Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!wuarchive!zaphod.mps.ohio-state.edu!sdd.hp.com!ucsd!helios.ee.lbl.gov!nosc!halibut.nosc.mil!koziarz
From: koziarz@halibut.nosc.mil (Walter A. Koziarz)
Newsgroups: comp.sys.ibm.pc.hardware
Subject: Re: PC Brand 386 Machines
Keywords: Cache size
Message-ID: <2650@nosc.NOSC.MIL>
Date: 12 Aug 90 21:28:03 GMT
References: <7557@ucdavis.ucdavis.edu>
Sender: nobody@nosc.NOSC.MIL
Reply-To: koziarz@halibut.nosc.mil.UUCP (Walter A. Koziarz)
Organization: Naval Ocean Systems Center, San Diego
Lines: 13
X-Local-Date: 12 Aug 90 14:28:03 PDT

In article <7557@ucdavis.ucdavis.edu> lasky@iris.Berkeley.EDU (Ty Lasky) writes:
>
>Finally, on a loosely related topic, would a 386, 25MHz with 387 be
>faster than a 386, 33MHz w/o 387?  For background, I do the usual word
>
For pure floating point math -- the answer is an unqualified YES.  As one data
point I provide the following -- My 10MHz, 0 Wait-state, Zenith Z-100  (8088
plus 8087; and both run at 10MHz, btw). is almost 4X _FASTER_ in floating point
math than a 25 MHz 80386dx machine sans 80387.  This test used THE SAME
EXECUTABLE for both systems, obviously an executable *using* '386-code (rather
than 8086-code) may alter the results.

Walt K.