks for clearing that
up.
(And also for clarifying the SGI compiler issue.)
Robert
----- Original Message -----
From: "Nathalie Godbout" <godbout_at_sgihud.hudson.sgi.com>
To: "R. M. Fesinmeyer" <rmf_at_u.washington.edu>
Cc: <amber_at_cgl.ucsf.edu>
Sent: Wednesday, February 14, 2001 12:00 PM
Subject: Re: Athlon 1GHz (was: Dual Origin vs dual Alfa)
> "R. M. Fesinmeyer" wrote:
> >
> > The benchmarks from Nathalie Godbout show an SGI P3 500 completing the
> > 4096WAT benchmark in 223.14s and an SGI P3 700 taking 166.22. Do these
> > numbers more or less match those obtained from "stock" (non-SGI) P3
systems?
> > If so, and if I also assume a linear decrease in computation time with
MHz
> > rating, a P3 1000 should complete the 4096WAT benchmark in something
like 81
> > seconds. This is interesting since Tru Huynh just reported an Athlon 1
GHz
> > completing this benchmark in 105-136 seconds (PGI vs. GNU compiler).
Given
> > what I've heard about the speed of athlons on this mailing list, these
> > statistics don't seem to add up.
> >
> > So is the SGI system just that-much faster than stock systems, or do
they
> > ship with a better optimized compiler? Does anyone have any benchmarks
that
> > would clarify this? (Perhaps a P3 700 and a P3 1000 on a stock system?)
> >
> > Thanks,
> > Robert Fesinmeyer
>
> Hi Robert,
>
> I think that a better estimate for a 1000MHz P3 system single-cpu time
> for 4096wat would be 111-116 seconds (this is calculated using clock
> speed ratios). sgi does not ship these systems with any compilers. I
> used PGI compilers.
>
> Nathalie
> --
> ------------------------------------------------------------------------
> Nathalie Godbout, Ph.D. SGI
> godbout_at_hudson.sgi.com Chemistry and Biology Applications Group
> (781)839-2147 One Cabot Road, Suite 250
> (978)562-7450 fax Hudson, MA
> vnet:483-2147 01749, USA
> ------------------------------------------------------------------------
>
Received on Wed Feb 14 2001 - 12:16:51 PST
