Kish Says No Moore

I was one of the TAs for a microelectronics processing course in the Fall 2002 semester. Our boss, Dr. S, an industry veteran turned university lecturer, is a very funny smart man. A couple days ago, J visited Dr. S at his 3rd floor office. He gave J a copy of the "Kish Says No Moore" headline.

[Things in italics are for some seasoning.]

J: Who (in the name of J) is Kish?
Dr. S: Well, turn around & face the blackboard, use your x-ray vision.
J: (*puzzled*) Ha? I don't (quite) understand.
Dr. S: He is next door, (you unfortunate looking moron).

Anyway, Dr. Kish and his research on thermal noise predict the end of Moore's Law. {*come again?*} In 1965, Gordon Moore of Intel predicted that the increase in the number of transistors per IC is indeed exponential. Semiconductor industry pretty much follows the observation, directing the best effort to come up with a new processor every 18 months. So as we all observe, Pentium IV already had 100 million transistors compared to the 2000 transistors in the 4004 model of 1971.

The threat to the law comes from the pysical limitation of the thermal noise and the decrease in noise tolerance on the densely packed substrate.

But not to worry, there are many efforts placed to delay the death of Moore's Law, for at least a couple of scores. Folks at companies and universities across the world have been working on alternative materials, high-k dielectrics, novel processes, etc. The somewhat related question is: Have we found those killer apps in this journey of building super processor?