Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!uakari.primate.wisc.edu!gem.mps.ohio-state.edu!rpi!crdgw1!crdos1!davidsen
From: davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr)
Newsgroups: comp.arch
Subject: Re: Fed up with MIPS
Message-ID: <1322@crdos1.crd.ge.COM>
Date: 23 Oct 89 13:40:42 GMT
References: <76700077@p.cs.uiuc.edu> <318@ssp1.idca.tds.philips.nl>
Reply-To: davidsen@crdos1.UUCP (bill davidsen)
Organization: GE Corp R&D Center
Lines: 44

In article <318@ssp1.idca.tds.philips.nl>, roelof@idca.tds.PHILIPS.nl (R. Vuurboom) writes:
|                                                       Racing cars today
|     no longer represent the forefront of passenger car technology but have
|     become a separate technology.

  I admit that I have taken this out of context, but what are you
driving? The cars on your showroom floor are loaded with stuff which was
racing only 20 years ago, like fuel injection, supercharging (and
turbocharging), 4 valves/cylinder, overhead cams, etc. These things are
not on the sports models only, but in the sedans and station wagons.

  I think that micros will continue to enbody features developed in more
expensive systems. Note the trend to Harvard archetecture, pipelines,
fewer cycles per instruction, etc. New designs are adding the f.p.
to the main CPU, while keeping it as a coprocessor in the sense of
operation in parallel with the main CPU. By the turn of the century I
expect to see vector processing on the top machines, using the burst
memory access modes already in limited use.

  I don't see the feature migration or more mips ending as trends any
time in the next two decades. Advances in graphics need more address
space and CPU. All this means more bandwidth needed. Better graphics
and more CPU allows more engineering tasks, which in turn needs more
CPU. Multitasking o/s's for micros are gaining market share and drive
the need for power, as does the availability of large cheap mass
storage which leads to bigger disk servers. If AI ever really gets
going (someone finds a mass application) that will keep pushing the
need for more CPU.

  There certainly are limitations to demand, we're just not anywhere
near them. When displays reach the limit of the human eye to extract
information, the growth in that area will stop. The limit is somewhere
around 1200 dpi, 24 bits or color, and whatever screen size you need for
your working distance. The CPU limit will be reached when any problem
can be solved and the answer displayed in less than 20 ms (the time the
eye takes to notice a change). I haven't noticed that we're getting
particularly close to any of these limits, so I don't expact the demand
for progress to stop any time soon. The limiting factor will be the
ability of vendors to provide advances at effective costs.
-- 
bill davidsen	(davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen)
"The world is filled with fools. They blindly follow their so-called
'reason' in the face of the church and common sense. Any fool can see
that the world is flat!" - anon