Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!unmvax!ogccse!cvedc!nosun!neighorn
From: neighorn@nosun.UUCP ( SE Sun/PDX)
Newsgroups: comp.sys.ibm.pc
Subject: Re: Norton Si
Message-ID: <408@nosun.UUCP>
Date: 31 Aug 89 21:00:43 GMT
References: <478@v7fs1.UUCP> <224@orchid.warwick.ac.uk> <1989Aug25.162653.25755@ziebmef.mef.org>
Reply-To: neighorn@nosun.UUCP (Steven C. Neighorn - SE Sun/PDX)
Organization: Sun Microsystems, Portland, Or.
Lines: 44

In article <1989Aug25.162653.25755@ziebmef.mef.org> stephen@ziebmef.mef.org (Stephen M. Dunn) writes:
>   Of course, there have been many complaints about the fact that Norton's
>SI seems skewed in favour of 286s (and higher) due to a (perhaps) inappropriate
>choice of benchmarking algorithm.  I don't know if any adjustments have been
>made or not.

"inappropriate" is an understatement. I shudder to think how many
computer salesmen and advertisements and jokers who don't know what they
are talking about quote that good old Norton SI number.

It is just less than completely useless!

The compute performance section of SI appears to check the multiply
instruction in a loop. Now, how many programs do you know of that do
nothing but multiply integers? For such programs, the SI number is
accurate, but for anything else???

Take a 16 MHz, 1 wait state 80386 machine. It will get a SI rating of
around 17. Now go and write your own little loop/mul program, and lo
and behold, it runs approx 17 times faster as well.

The oft-flamed Dhrystone benchmark is a magnitude better than this.

Mis-information is worse than no information at all, because people who
want answers will stop looking for them when they get *something*.

>   And, of course, any benchmark will only be an approximation based on what
>the author thinks a typical user's usage patterns will be.  If I use my
>hard drive more than you do on an identical machine, we'll get different
>performance levels even though SI won't find any difference because it uses
>some assumed weighting in computing the overall performance score.

And then there are really many "types" of hard disk use. Reading/Writing
different size blocks, sequential/random, etc. Most of the cheap and
dirty disk benchmarks only scratch the surface of what is possible to
test.

Of course the only real benchmark is the application(s) you really want
to run! :-)
-- 
Steven C. Neighorn           !tektronix!{psu-cs,nosun,ogccse}!qiclab!neighorn
Sun Microsystems, Inc.      "Where we DESIGN the Star Fighters that defend the
9900 SW Greenburg Road #240     frontier against Xur and the Ko-dan Armada"
Portland, Oregon 97223          work: (503) 684-9001 / home: (503) 641-3469