Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!apple!mips!mash
From: mash@mips.COM (John Mashey)
Newsgroups: comp.arch
Subject: Re: VLIW Architecture
Keywords: VLIW
Message-ID: <29400@winchester.mips.COM>
Date: 13 Oct 89 08:59:58 GMT
References: <251FCB3F.12366@maccs.dcss.mcmaster.ca> <1050@m3.mfci.UUCP> <13050@pur-ee.UUCP> <1630@l.cc.purdue.edu> <1989Oct5.025841.2046@esegue.segue.boston.ma.us> <3449@alliant.Alliant.COM> <1071@m3.mfci.UUCP> <3460@alliant.Alliant.COM> <35596@lll-winken.LLNL.GOV>
Reply-To: mash@mips.COM (John Mashey)
Organization: MIPS Computer Systems, Inc.
Lines: 41

In article <35596@lll-winken.LLNL.GOV> brooks@maddog.llnl.gov (Eugene Brooks) writes:
....
>There is no defense against the ATTACK OF THE KILLER MICROS!

Well, even if I agree with the fundamental point, this might be a teeny bit
strong :-)

1) Supers and mini-supers and mainframes still do things that micros
don't yet, like huge memories, and very fast I/O, or very high
connectivity, or very high bandwidths in more places.  Those things
cost money.  One can hope that having fast cheap micros will induce people to
build more fast cheap peripheral chips to help some parts of the problems.
Some things will always cost money, no matter what.

2) It's hard to believe that there will not always be a niche for:
	a) The fastest box, at almost any cost
	b) Very fast boxes with good cost/performance

3) The issue raised really is:
	How big are those niches?  How many companies can thrive in them?

4) OPINION: 1-2 companies each in both a) and b).
	This is a high-stakes game, and the ante to play keeps rising.

5) OBSERVATION: RISC micros are munching away at the lower levels of the
minisuper business; going up (i.e., CONVEX) was a good strategy.
Trying to fight it out with them where they can do the job, is like the
fight that huge horde of 1970s mini- makers had with CISC micros:
once the latter got to be reasonably competitive on performance,
the former just got run over, except for the very biggest players.

6) An interesting paper is: "SUPERCOMPUTING, Myth & Reality", by
George J. Luste, of the Physics Department at U. Toronto.
(I think this was in Supercomputing Symposium '89). This was a nice
intro to some of the issues of vector code versus scalar code,
and cost/performance issues of supercomputers versus RISC micros.
-- 
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: 	{ames,decwrl,prls,pyramid}!mips!mash  OR  mash@mips.com
DDD:  	408-991-0253 or 408-720-1700, x253
USPS: 	MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086