Xref: utzoo comp.sys.mac:46030 comp.protocols.appletalk:2891
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!rpi!nyser!rodan!isr
From: isr@rodan.acs.syr.edu ( ISR group account)
Newsgroups: comp.sys.mac,comp.protocols.appletalk
Subject: Re: MIP rating for the Mac II
Keywords: performance
Message-ID: <1712@rodan.acs.syr.edu>
Date: 11 Jan 90 16:14:48 GMT
References: <2785@mtuno.ATT.COM>
Reply-To: isr@rodan.acs.syr.edu (Michael S. Schechter - ISR group account)
Organization: Syracuse University, Syracuse, NY
Lines: 60

In article <2785@mtuno.ATT.COM> rjk@mtuno.ATT.COM (Roberto Kohler) writes:
>Does anyone out there know how the Mac II, Max IIx, and Mac II CX compare
>in performance with the 80386 based machines?  I'm trying to find out the 
>MIP rating of the Macs and how they compare with 386 based machines.
>Would the Mac IIx be equivalent to a 20 MHz 386 machine, etc.

I would reply via Email, but this is a VERY bad comparison. The idea
of the mythical "MIP" is an idead propagated by computer sales(wo)men.
A MIP is ONLY useful when comparing proccessors with the same instruction
set. For example, in my Mac, I have a 16mhz 68020/68881 which can put
out anywhere from 5 MIP->0.1 MIP depending on what it's doing. (These 
figures are off the top of my head, but I believe I'm pretty close that
a NOP takes 3 cycles from start to completion and theres a 68881 instruc
that takes about 60). Now, if a 386 machine was running at say, 20Mhz,
you could see how many cycles each of it's instructions take up and get
general MIP ratings for it's instructions.
  Now, for the big news:
NONE OF WHAT I JUST SAID MATTERS AT ALL!!!!
It doesn't matter how many MIPS my 68020 or your 80386 is doing; This
is because you have to compare what each CPU has to do in order to get
a job done. For example, if you have a very large array in memory,
(>64kbytes) on a 68020 it is very easy to access any portion of it at
any time, just address it. On a 80286, you have to go through arcane
and miserable playing with "segment registers". This  means you have
to move a segmentregister to the stack, move an address to the stack, move the
stack to segmentregister, access you data, and then move the stack
back to the segmentregsiter.
  Four instructions for the 80286 where the 68020 requires one.
There may be similar instances where the 80286 is better than the
68020. In any case, I'm just trying to show that comparing MIPS of
proccessors is like comparing Apples and Peanuts - It don't work.
 There are standard ways of comparing speeds of computers - 
Like running the same program - one that does a simulation of typical
tasks on both computers and timing it. But that only brings in the
question of was each version written to take advantage of each CPU's
particular (in)efficiencies?? And what about the quality of the
compiler output?? Perhaps you are comparing the best compiler for
CPU A to the worst for CPU B ???

THERE IS ONE TRUE TEST.
put both machines side by side, each of them running the program that
YOU WILL BE USING for most of your work, and have them do the same
task. Do this task from start to finish on both machines several
times, switching back and forth from one to the other between trials.
This will let you get used to both and give you a good comparison.
Whichever machine is faster for your task is the faster one.

Any other comparison is total drivel.
MIPS are for morons and DSPers

Mike Schechter
Institute for Sensory Research
isr@rodan.acs.syr.edu




-- 
Mike Schechter, Computer Engineer,Institute Sensory Research, Syracuse Univ.
InterNet: isr@rodan.acs.syr.edu   Bitnet: SENSORY@SUNRISE