What do you guys thing about the P4

[Phillip D. Matz]

OK!  So when is someone with one of these P4 systems going to run G98
comparison tests?  The Athlon is faster than a P3 for routine G98 stuff (HF,
DFT, MP2) but if we go by Tim's statement below then the P4 should blow away
an Athlon.  After all, inverting matrices is essentially computational
chemistry (loosely speaking of course) and I would really really like to see
those numbers in case it is true!

Respectfully,

Phillip Matz
matz at wsunix.wsu.edu

----- Original Message -----
From: "Tim Carlson" <tim at pnl.gov>
To: "Robert G. Brown" <rgb at phy.duke.edu>
Cc: <Eugene.Leitl at lrz.uni-muenchen.de>; "Scott Shealy"
<sshealy at asgnet.psc.sc.edu>; <beowulf at beowulf.org>
Sent: Wednesday, April 04, 2001 2:19 PM
Subject: Re: What do you guys thing about the P4


> On Wed, 4 Apr 2001, Robert G. Brown wrote:
>
> > I have nothing for the P4, and at their current prices am not likely to
> > in the short run, although I confess that I'm a slut and if somebody
> > wanted to give me one I'd be more than happy to run it through its paces
> > and report on it.
>
> Well if all you are doing is inverting (big) matricies all day long.. Then
> you just have to buy a P4. :)
>
> Here is my favorite benchmark.. How fast can octave invert my random
> 2000x2000 matrix. Non-blas plain vanilla octave. Same RPM on both
> machines. Same kernel.
>
>
> time octave < test.m
>
> ----test.m-----
> inv(rand(2000));
> quit
> ----
>
> 1.5G P4 with 512 PC800 RDRAM
>
> real    1m11.479s
>
> vs. a dual P3/1G with 512M PC133 SDRAM
>
> real    4m13.391s
>
> I mean there is really no comparison. The P4 with a 50% faster CPU beats
> the PIII by a factor of almost 4. Pretty much in line with the standard
> stream numbers.
>
> Tim
> ---
> Tim Carlson
> Voice: (509) 375-5978
> Email: Tim.Carlson at pnl.gov
>
>
>
>
>
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
>