Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!tut.cis.ohio-state.edu!zaphod.mps.ohio-state.edu!mips!winchester!mash
From: mash@mips.COM (John Mashey)
Newsgroups: comp.arch
Subject: Re: Benchmarking minisupers...
Keywords: Alliant, Stardent Titan, IBM RS6000
Message-ID: <39283@mips.mips.COM>
Date: 9 Jun 90 22:34:15 GMT
References: <174@adam.adelaide.edu.au>
Sender: news@mips.COM
Reply-To: mash@mips.COM (John Mashey)
Organization: Your Organization Goes Here
Lines: 36

In article <174@adam.adelaide.edu.au> jsyktus@adam.ua.oz (Jozef Syktus) writes:
>I have been  following discussion about new 'killer micros' and I wonder if
>anybody have any data to compare new Stardent 3000 four CPU mini with 33-ns
>IBM RISC system 6000 (e.g. server 540) performance ??? Also, how new Alliant
>FX/2800 (minimum system, i.e. 8 processor) might compare to above.

There is the SPEC data collected so far, including the 32Mhz Stardent,
and the 30Mhz (avail late this year) IBM 540, and just for fun, a MIPS 6280.
Following shows the 10 benchmark, makring the integer ones with I,
followed by the geo-means of the Integer subset, the FP subset, and the
overall SPECmark.  For each machine, the benchmarks where it is fastest
are underlined.

1-I  8-I  13   15   20   22-I 23-I 30   42   47   I-m  F-m  SPEC
44.5 43.7 37.7 38.3 39.6 44.9 35.6 52.4 52.2 36.8 42.0 42.3 42.2 MIPS 6280,60Mhz
---- ---- ---- ----      ---- ----
21.0 24.9 33.2 33.1 43.4 23.7 26.7 26.5 65.8 91.0 24.0 44.3 34.7 IBM 6000/540
                                        ---- ----
17.8 20.3 14.7 19.7 62.9 18.1 18.2108.5 29.1 61.9 18.6 39.1 29.0 Stardent 3010
                    ----          -----

One of my standard foils uses a SPEC-style graph of these three together
to illustrate why it is so important to 1) use your own benchmarks, or 2)
When looking at the SPEC stuff, be very careful to avoid generalizing too
much from the means, especially on floating point.
(This is why SPEC insists on publication of all 10 numbers....)
In particular, although the set of the 4 integer numbers appears to be a
decent predictor of consistent relative performance, let me pick the
which FP benchmark, and I can prove ANY ordering of these 3 machines that
I want :-)  Note that the 3 machines are, in order: a scalar RISC with
a fast clock rate, a super-scalar, and a vector machine. 
-- 
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: 	 mash@mips.com OR {ames,decwrl,prls,pyramid}!mips!mash 
DDD:  	408-524-7015, 524-8253 or (main number) 408-720-1700
USPS: 	MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086