Charles M. Hannum said:
> 
> Manuel Bouyer <bouyer@ensta.fr> writes:
> 
> > 
> > I have a cyrix 6x86-P150 (a 120 Mhz CPU) to test. I've run some private
> > benchamrks on this CPU, and a pentium 120. I found the p120 from 15 to
> > 50% fater in most of the case, and in all tests involving interger or
> > float calculations. According to cyrix, the 6x86 should have been faster.
> 
> Are you using the CPUs in the same machine?  If not, how big are the
> caches?  What kind of cache?  What speed are the memory and I/O busses
> running at?  How fast is the memory?  EDO, FPM?  There are a huge
> number of variables that could affect performance.
> 
> > At boot, NetBSD reports the CPU to be a "i486DX (486-class CPU)". Perhaps it
> > doesn't make use of pentium-specific stuffs ?
> 
> Did you build the kernel with `options I586_CPU'?  If not, it will be
> used like a 486.  For normal applications, this won't make any
> noticable difference.

Yes, it has options I586_CPU. With a Pentium, the cpu-type is correctly
probed.

Brian de Alwis (OTT) said:
> 
> I bought 'PC Magazine' this weekend since it had a feature on buying new PCs 
>  -- I wanted to refamiliarize myself with the new technology. They said that, 
> from their tests, the Cyrix 6x86 performed better than the Pentium on 16-bit 
> OSes, which includes Windows 95, but performed worse than the Pentium on 
> 32-bit OSes such as NT.
> 


Well, the cyrix was on an new Asus Teck P55T2P4 motherboard, as the pentium
is on an old Intel motherboard. The Asus is supposed to be faster but ...

I plugged the pentium on the Asus Teck, and re-did the tests
(with the same drive). This doesn't change significally the results.

I did'nt make any tests on 16bits apps (I don't have time to loose installing
windoze), but on 32 bits apps, the pentium is clearly better than the 6x86,
from 7%  for double precision computations, 25% for integer
computations, to 60% for computations involving arrays. The only points where
the cyrix is better is on a test using the stack (for a recurcive computation).

The 6x86P150+ should be called "6x86P100-" :)

--
Manuel Bouyer, Ecole Nationale Superieure de Techniques Avancees, Paris
email: bouyer@ensta.fr
--