Re: Software directions to ponder over...

Date view Thread view Subject view Author view

From: Eugene Leitl (
Date: Tue Sep 26 2000 - 02:47:36 PDT

Koen Holtman writes:

> Actually, I recall reading somewhere that 3 *is* the optimal base.
> The argument goes something like this: suppose you have a linearly scaling
> manufacturing process, so that for any N, a storage cell with N states

You can't scale below molecular component size.

And if you want to process these bits in a lively fashion, you better
put your processing logics right into the vicinity of your
storage. Relativistic latency already starts biting even on the die
scale, when doing it with molecules a micron is a already a large
distance. A provably optimal computer architecture is a hardware
cellular automaton machine, at the high end from molecular components.
(I have a hunch, one with an Edge of Chaos rule, but that one has not
been proven yet).

> will have a size s*N for some constant s. How large should you make your

Using quantum energy levels, you can store several bits in essentially
the same volume. Of course, then you have to deal with implementing a
more or less deterministic machine with probabilistic elements, which
will be a bear.

> cells to get optimum storage density when you pack lots of cells together?
> Turns out that the optimum is N=e (2.718....) which you should round to 3
> for practical purposes.
A yet another useless theoretical result. Ternary logics, huh?

Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Sep 26 2000 - 03:55:36 PDT