Path: utzoo!attcan!uunet!amdahl!ames!killer!elg
From: elg@killer.Dallas.TX.US (Eric Green)
Newsgroups: comp.arch
Subject: Re: was: Don't look back. What about looking east?
Message-ID: <7499@killer.Dallas.TX.US>
Date: 11 Mar 89 03:12:15 GMT
References: <7287@spool.cs.wisc.edu>
Organization: The Unix(R) Connection, Dallas, Texas
Lines: 87

in article <7287@spool.cs.wisc.edu>, shekita@provolone.cs.wisc.edu (Eugene Shekita) says:
> While we're on the subject of looking back, forward, etc., I was wondering 
> how much you architects out there worry about our microprocessor market 
> falling prey to Japan. They've squashed our masking tools market, the DRAM 
> market, and they seem to have better production yields, so isn't the next 
> logical step the microprocessor market? 

Yes, and, in fact, they're going after it. The biggest Japanese
microprocessor exporter right now seems to be NEC, with its V-series
chips. Now that the Intel suit is over, you can expect to see a lot of
design-ins. 
     The Japanese seem to be going more after the high-end
microcontroller market than the systems market. Or maybe that's just
what they're most effective at. For example, my brother's company is
designing-in the V25 into their top-end data aquisition product. The
V25 has everything but RAM and ROM onboard for such an application --
it has 24 i/o lines, 8 analog inputs, two UARTs, interrupt controller,
2 DMA channels, DRAM refresh, etc. etc. all in an 84-pin package. And
in low power CMOS -- under 50ma, a prime consideration when the thing
has to be solar-powered. That level of integration one of the reasons
they're using it.  The other reasons: its performance is more than
adequate for most applications, and, best of all, it is 8086
instruction-set-compatible.  
      I cannot stress how much instruction set compatibility means to
a smaller company making specialty products. If it is 68K or
8086-compatible, they can use off-the-shelf development systems and
"C" compilers. If they wanted to use the 29K, on the other hand, (as
if an adequately integrated version of any RISC chip existed for
imbedded markets) they'd have to buy an upteen-K$ development system,
which they'd then have to spend weeks learning. I think CSI spent
about $5K on their initial design stage, including buying new PCB
layout software and hiring an additional engineering tech to do the
layout, final schematic, and (when the time comes) fabrication. And
did it in less than three weeks, INCLUDING software, due to the
instruction-set compatibility -- they had previously done PC-based
data aquisition products, and most of their (Microsoft "C") code was
easily modified for the new environment. I suspect total development
time will be under three months, because the engineers are familiar with
the application and the development environment... 

> pipelining, etc. are pretty well understood at this point. Seems like the 
> winners in the near future will be the ones who can pack the most on one 
> chip (witness the N-10) and still get decent yields. The Japanese are
> obviously well-positioned to do just this.

Yep! (see above).

> On top of everything else, many of the Japanese computer companies
> are hugh and will be far less effected by industry slow-downs. They also
> have gobs of money to spend on research, which allows them to go out
> on a limb with technologies that could have a big payback.

The Japanese rarely go out on a limb with "unsafe" technologies. Even
their most noted innovator, Sony, generally only takes existing
technology and brings it to a marketable form.  Their advantage is
that they are more long-term oriented, and won't put off innovation in
order to improve the short-term bottom-line. For example, the high-end
Japanese autos have 16-valve engines, multi-point fuel injection, and
four-wheel independent suspension.  None of those are "revolutionary"
technologies -- 4 valve cylinders have been used in motorcycles for
years, and expensive Euro-cars have long had fuel injection and
independent suspension. But all of these technologies were late to be
adopted by Detroit, because the American automakers were "running
scared"... they were having enough trouble with redesigning their
entire car lines and making capital investments to bring their
factories into the '80s after a decade of neglect (the '70s saw no
innovation at all from Detroit, nothing like the massive redesign of
the '80s that has resulted in most Detroit cars being front wheel
drive and 2,000 pounds lighter than in 1970). 

> So how much do you guys worry? If you believe that the Japanese will
> never overtake our microprocessor market (we're talking a 10-year horizon),
> then I'd be curious to hear why. Of course, I hope it doesn't

I don't think they'll overtake our "systems" markets. Intel and
Motorola come out with new "standard" processors faster than the
Japanese can catch up, in that market, and the Japanese don't have the
software clout to go non-standard (i.e., RISC). But I do think that in
the embedded processor market, they're going to mop up. Their
highly-integrated code-compatible processors simply are more cost
effective than what American companies, beglamored by the flashy
systems market, are putting out.

--
|    // Eric Lee Green              P.O. Box 92191, Lafayette, LA 70509     |
|   //  ..!{ames,decwrl,mit-eddie,osu-cis}!killer!elg     (318)989-9849     |
| \X/         amoy mi amiga, pero no me gusta bazofia-DOS.                  |