roberts@icst-ecf.arpa (05/22/87)
A few years ago, an article was circulated pertaining to the question
of whether or not it is theoretically possible to construct a device
which performs computations with zero net dissipation of energy.
The basic premise was that changing the value of a binary logic element
always requires a change in energy level, and that quanta of energy
are pumped around in a manner somewhat analogous to the operation of
a heat engine. In one of the models described, the possibility of
dissipationless computing requires that a certain relationship exist
between the initial state and the final state of the processor; thus,
the proper handling of any unneeded bits of information becomes vitally
important.
We could never figure out whether the article was serious or not.
(For real-world applications, this will not be a problem
until the development of the peanut-sized Cray-97, for which the
dissipation of any measurable amount of energy per operation would
vaporize both the processor and the surrounding building.)
John Roberts
roberts@icst-ecf.ARPA
------mangler@cit-vax.Caltech.Edu (System Mangler) (05/31/87)
In article <7505@brl-adm.ARPA>, roberts@icst-ecf.arpa writes: > A few years ago, an article was circulated pertaining to the question > of whether or not it is theoretically possible to construct a device > which performs computations with zero net dissipation of energy. Only a few years ago, it was thought that there was a lower limit. Now it is generally believed that energy dissipation can be made arbitrarily low, either by doing the computation slowly enough, or reducing the temperature enough. (The latter is cheating, because you're really just moved the energy dissipation to the refrigeration unit). Reference: "1985 Chapel Hill Conference on VLSI", Ed: Henry Fuchs, Computer Science Press, Rockville Maryland, 1985; pp 1-3. Don Speck speck@vlsi.caltech.edu {seismo,rutgers,ames}!cit-vax!speck