Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!news.cs.indiana.edu!purdue!krk
From: krk@cs.purdue.EDU (Kevin Kuehl)
Newsgroups: comp.sys.mac.misc
Subject: Re: The Fate of the Macintosh
Message-ID: <14020@medusa.cs.purdue.edu>
Date: 23 Mar 91 16:52:45 GMT
References: <1991Mar22.154811.8691@rucs2.sunlab.cs.runet.edu>
	<YbucwQ600iV080lo0x@andrew.cmu.edu>
	<1991Mar23.055725.27761@ux1.cso.uiuc.edu>
	<1991Mar23.064856.4877@cs.ucla.edu>
Sender: news@cs.purdue.EDU
Organization: Purdue Univserity Computer Science Department
Lines: 47

In article <1991Mar23.064856.4877@cs.ucla.edu> lange@lanai.cs.ucla.edu (Trent Lange) writes:
   You're right, of course, but only to a point.  On the same kind of
   machines, running the same software, MIPS *is* a fairly reliable
   measure of performance between models of that machine.  The fact
   that a Mac Classic is about 0.8 MIPS and a Mac IIfx is about 6 or 7
   MIPS tells you something about their relative performance, does it not?

Sure it does, but so do a lot of other things -- like price.  And not
too many people are trying to decide between a Classic and a IIfx.  A
better example is a IIsi and an SE/30.  And even then MIPS fool you
because you would think that a IIsi is always faster than an SE/30,
but timing comparisons posted to the net show that ain't true.

   And, on straight number-crunching and/or programming tasks, the MIPS
   ratings of machines, even across vendors and machines, does give a
   decent ballpark estimate of relative performance.

I agree for number-crunching, but not for programming tasks.  When you
say programming tasks, I assume you are talking about such things as
editing, compiling, debugging, etc.  In these cases brute CPU power is
quite meaningless because the speed of your I/O devices such as the
keyboard, disks and memory are the true limits to speed.  A good way
to find this out is to use a timing program that counts the real time
in a kernel, on the cpu and the total time taken.  From my estimates
on Unix machines, a compiler spends about 2-3 seconds blocked for I/O
for every 1 second it spends on the CPU.  And editing and debugging
are going to be in the range 10-1000:1 for the most part.

   But this is why MIPS ratings *are* somewhat useful when comparing
   Macs and NeXTs, since the NeXTs *do* have a graphical interface that
   is arguably at least as nice as the Mac's, and are comparable in many
   other ways (even philosophically).  With those things the same, raw
   performance definitely becomes an issue again for some people -- which
   is where the MIPS come in.

But what you forget is that the Mac and the NeXT have two completely
different operating systems, buses and display systems.  The NeXT is
going to seem somewhat slower than the MIPS figures would lead you to
believe because of the overhead in displaying on such a large screen
with PostScript.  Also it has MACH which will provide more of an
overhead than the MacOS will because MACH provides a lot more services
than the MacOS does.  They are somewhat useful, but only realizing
that a NeXT is going to be faster than a IIfx.
-- 
Kevin Kuehl
krk@cs.purdue.edu
kuehlkr@mentor.cc.purude.edu