Moore's Law no more?

Or maybe we shouldn't even consider it the Intel founder's greatest legacy

Or maybe we shouldn't even consider it the Intel founder's greatest legacy

Can processor power continue to double every 18 months? And why does it have to? Ben King looks at the accidental birth and surprisingly long life of one of the classic high-tech maxims. Gordon Moore is best known as the man who said computer power doubles every 18 months, while prices for a given product tend to fall. He's also widely known as one of the founders of Intel, with Bob Noyce, back in 1968. Oddly, though, Moore is much more likely to confess to founding Intel than coming up with Moore's Law. He traces the origin of the law back to 1965 when he was working for his first start-up, Fairchild Semiconductor. The birth of Moore's Law was unintentional, and its progenitor was not altogether happy about it. It started from a magazine article in 1965, he says in an oral history interview on Intel's website. "I was asked to predict what would likely happen in component technology for the next 10 years. And I looked at what had happened for the first few [years] in integrated circuits and extrapolated it for the next 10," he says. "I predicted that the cheapest way to buy components would be as one of a very complex circuit, with something like 65,000 components on it, which is a very long extrapolation from 60. A thousand-fold extrapolation. And it turned out to be surprisingly correct. And somebody ended up dubbing it Moore's Law." The fact that people still call it Moore's Law probably owes as much to the fact that the two words rhyme as it does to anything else. Now, as Moore says: "The term has been applied to almost anything that changes exponentially in the industry, and I'm happy to take credit for all of it." Nonetheless, Moore himself remains one of the biggest sceptics about the long-term viability of his own law. He regularly appears on record saying that the processor industry cannot keep up with it, even though he has thus far been proved wrong. As Craig Barrett, Intel's current president and CEO told Forbes magazine in 1997: "I have been at Intel for 23 years now, and Gordon Moore has been saying all along that the industry couldn't keep up the pace of Moore's Law. But it does." It remains true even today: Intel launched the 1GHz Pentium III chip in March 2000 for $990. Almost exactly 18 months later, it launched the 2GHz Pentium 4 in August 2001 for $1499. All kinds of reasons why Moore's Law must eventually fail crop up from time to time. To make faster processors, the transistors which are the basic building blocks of computer chips have to be smaller and smaller. But the smaller they get, the harder it is to make them. How will they make transistors smaller than the wavelength of the lasers that etch them onto the silicon wafers? Will they dissipate so much heat that no fan will be able to keep them cool? What happens when we try to make chips no bigger than the atoms of silicon on which they are etched? Technical problems such as these have been cropping up ever since 1965 and an answer has always been found. We're still some way from needing transistors smaller than atoms, but research is already underway on "quantum computing" which uses individual atoms instead of transistors as the building blocks of logical circuits. For the moment, the biggest threat to Moore's Law is not the laws of physics but the cold reasoning of the marketplace. There is another law which states the cost of building chip factories doubles every 18 months and they're already several billion dollars a pop. Will it be possible to justify this level of investment, when the technology industry is in recession and seems unlikely ever to grow as fast as it once did? Not if the PC market fails to wake up from its current state of catatonic inactivity. Nowadays most PCs are perfectly adequate for the tasks they are used for. Neither businesses nor consumers feel motivated to upgrade their current machines, particularly with a recession on the horizon. If no one is buying new PCs, no one is going to buy the most common of processors, no matter how fast they are. Tony Lock, senior analyst at Bloor Research, said: "I think that the next 12 months will be crucial. If things don't pick up, we could see the prices of new processors start to rise, and products launch every 24 months instead of 18." In other words, it'll be the end of Moore's Law. In the high-end server and handheld device markets, there is still a need for faster processors, and Lock expects to see these areas to continue to develop in accordance with Moore's Law, even if processors for PCs can't keep pace. Despite the longevity of the law, Gordon Moore's major contribution to the history of high-tech is arguably the invention of the Silicon Valley start-up culture. With Intel, Moore and co-founder Robert Noyce pioneered the model of building huge companies from scratch by making small teams of bright young workaholics slog their guts out for share options instead of cash. While many of those companies didn't survive, Moore's start-up model did, and it's now so commonplace that it hardly gets noticed. While Moore's Law would be lucky to survive the next five years, there's no reason why the Intel-style start up won't be around for centuries. So the fact that Gordon Moore is only known as the father of Moore's Law would be a bit like Bill Gates being known as the man who brought us Mr Clippy.