LESS IS MOORE

Dick Pountain/Wed 20 April 2005/11:37 am/Idealog 129

Predicting the end of 'Moore's Law' has become a right of passage for the aspiring IT journalist, but if you're not careful it can develop into an irritating tic, on a par with 'This is going to be the year of Unix' or 'One day we'll all be wearing our computers like clothes'. I consider myself lucky to have got it out of my system early in a piece I wrote for Byte back in 1997, and I try to keep silent on the subject nowadays. However that's not always possible because whenever time Intel makes a major announcement of new silicon, we get a new rash of 'Moore's Law is finished' (or not) stories.

Intel's recent launch of the dual-core Pentium Extreme Edition - along with IBM's announcement that it's working on a dual-core 970MP Power PC chip - has prompted a new round of speculation, on what is rather ambiguous evidence. While Moore's Law operates we keep shrinking circuits by roughly half every 18 months, and that ability can be exploited three ways: shrink an existing chip design (say for use in mobile phones); make its architecture more complex (deeper piplines, extra function units, more cache) for the same size; or else put two (or four, or eight) processors on a chip that used to hold only one. The mere fact of a dual-core Pentium doesn't mean Moore's Law is dead - Intel may be making the strategic decision that multi-core is now a more efficient way to exploit extra transistors than boosting the cache on a single-core. However I suspect that this sort of ambiguous interpretation won't satisfy people who need to know whether the Law still Rules or not.   

The End of Moore's Law is starting to look rather like the Second Coming, an event that sorts out people according to what they believe about it. Some, myself included, regard it (the Last Judgment that is) as a myth, albeit a powerful one that provided the emotional fuel to spread the Christian religion. Moderate Christians treat it as a real event, but one that lies always in the future, certainly not in our lifetime. And then there are those who believe it's going to happen the year after next, in Israel, after the US Airforce has carpet-bombed the armies of Beelzebub. Similarly Moore's Law separates the techno-realists from the optimists and the pessimists.

Moore's Law is not, and never was, a law in the sense that Newton's Laws of Motion are (or were, because of course in the wake of Einstein's we can see that even scientific laws are only provisional). It was a good rule-of-thumb, an heuristic, well spotted by a clever and very lucky man with a deep grasp of his industry. Gordon Moore spotted what might be better called 'Moore's Trend' in 1965, and at a recent 40th anniversary bash he admitted 'Frankly I didn't expect to be so precise - it was much more precise than it had any reason to be.'

So what reason *did* it have to be? Part of the explanation does lie in physics, namely the peculiar scaling characteristics of the CMOS fabrication process whose transistors speed up and use less power as they get smaller (see my feature in PC Pro June 2003 for details). But at least as much of the reason is economic, more particularly economic psychology. Moore's Law is actually a statement of intent rather than a statement of fact. The number of transistors on a chip only doubles every 18 months because Intel Corporation (and to some extent its rivals) wills it so: that is, because Intel invests a large proportion of its vast revenues in new fabrication plants to build the next generation of chips, rather than, say, distributing the profits to shareholders to buy yachts with. As such it's perhaps the purest expression of the spirit of modern, dynamic American capitalism. Gordon Moore might have chosen to stick with the 4004, and we now wouldn't have PCs, PDAs, mobile phones or the Internet, but rather desk calculators the size of breadbins. Of course that was never really a choice, because once he'd seen it, he (or someone else) had to go for it.

I like to think of modern techno-capitalism as a bit like riding a unicycle across a tight-rope. If you're very good, and above all very confident - and if you don't look down - you'll just zoom along. Oh, I forgot to mention there isn't any far side. This attitude was beautifully captured in an International Herald Tribune headline last year that read 'Thrifty Euros Threaten Recovery', which suggested that we cheese-eaters were not going deep enough into debt (ie. not pedalling fast enough) to float the US economy. So belief in Moore's Law is in one sense a belief in this ultra-dynamic capitalism, about which I find myself in roughly the same position as about the Last Judgement. That is, Moore's Law has been a useful and necessary myth, but when it does stop working, as it certainly will one day, that won't be the end of civilisation. It will cause some grief, some big bankruptcies, and it may - in conjunction with other, for example enviromental, factors - cause the pace of innovation to slow down a tad. My hunch is that we could learn to live with that...