At 16:05 20/04/97 -0400, you wrote:
>Robin Whittle wrote:
>> 
>> 
>> I can't compare its performance to a Pentium, but it runs rings
>> around the 486 100 MHz which I have been using for over a year.
>> It is between 5.6 and 6 times the speed!
>> 
>
>As a side note, there seems to be a problem with the Pentium Pro cpu. A
>Linux benchmark called Oscillates,  which simply tries to send as many
>wavetable-lookup oscillators as it can to the dsp, returns a score of 42
>oscillators on a P200 w/ 32Meg, 36 oscilators on a P166 MMX, and only
>28 on a PPro 200 w/ 32Meg and 27 on a dual PPro 180 w/ 32Meg.
>
>Any thoughts? The code for the benchmark by Dave Topper is included.

As I understand it this comes down to cache management and the efficiency
of the branch prediction. And we already know that in a Windows environment
the PPro is worse at legacy 16-bit code than a Pentium.

I once came across a case where someone was *convinced* a Pentium was
slower than a 486, because it ran their own home-brew bench mark slower. It
turned out that this rather special test was tripping up the Pentium's
branch prediction scheme. This is usually more efficient when dealing with
'real' code, but incurs a bigger time penalty than the 486 when it gets it
wrong. 

As a result the 486 was actually faster in this one particular accidentally
contrived situation. Of course in real life, with real apps, this would
never (well - hardly ever) happen.

Aside from this - were these machines identical in other ways?
Particularly, was the memory the same type and speed? SDRAM, EDO and
standard RAM all make a difference. Some PCs run with 60nS and others 70nS
RAM, and real speed freaks can get 50nS. 

Also - how closely can you map Oscillates to a Csound run? Csound seems
like a more general and wide-ranging workout for a processor. I'd guess
that the only way to see how well a processor does is to try it out.
There's only a 'problem' if it's a problem for real apps. 

In any case,  I'd be interested to see how well the K6 does with this
benchmark just as a rough comparison. Caching,  branch prediction and
instruction decoding are all supposed to be improved over the Pentium and
PPro. 

I'll be getting a K6 in the next month or three, and it will be fun to
compare Robin's timings with that and my current P133.

R.