Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!sdd.hp.com!ucsd!ucselx!crash!pnet01!uzun
From: uzun@pnet01.cts.com (Roger Uzun)
Newsgroups: comp.sys.amiga
Subject: Re: Real System Comparisons
Message-ID: <3892@crash.cts.com>
Date: 11 Aug 90 02:36:01 GMT
Sender: root@crash.cts.com
Organization: People-Net [pnet01], El Cajon CA
Lines: 42

If anyone is interested I use a MS DOS 3.3 based 20 Mhz PC clone
witrh cache and a 40MB IDE controller at work.  I also have an Amiag
3000/25-100 I use at home/work.  I use lattice C 5.10 on the Amiga and Turbo C
2.0 on the PC.  Trubo C is set to compile for 80286 code (no 386 option with
that compiler) using word boundaries optimized for speed.  The 386 has NO
match chip.  In day to day work and system development type tests,
subjectively, the 3000 seems MUCH faster.  Here are some benchmarks I have
run.  Please note that the 68000 instruction set is VERY close to the 030
instruction set, but the 286 instrction set is MUCH WORSE than the 80386 set. 
This skews this becnhmark for the amiga BUT all my DOS code is compiled for
the 286 or the 8086 (even worse) many of my amy applications, in fact most,
are compiled for the 68030/882 and even those compiled for the 68000 are still
close to the 030 in execution speed, so you can get MUCH more out of an amiga
running AmigaDOS than you can out of a PC running DOS.  In any case here are
some CPU benchmarks, I ran no fp benchmarks on the pc since it has no math
chip and would be DOG slow.
 
           Amiga 3000            20 Mhz 80386 cache
Qsort        4 secs                 8 secs
sieve(100 iters) 2 secs             4 secs
drystone 1.1   12,500               5,000
Drystone2      10,000               ?????
savage(25,000 iters) 2 secs
whetstones     1,666,667 whets/sec  ????

In any case looking through Personal Workstatin magaziine, using a highly opti
mized 386 compiler running under a 386 DOS extender, a 25Mhz Cache 386 system
gets 11,939 Drystone2's/sec vs the Amiga 3000's 10,000 this is within
measurement error, they are approx equal.  So when you can get a 386 native
mode application it will perform about the same on a 25mhz 386 as on the amiga
3000, when you geta 286 compiled application it will perform about half as
fast as on the 3000.  This is CPU dependent stuff (integer) only.

IN my disk tests the Amiga 3000 is MUCH faster, but I need someone to suggest
a quick and dirty C benchmark I can compile and run on each machine.  In
psuedo code would be fine.

-Roger

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