Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!cs.utexas.edu!sun-barr!newstop!texsun!digi!jgay
From: jgay@digi.lonestar.org (john gay)
Newsgroups: comp.sys.ibm.pc.hardware
Subject: Re: 80387 25MHz & 33MHz Questions (Was : 80387 25MHz question)
Message-ID: <1991Apr24.165319.3600@digi.lonestar.org>
Date: 24 Apr 91 16:53:19 GMT
References: <1991Apr23.081807.27116@cs.mcgill.ca>
Organization: none
Lines: 25

From article <1991Apr23.081807.27116@cs.mcgill.ca>, by tinyguy@cs.mcgill.ca (Yeo-Hoon BAE):
> In article <1991Apr18.211744.12564@vuse.vanderbilt.edu> jsims@vuse.vanderbilt.edu (J. Robert Sims) writes:

>>I have seen comparisons (PC Magazine?) that indicated that the Weitek's 
>>performance was quite exaggerated.  It was only blindingly fast with
>>certain special optimized benchmark programs, and in real applications
>>was only a little bit faster.  The only league difference is price.

>>Rob


> Isn't some of the Sun Sparcs use Weitek as their Math Chip? It's
> rated figure is about twice of that of 486 at same clock speed...
> (I'm talking about SPECmarks for floats)
> So it _should_ have some significant improvement over 387...



I think that most of the speed gained by the Weitek coprocessor is because
it is memory mapped.  The new Cyrix memory mapped '387 is supposed to be
just as fast as the Weitek (or very close) and can be used as either
memory mapped or interrupt driven.  There is an article about 386 coprocessors
in the May '91 Dr. Dobb's.  Give it a read.

john gay.