Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxn!ihnp4!ima!johnl
From: johnl@ima.UUCP (John R. Levine)
Newsgroups: net.arch,net.micro.att
Subject: Re: AT&T MIPS claim
Message-ID: <124@ima.UUCP>
Date: Tue, 3-Jun-86 10:55:15 EDT
Article-I.D.: ima.124
Posted: Tue Jun  3 10:55:15 1986
Date-Received: Sat, 7-Jun-86 08:16:20 EDT
References: <577@scirtp.UUCP> <124@bakerst.UUCP> <583@scirtp.UUCP>
Reply-To: johnl@ima.UUCP (John R. Levine)
Distribution: net
Organization: Javelin Software Corporation
Lines: 29
Xref: watmath net.arch:3379 net.micro.att:1262
Summary: for single user performance, it's not so far fetched

It's certainly a lie to claim that you can put 75% of the load of a 780 on
a 68010 box.  It's probably not a lie to claim that you get 75% of the single
user performance.  A few years back, people I work with were benchmarking a
port of System III between a 286 box and a Vax 750.  The 286 box was one of
Intel's development systems that roughly resembled a PC AT except the I/O was
multibus rather than PC bus.  For the sorts of things that Unix users do, such
as compiling C programs and running documents through troff, the 286 was at
least as fast as the 750, and often close to a 780 in performance.  I think that
the comparison was quite fair, as both machines were running similar versions of
Sys III, so that most of the kernel code and almost all of the user code was
compiled from the same C sources.  Both compilers were Sys III PCC based.

The biggest difference these days between a micro and a big computer isn't CPU
power -- it's I/O bandwidth.  With one user on a computer, the CPU performance
is relatively more important than the I/O bandwidth.  As you load up more
users, I/O becomes more important, as you start paging and swapping, and the
larger computer's more sophisticated I/O becomes more useful.  If there's only
one process, it hardly matters if you can overlap I/O and computing.  If there
are a lot of processes, it's quite important.  The Vax's separate I/O and memory
buses pay off then.

Anyway, if you interpret their claim as meaning that having a Unix PC is 75%
as good as having your own Vax, it's still hype but it's a lot closer to
reality.
-- 
John R. Levine, Javelin Software Corp., Cambridge MA +1 617 494 1400
{ ihnp4 | decvax | cbosgd | harvard | yale }!ima!johnl, Levine@YALE.EDU
The opinions expressed herein are solely those of a 12-year-old hacker
who has broken into my account and not those of any person or organization.