I saw this item on Slashdot the other day.  arstechnica.com has published an article that discusses the theoretical ultimate limit of processor speed.  Using a hypothetical 1 kg computer, the author computes the available energy (E=mc^2), and the amount of energy needed to effect a state change in the computer.  It’s a fascinating article.  The math is approachable, but not entirely necessary to understand the article.  Note that this is a theoretical article that doesn’t concern itself with manufacturing processes, heat dissipation, or even the problem of where to get the energy to drive the thing.  On the contrary, if you converted the entire mass to energy in order to effect a state change, there’d be no more matter left to change state.  So the practical limit is surely much lower than the theoretical limit.

This is the first article I’ve seen that approaches the problem from a strictly theoretical standpoint.  All of the other articles I’ve read discuss limits of the current manufacturing processes or materials.  Good stuff, and well worth the read.