Where do we go from here? (was: Re: Can anyone identify this card?)

[mjmahon@aol.com (Michael J. Mahon)]

Matt Jenkins replied:

>On Tue, 09 Dec 2003 02:19:23 +0000, Michael J. Mahon wrote:
>
>> Moore's "Law" used to give us about a factor of two
>> every year or two--usually every year in the processor
>> race.
>
>Fortunately we're still a ways from the limit of Moore's Law
>in terms of what's physically possible with the manufacturing
>processes we use. We should really get another 15 years out
>of it yet. I think though that over the next few generations
>we'll see some alternative strategies used in processor design.
>I'd be expecting (mainly from AMD) to see multi-core units start
>to appear. Increasing density in data centers is a potentially 
>large market. The ability to have 4 or even 8 2+ Ghz CPU's per
>unit of rack space would be a huge advantage over the current 2.

I agree that we can expect densities and speeds to continue to
increase, but there is serious question about whether the market
will continue to reward these advances sufficiently to repay the
exponentially increasing investments required for advancing
processes and fabs.

AMD is a rather small player in the development of new processes
and, without a considerable profit boost, will be unable to fund the
new fabs required beyond the next generation or so.

Intel and IBM certainly have the money to keep going, but with
the deceleration of the PC (and computer!) marketplace, there is
caution about the payback interval on continued huge investments
in semiconductor technology.  (The research isn't cheap, but no
major player can opt out--the real issue is the multi-billion dollar
investments required to construct next-generation fabs, without
which the research cannot lead to product).

Moore's "Law" was always an economic principle, not a physical
principle (though the phycal "scalability" of silicon structures is
what makes it possible, at least for another few generations).

Seen in this way, Moore predicted that "if semiconductor makers
invest at a rate to allow the capability of semiconductor products to
double every year or two, then the marketplace will accelerate
demand for those products so that the investment can be sustained."

This is an equilibrium statement about a condition of prolonged
exponential growth, which must, almost by its nature, be limited
after a certain number of doublings by either physical principles,
or by market weakening, or both.  While I am concerned about the
eventual problems of few-atom transistors, I am even more concerned
about the several years that the industry has had to go without major
acceleration in demand.

If anything interrupts the "virtuous cycle" of Moore's "Law", then the
result is either relative stagnation or even economic collapse of the
industry.  While I don't yet think the "jig is up", I do feel that we are
experiencing a double whammy of 1) tech capital evaporation, and
2) failure of imagination...

>> Sad to say, the market in computer applications is even more
>> sedate--everything moving from version 6 to version 7, without any big
>> improvements in capability or innovation.
>
>It's a big problem. What we really need a better software development
>environments and languages that can take advantage of the computing
>horsepower available today to dramatically reduce the amount of
>man-hours needed to develop today's software applications. In
>many ways, it the tremendous amount of human effort and hence cost
>which is stifling growth in the software industry - the corresponding
>flow-on affect to the hardware industry might even make a liar out of
>Moore!

You are suggesting that the reason we do not have a new crop of
compelling applications is that our software tools are too weak to
allow them to be developed.  Interesting.  Certainly better tools are
always useful, but one must have an idea of what to do with them.

If the past is any guide to the future, I expect the next generation of
compelling apps to be invented outside the corporate sphere, and
implemented by a few free agents.  The power of their application(s)
will not be in their cute interfaces, but in their stunning ability to do
things that people have wished to do, but considered impossible.

In short, the fault is not in our tools, but in ourselves (with apologies
to Shakespeare  ;-).

I cannot escape the irony and appropriateness of this discussion in
a forum dedicated to one of the machines that truly ignited the world
of applications that we use today.  Perhaps someone involved in that
first ignition can point the way to another.

-michael

Check out amazing quality sound for 8-bit Apples on my
Home page:  http://members.aol.com/MJMahon/