🚀 go-pugleaf

>(AI is very processer intensive) and weight/power are pretty critical in these
>situations.  Still  I doubt that the soviets will make a major industry out of
>micrograv. crystals, but if they make a few super applications with them it

For super-performance, we can use parallel processing.  Granted, some problems
cannot be parallelized (such as compiling a program).

Bear in mind that the super-computer built from microgravity silicon and
meteoric iron will be surpassed in performance by conventional technology
in 2 to 5 years.  (Of course, the new technology 2 to 5 years from now
could also benefit from super-perfect crystals;  perhaps this suggests
stockpiling these crystals, and only turning them into IC's in the last
months before war.)

An inaccuracy in my previous statement on this topic is the idea you would
use low-defect silicon to make tight-geometry chips.  Actually, you would make 
larger chips.  A more advanced silicon technology would be used to tighten
geometries.  Larger chips might be linear extensions of conventional
technology chips.  For example, you could take the standard CPU+cache chip
and make it with a cache 10X larger.

I once toyed with the idea of a burst computer.  This would be a machine
which would provide 100X or 1000X performance for, say, 10 or 100 milliseconds.
I figured it would mostly be useful for compiling programs and refreshing
raster-graphics displays.  A friend of mine suggested using it for solving
complex pattern recognition problems during a missile attack.  Because a
missile attack could take several minutes, I imagined a long row of burst
computers being gobbled up one-at-a-time.

(BTW, I was just kidding about using meteoric iron.  But you should consider
microgravity silicon a similarly scarce resource.)