Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!uwm.edu!bionet!agate!web-4h.berkeley.edu!c60b-1eq
From: c60b-1eq@web-4h.berkeley.edu (Noam Mendelson)
Newsgroups: comp.sys.ibm.pc.hardware
Subject: Re: Difference between 386/33 & 486/25 not counting fp
Message-ID: <1991Apr12.055944.14472@agate.berkeley.edu>
Date: 12 Apr 91 05:59:44 GMT
References: <1163@gistdev.gist.com> <1991Apr9.085749.4568@agate.berkeley.edu> <1164@gistdev.gist.com>
Sender: usenet@agate.berkeley.edu (USENET Administrator)
Organization: University of California, Berkeley
Lines: 60

In article <1164@gistdev.gist.com> flint@gistdev.gist.com (Flint Pellett) writes:
>>>>In article <Qbzekoy00Uzx42w1Fh@andrew.cmu.edu> dd2x+@andrew.cmu.edu (David Eugene Dwiggins) asked:
>>>>>Is there a significant difference in speed between a cached 33 Mhz system
>>>>>and a 486/25 system not counting floating point performance?
>I responded:
>>>The 486 is a lot faster at everything: generally >2 times as fast, at the same
>>>clock speed.  That means a 25 MHz 486 runs about half again faster than a 33
>>>MHz 386.
>>c60b-1eq@web-1e.berkeley.edu (Noam Mendelson) then stated:
>>NOT AS FAR AS CPU PERFORMANCE IS CONCERNED.  Given the same clock speed,
>>the 486 runs roughly 50% faster.  What you are talking about is a complete
>>system's performance.  This depends on memory, disks, etc., as well as
>>the CPU.  If you perform a CPU benchmark, please post the results.
>My response, as polite as I can make it:
>Mr. Mendelson doesn't know what he's talking about: his 50% figure is
>just plain wrong.  Ok, I will post a benchmark, but not mine, this is
>from Personal Workstation magazine.  (Excerpted without permission: I
>doubt they will mind.  If you want to see the full details about the
>benchmark, I suggest you acquire any copy of this excellent mag, as
>they post benchmark results every month.)  This excerpt shows only one
>category (DOS) and only 33 MHz machines, and only the Dhrystone tests
>(they publish results of 5 different tests in 4 different OS'es and
>about 4 different price categories.  Performance ratios between 386
>and 486 machines seem to follow the same ratios as the ones below.) 
>DOS 386 Systems:	Dhrystone	Price
>Micro Express /33	15,870		$4,998
>Laser Ditigal 386/33	15,750		$4,296
>Arche Legacy 386/33	17,134		$8,665
>Northgate 386/33	17,131		$8,919
>
>AST 486/33		34,192		$4,490 (upgrade board)
>Club American Hawk III	35,923		$6,495
>NCR PC486/MC 33		35,234		$14,995

My response, as polite as I can make it:  You quoted only 4 386/33 systems
and 3 486/33 systems.  I would like to see the entire survey.  You also failed
to specify the size of the cache (or lack thereof) on each of the systems.
You also failed to provide additional benchmarks on said systems to
support your claim.
If you examine the article in PC Magazine, you'll find that they also
performed additional benchmarks on their systems, but since the results
came out to be identical to those yielded by the instruction mix, they
did not bother to print them.

> . . .  I'd be interested to
>hear what you think they are, assuming you have some concrete
>information, not misinformation.  (Compiler performance can affect
>Dhrystone results, but on 386 and 486 machines they are using the
>same compilers, so that isn't it.)

Provided you specify the issue, I'll check out the article in Personal
Workstation magazine.  Also, I will collect more benchmark data and
post it, since you obviously doubt the PC Magazine article.
I would advise readers of this newsgroup to do the same.

-- 
+==========================================================================+
| Noam Mendelson   ..!agate!ucbvax!web!c60b-1eq | "I haven't lost my mind, |
| c60b-1eq@web.Berkeley.EDU                     |  it's backed up on tape  |
| University of California at Berkeley          |  somewhere."             |