/ 21 April 2005

Moore and more

This week, Moore’s Law reached its 40th birthday. Well, it’s 40 years since Electronics magazine published an article by Gordon Moore, co-founder of Intel, entitled ”Cramming more components on to integrated circuits”.

Moore had noticed that the number of components on a chip had doubled, from 30 to 60, in a year. ”Over the longer term,” he wrote, ”the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years. That means by 1975, the number of components per integrated circuit for minimum cost will be 65 000.”

Today, the number of transistors on a microprocessor can reach about 1,7-billion. Moore didn’t call his observation a ”law” — it wasn’t, and never has been. He didn’t say it would last forever — exponential growth never does. He didn’t say microprocessor clock speeds would double every 18 months — microprocessors hadn’t been invented. He didn’t say anything about disk drives or other things that have also enjoyed exponential growth in the past four decades.

None the less, the very idea of Moore’s Law had a number of dramatic effects. For one thing, it became a self-fulfilling prophecy. Chip manufacturers working on projects that took three or four years to complete had a target: they knew more or less what they would have to achieve for their products to be competitive.

It also enabled hardware and software designers, and even users, to embark on projects that might not make commercial sense at the time — or were actually impossible — but would become feasible and affordable if the chip manufacturers kept delivering more and more power on schedule.

And they have, very roughly, tracked Moore’s law as he revised it in 1975, to say that the number of components on a chip would double every two years instead of every year. In a conference call last week, Moore said: ”I have been amazed that we’ve been able to keep it going so long. I never expected to predict so far in the future.”

What will happen in 10 or 20 years, when chip miniaturisation bumps up against fundamental limits? We don’t know. However, Ray Kurzweil has observed that Moore’s Law is just one example of what he calls The Law of Accelerating Returns. Earlier technologies such as mechanical calculators, and valve-based and transistor-based computers also followed Moore’s Law, redefined to mean ”increased computational speed per unit cost”.

Kurzweil plotted 49 machines and was surprised to find that ”what I got was another exponential curve. In other words, there’s exponential growth in the rate of exponential growth. Computer speed (per unit cost) doubled every three years between 1910 and 1950, doubled every two years between 1950 and 1966, and is now doubling every year.”

It remains to be seen what, if anything, will start to take over from integrated chips in around 2019. Current ”new technologies” such as quantum computing don’t look promising. However, lots of people must have thought chips didn’t look promising in 1965. – Guardian Unlimited Â