Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!apple!sun-barr!rutgers!cbmvax!daveh
From: daveh@cbmvax.commodore.com (Dave Haynie)
Newsgroups: comp.sys.amiga.advocacy
Subject: Re: Commodore Research and Development.
Message-ID: <17616@cbmvax.commodore.com>
Date: 15 Jan 91 18:29:20 GMT
References: <187e4f65.ARN097c@easy.hiam> <1991Jan09.153108.17485@convex.com> <17561@cbmvax.commodore.com> <242^HH|@rpi.edu>
Reply-To: daveh@cbmvax.commodore.com (Dave Haynie)
Organization: Commodore, West Chester, PA
Lines: 42

In article <242^HH|@rpi.edu> peck@ral.rpi.edu (Joseph Peck) writes:
>In article <17561@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Haynie) writes:

>>Though maybe that helps explain RS6000 performance.  Its a real screamer if
>>you're a single user with one big task doing Workstation type floating point
>>operations.  

>Well, I don't know about the more users comment.  My dad's company does
>business applications using Informix, and with 20 or so users on a 
>base RS6000 it still seems awfully fast.....  even though lot's of 
>I/O is going on. 

I don't think I/O is really a problem.  The RS6000s have the double-speed 
MCA bus (peaks at 40 MB/s), as well as a 20 MB/s "optical bus".  That's 
generally plenty of I/O to support a number of users.  The problem seems to
be that lots of task switching can bring this machine to its knees.  I don't
use RS6000s myself, but like for any new CPU architecture, I've been 
watching this one.  In various reviews (the January Personal Workstation,
and a UNIX Review a few months back), they find these systems have better
than [workstation] average hard disk performance, excellent single user math
and CAD performance, reasonable but not amazing [workstation class] graphics
performance, and rather moderate performance under a heavy task load (like
the AIM benchmark series).  

>I don't know what would happen if you had all 20 people running Matlab 
>though.... (but then again, even one person running Matlab tends to screw 
>our Sun4's at work.... :)

Actually, the same kind of thing seems to be true of Sun SPARC machines.  
Though the SPARCs seems to have kind of a plateau effect -- they drop off
linearly for CPU hog tasks 1..N, then all of a sudden take a nose dive.   I 
don't know if this is a Sun 4 implementation detail, or an expected effect of 
the SPARC architecture, though.

>Joe Peck


-- 
Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests"
   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
	"Don't worry, 'bout a thing. 'Cause every little thing, 
	 gonna be alright"		-Bob Marley