[Starkit] Re: Extraneous benchmark between Red Hat Lynux andUnix

[Arjen Markus]

"Techentin, Robert W." wrote:
> 
> Arjen Markus wrote:
> >
> > "Eduardo Rodriguez (Gerencia de Desarrollo)" wrote:
> > >  P4:
> > >   256 MB ram
> > >   1 Intel(R) Celeron(R) CPU 1.70GHz
> > > Alpha Server:
> > >   1 Gb ram
> > >   2 alpha processors at 600 MHz each one.
> >
> > But the raw CPU speeds differ by a factor of 3 already - isn't that
> > about the difference you saw? I'd say: look no further ;)
> 
> Not necessarily.  Intel and Alpha CPUs are very different, and you can't
> compare MHz/GHz clock speeds in any meaningful way.
> 
> If you look at CPU benchmarks at http://www.spec.org, you'll see a Compaq
> AlphaServer DS10 6/600 with a SPECint of 364.  Nobody bothers benchmarking
> Celerons for CPU performance (because they're slow) but SPECint for "real"
> P4 1.7 GHz CPUs runs around 590.  That's a difference of 1.6X.
> 
> And Celerons are slow when compared to the Pentium.  The alpha servers have
> lots of multilevel cache for just this sort of application.  I wouldn't be
> surprised if the Alpha outperforms the Celeron in raw CPU power.  So I think
> looking further is reasonable.
> 
>

Okay, I was unaware of these details ;). But if they are so important,
then
any conclusions drawn have to be confined to the particular class of
programs
used in the benchmark ....

Regards,

Arjen