Every EDA marketing presentation starts off by pointing out that Moore’s law is making some problem worse. Of course, just the problem that the EDA product pitched in the rest of the presentation is designed to solve.
Everyone sorta knows Moore’s law but few people realize just what it was he said over 40 years ago, and just how prescient he was, or have even read his original paper (pdf).
In 1965, Gordon Moore was Head of R&D at Fairchild. This was several years before Moore left Fairchild to found Intel. Moore noticed that the number of transistors on the integrated circuits that Fairchild was building seemed to double every two years, as shown in the graph above from Moore’s original article. As he pointed out there, "Integrated circuits will lead to such wonders as home computers, automatic controls for automobiles, and personal portable communications equipment." Remember that this was 1965, when an integrated circuit contained 64 transistors: this was an extraordinary prediction.
Surprisingly, over 40 years on, semiconductors seem still to be increasing in complexity at this rate. Gordon Moore’s original remark is, of course, now known as “Moore’s Law” and is expected to continue for some time.
Exponential growth like this over a sustained period of time, rather like compound interest, has a dramatic effect. In the seventies a chip may have contained a few hundred transistors. Today a chip can contain billions of transistors and in the future the predictions are for chips with several hundred billion transistors. This is how all the electronics for a high-end mainframe computer can be compressed into a single chip. Only we attach a radio to it and call it a cell-phone. Or we put a lens on it and call it a digital camera. Or we attach a dish to it and call it satellite TV. Or we take it on a plane and use it to blog on Moore’s Law.
But it is possible to look at Moore’s Law the other way round: the cost of any given functionality implemented in electronics halves every two years or so. Over a period of twenty years this is a thousand-fold reduction. A video-game console, which is so cheap that children can buy them from their allowances, has far more computing power and much better graphics than the highest-end flight simulators of the 1970s, which cost millions of dollars. An ink-jet printer has far more computing power than NASA had at its disposal for the moon-shots (supposedly a total of 1 MIPS1 on all the computers they had put together). It is this exponential driving down of electronic costs that had transformed so many aspects of our lives in the last twenty years or so since integrated circuits became cheap enough to go into consumer products.
Here is Gordon Moore again, this time from a 1995 Fortune article: “The whole point of integrated circuits is to absorb the functions of what previously were discrete electronic components, to incorporate them in a single new chip, and then to give them back for free, or at least for a lot less money than what they cost as individual parts. Thus, semiconductor technology eats everything, and people who oppose it get trampled.”
Moore’s Law started as an observation, became a prediction, and eventually transformed into a blueprint for the semiconductor industry. The International Technology Roadmap for Semiconductors is largely an analysis of what it will take to make Moore’s Law continue to be true. Moore’s Law has thus become a self-fulfilling prophecy for the time being.
Of course the really interesting question is for how much longer?
1 Have you noticed that people like to write or say 1 MIP as if MIPS was plural. But, of course, the S stands for “seconds”. End of today’s nitpicking.