Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!purdue!ames!attctc!chasm
From: chasm@attctc.Dallas.TX.US (Charles Marslett)
Newsgroups: comp.sys.ibm.pc
Subject: Re: V20 vs. 8088: Is there an
Summary: New coprocessors are from IIT, not Witek
Message-ID: <8961@attctc.Dallas.TX.US>
Date: 12 Aug 89 05:40:20 GMT
References: <89080907301922@masnet.uucp>
Organization: The Unix(R) Connection, Dallas, Texas
Lines: 36

In article <89080907301922@masnet.uucp>, william.richardson@canremote.uucp (WILLIAM RICHARDSON) writes:
> . . .  Has anyone noticed the new 80287/80387 from
> Witec? They claim 200% increase with the same clock rate on the 80287.

I don't know about such from Weitek, but a startup company (the founders are
from Weitek, if I remember my rumors correctly) called IIT has started makeing
287 and 387 clones.  I got one!  And it is fast -- one funny, though, is that
...

> Fun note:  most ATs' run their Co-Processer at 2/3 the system speed!
> That can be fixed though, I'm running mine & 16Mhz on a 12Mhz AT !

As you mentioned, a signal can be used to divide the incoming clock (usually
twice the CPU clock or equal to the CPU clock) by 3 before using it in the
287 -- the IIT chip divides the clock by 2 rather than by 3, so you get an
additional speedup if the box runs with the divide-by-3 enabled.

I wrote a benchmark suite (that includes one demo written by IIT programmers
and it indicates you can get anywhere from a few percent improvement to
about 4x depending on the type of code (if you load the FPU and bang away --
like my fractal demo -- it runs away with you, but if you just do the usual
plodding C code -- load, add, store sequences -- then you'll get something
more like a 10 or 20% improvement over a similar speed 287).  Most programs
are somewhere in between.

> ---
>  ~ EZ-Reader 1.20 ~  -ex
> 
>  * QNet 1.03a3: CanConfMail: . HQBBS / ~ Toronto ~ Canada ~ 416y566y1765


===========================================================================
Charles Marslett
STB Systems, Inc.  <== Apply all standard disclaimers
Wordmark Systems   <== No disclaimers required -- that's just me
chasm@attctc.dallas.tx.us