Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!asuvax!ncar!elroy.jpl.nasa.gov!swrinde!zaphod.mps.ohio-state.edu!think.com!hsdndev!cmcl2!kramden.acf.nyu.edu!brnstnd
From: brnstnd@kramden.acf.nyu.edu (Dan Bernstein)
Newsgroups: comp.arch
Subject: Re: Novice question:  measuring speed
Message-ID: <3516:Mar1319:50:3291@kramden.acf.nyu.edu>
Date: 13 Mar 91 19:50:32 GMT
References: <645@ssdc?>
Organization: IR
Lines: 26

In article <645@ssdc?> jbasara@ssdc (jim basara) writes:
> Could someone please
> provide me with a descriptive explaination of why MIP ratings are not a 
> good way of comparing processing speed for RISC machines as opposed to MFLOPS?

I've got this Turing machine that runs at 5000 MIPS. That's right, it
can move the tape back and forth 5 *billion* times a second. Impressed?
You shouldn't be: it takes so many instructions to get something done on
a Turing machine that the MIPS measurement is pointless.

In contrast, MFLOPS measure some (supposedly) real amount of work
getting done. The number of floating-point operations in a typical
computation is relatively independent of the machine at hand. Of course,
MFLOPS don't tell you whether floating-point divisions are ridiculously
slow, and they don't tell you how non-floating-point computations will
run, but they're at least a bit more solid than MIPS.

> I would also like to know why companies are using MIP ratings for those machine
> s if they are not accurate.

If you were trying to make a machine look good, you'd take the most
favorable measurement you could get, right?

(I don't actually have a 5000-MIP Turing machine, btw.)

---Dan