Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!husc6!necntc!necis!encore!fay
From: fay@encore.UUCP (Peter Fay)
Newsgroups: comp.arch
Subject: Re: chewing up mips with graphics
Message-ID: <1677@encore.UUCP>
Date: Fri, 19-Jun-87 16:43:57 EDT
Article-I.D.: encore.1677
Posted: Fri Jun 19 16:43:57 1987
Date-Received: Wed, 24-Jun-87 06:53:15 EDT
References: <8270@amdahl.amdahl.com> <359@rocky2.UUCP> <6240@steinmetz.steinmetz.UUCP>
Reply-To: fay@encore.UUCP (Peter Fay)
Organization: Encore Computer Corp., Marlboro, MA
Lines: 39

In article <6240@steinmetz.steinmetz.UUCP> OCONNORDM@ge-crd.com@ARPA (Dennis Oconnor) writes:
>
>Enuff on how graphics will chew MIPS. It will, but not like SIMULATION !
>A revolution is occurring today in engineering, where more and more
>computers, chemicals, parts and systems are thoroughly SIMULATED before
>even a first prototype is built. On the horizon are even MORE complex
>simulations, like the effect of drugs on biological systems, and
>accurate faster-than-life atmospheric simulation. Even really simple
>simulations ( like 100k-transistor CPUs ) eat MIPS. Simulation will
>gobble up parrallism as well - what a bargain !
>
>When it comes to producing new products in competive arenas, simulation
>has been shown to be faster and cheaper than traditional methods. The
>more MIPS you have, the more complex a system people will be trying
>to simulate. So bring on the MIPS ! ( MFLOPS too, of course. )

I was wondering how long it would be before someone said this. There is no
question in my mind that this will become a very huge market. Imagine
simulating cache coherency of a 1000-MIP machine with 128 cpus and
hierarchical caches. And that's only what we're doing today, not a few
years from now.

There's no question that this current work would be impossible without
nice machines like our 35-MIPS multiprocessor. (Now we can run the Mach
operating system - picture one Mach simulation task running with 2000
light-weight threads of control instead of one Unix job creating 2000
processes.)

Of course it would be nice to have 30 times our current power, which is
why we're designing it.



				peter fay
				research
				encore computer

{allegra|decvax|ihnp4|linus|princeton|pur-ee|talcott}!encore!fay
				fay@multimax.arpa