Re: [OFFTOPIC] Re: Pentium II Math Bug

Rogier Wolff (R.E.Wolff@BitWizard.nl)
Tue, 6 May 1997 13:06:50 +0200 (MET DST)


Juhani Rautiainen wrote:
> > - and don't go see the page about comparison between Pentium-II and K6
> > (and others) because the guy has sufficiently little knowledge in processor
> > to think that processors have to be compared at similar frequencies,
> > As if you'd an Alpha 21164 to 100Mhz in order to compare it with some POWER2
> > processor. Worse yet: the 486DX2-50 would end up being faster than the
> > 486DX-50 (since they'd be both compared at a speed of 25Mhz which the 486DX2
> > would double).
>
> Another example of how well you can read. I'm not very good at processor
> technology but those technical documents that Collins (that guy) has on
> his site seem to tell that he knows very much about x86 technology. Maybe

Ehmm His knowlege about Intel processors doesn't mean he knows how to
perform a benchmark.

I could now plug a 200MHz Pentium/MMX into my motherboard. Now
suppose I made a new processor that runs at 1GHz, and fits into the
same socket. At 1000MHz it performs twice as fast as a Pentium/MMx.
Collins suggests that I should benchmark this processor at 200MHz,
and conclude that it runs about 2.5 times slower than a Pentium?

Real life performance comes from how much a processor can do every
clock cycle, and how fast a clock it can run. My hypothetical
processor above could do very little in every clock, but makes up very
well by running at a high clock speed. If Intel could pump their P200MMX
to 1GHz in a few weeks, then collins would be right.

The Pentium II seems to be a few percent slower than a Pentium Pro at
the same clock rate. However it can run at a higher clock rate to
make up the difference.

Roger.