Edgar Allan Poe’s poem “The Raven” describes a night when a black bird flew into the study of the man who is narrating. The raven answers all the narrator’s questions in the same way, by intoning the single word “Nevermore”, answering this way even his question about whether he will ever see his deceased love Leonore again, in the afterlife.
Moore’s Law describes the blistering pace of development that the semiconductor industry has followed, as a self-imposed discipline more than as a law of nature, for more than 50 years. It’s an exponential growth law. It says that you get 2 every 2. (i.e., the density of transistors in an integrated circuit increases by a factor of 2 every 2 years.)
Will we ever see our love LawMoore again, or are we destined to see her nevermoore?
Intel, a.k.a. chipzilla, sets the pace in the semiconductor industry. Their Sandy Bridge 32nm (nanometer) node-length chips are already available in the market, and their Ivy Bridge 22nm node-length chips are scheduled to arrive before the end of this year.
Intel has announced that it is ready to start construction this year of a $5 billion fab in Arizona that will manufacture 14 nm node-length chips. The fab is expected to go on line by 2013.
The 14nm node-length chips will be followed by 11nm node-length ones, which, if current trends hold, will arrive to market by approx. 2015.
The future beyond 11nm is hard to predict. Even the 11nm node-length chips are iffy at this point in time. That’s because the distance between atoms in an unstrained silicon crystal is about 0.5nm. A transistor with 11nm node-length will have a gate-length of about 6nm (about 12 Si atoms). Theoretical calculations predict that transistors with such tiny gate lengths will suffer from significant amounts of quantum tunneling. As a result, going below 11nm will require a paradigm shift: exploiting optical transistors or spin-electronics, or some other very new techniques and phenomena. No doubt, any of these new techniques will involve a large dose of quantum mechanics.
Quantum Mechanics is a Borg, and its message to chip manufacturers is: “Resistance is futile, you will be assimilated. And if you are going to do quantum mechanics, you might as well also consider quantum computation, which can solve some problems much faster (with a more favorable time complexity) than classical computation.”