Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!elroy.jpl.nasa.gov!decwrl!world!iecc!johnl
From: johnl@iecc.cambridge.ma.us (John R. Levine)
Newsgroups: comp.sys.ibm.pc.hardware
Subject: Re: Comparing CPU performance 386 / 486 -> use SI
Message-ID: <1991Apr10.182818.22123@iecc.cambridge.ma.us>
Date: 10 Apr 91 18:28:18 GMT
References: <stolk@fwi.uva.nl> <1991Apr10.105542.12938@fwi.uva.nl>
Organization: I.E.C.C.
Lines: 19

In article <1991Apr10.105542.12938@fwi.uva.nl> stolk@fwi.uva.nl (Jacco de Leeuw) writes:
>Norton SI is quite representable for the average DOS application, so
>I think the rates are quite relevant.

That's simply untrue.  SI times a small set of instructions, and
overemphasizes the speed of multiplication.  SI will give you a fair
comparison of two chips with the same internal architecture, e.g. 386
chips running at different speeds, but can be quite misleading when used
to compare different chips.

As has been noted elsewhere, the only real benchmark is your own application.
I find it useful as a rule of thumb that a 486 accomplishes 2 to 3 times as
much work per clock cycle as a 386, but that can get swamped by differences
in memory architecture (cache size, particularly) and peripherals.

-- 
John R. Levine, IECC, POB 349, Cambridge MA 02238, +1 617 864 9650
johnl@iecc.cambridge.ma.us, {ima|spdcc|world}!iecc!johnl
Cheap oil is an oxymoron.