Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!snorkelwacker!usc!ucsd!helios.ee.lbl.gov!nosc!crash!pnet01!uzun
From: uzun@pnet01.cts.com (Roger Uzun)
Newsgroups: comp.sys.amiga
Subject: Real World Benchmarks
Message-ID: <3902@crash.cts.com>
Date: 11 Aug 90 18:56:05 GMT
Sender: root@crash.cts.com
Organization: People-Net [pnet01], El Cajon CA
Lines: 29
X-Local-Date: 11 Aug 90 11:56:05 PDT

A couple of more real world benchmarks for the Amiga 3000, compiled
with Lattice C 5.10 using inline 882 code 030 option on, longword
alignment and forcing rounding of 882 to extended precision, nearest 
(this is faster than the default double prec rounding)

Linpacks (double precision) .214 MFlops
Livermore Loops (dbl prec)  .232 MFlops

Looking at the BEST performing 25 Mhz 387 machine under SCO Xenix/Unix
from June 1990 Personal Workstation, page 62

Linpack = .239 and Livermore loops = .250
So the BEST 25 Mhz 387 performs virtually identically to the Amiga 3000
in double precision fp tests.

I have written a small disk test using unix open(), read(), write(),
but results out the Amiga 3000 SO FAR in front of the DOS machine
with a 40M IDE Conner drive, I feel there must be a problem with the
Turbo-C read() write() library calls.  I mean is my Amiga 3000 really
over 25 times faster than a 20 Mhz DOS machine?  It feels about that
much faster and the benchmark indicates that it is, but I hesitate
to post anything here until I can get a better C compiler for the
MS-DOS machine.

-Roger

UUCP: {hplabs!hp-sdd ucsd nosc}!crash!pnet01!uzun
ARPA: crash!pnet01!uzun@nosc.mil
INET: uzun@pnet01.cts.com