Fastest Intel Processors

W Bauske wsb at paralleldata.com
Fri Jan 11 15:56:06 PST 2002


One could define it as SPECFP2000. AMD might not like that though cause 
it would show that P4's are faster (by virtue of their faster clock). 
It does show that XP's are about equal to a P4 clock for clock, which is 
something AMD seems to want to deny. Compare the XP1900 vs P4 1.6Ghz, 634 vs 637.

I think AMD would prefer SPECINT2000. There XP1900 vs 1.6Ghz P4 is 701 vs 565.

So, that implies if you find Athlons out-perform a P4 on your code, it's
probably has a large integer component. (assuming a good compiler for both)

Wes

Marc Cozzi wrote:
> 
> I'm sure most people here understand the relationships between architectures
> and clocks
> and not to move the subject but isn't AMD moving industry people to adapt
> a new measurement for the performance or relative performance of newer
> chips?
> Think I read the proposal was to get as far from the MHZ/GHZ thing as
> possible.
> 
> Hell, even Steven from Dell knows MHz from GHz.
> 
>    marc
> 
> -----Original Message-----
> From: W Bauske [mailto:wsb at paralleldata.com]
> Sent: Friday, January 11, 2002 2:38 PM
> To: beowulf at beowulf.org
> Cc: beowulf at beowulf.org
> Subject: Re: Fastest Intel Processors
> 
> Mhz/Ghz matter. So does IPC (Instructions per clock). Word width
> determines which architecture you must have to run on.
> 
> Consider that the IBM Power4 chip runs at 1.3Ghz yet it has the
> fastest SPECFP2000 of any machine on the list(1169). IBM Power chips
> have always used substantial on chip parallelism. A 2.2Ghz P4 is
> 766 for SPECFP2000.
> 
> Keep in mind that the way to measure how well a processor works is
> to use your own codes and decide which one does it best for the $$$$
> spent.
> 
> Wes
> 
> "C.Clary" wrote:
> >
> > Hi all,
> > I really think this Ghz. goobly-gook is way over rated.  I think, if you
> > look more closley at the problem, the key to superior proformance is in
> the
> > found in the actual construction or architecture of the processor.  The
> > Itanium, for example only runs at 733 Mghz but it has (on die)   2meg of
> > cashe on the L-2... The Alpha, in its previous incarnation ran at only 233
> > but did 64 bit processing... Quite amazing actually.
> >
> > A friend wrote me that:
> > "Red hat is working closely with Compaq to solidify its OS to the Alpha
> > architecture"... I don't know for sure, first hand... But it is, of
> course,
> > the next logical step, since FreeBSD, already provides a clean UNIX 64 bit
> > processing code for the Alpha.
> > Chip
> >
> > -----Original Message-----
> > From: beowulf-admin at beowulf.org [mailto:beowulf-admin at beowulf.org]On
> > Behalf Of Craig Tierney
> > Sent: Friday, January 11, 2002 11:13 AM
> > To: Rob Simac
> > Cc: beowulf at beowulf.org
> > Subject: Re: Fastest Intel Processors
> >
> > The 2.2 Ghz cpus have just been released.  This generation
> > of cpu is built at 0.13 microns and has twice as big of a
> > L2 cache (512 KB).   These are the single processor versions.
> > The Xeon (smp) chips should follow shortly (I am guessing).
> > Go look at the roadmaps at www.theregister.co.uk.  They tend to
> > be accurate.
> >
> > I have not heard that there is a problem with the 2.0 Ghz.  Is
> > it a problem with RedHat or the Linux kernel specifically?  We
> > had no problems with the 1.7 Ghz Xeon chips, but they are not the
> > 2.0 Ghz that you are talking about.
> >
> > Craig



More information about the Beowulf mailing list