A hot calculation...
- Reduce text size Decrease text size
- Increase text size Increase text size
- Print article Print
- Jump to comments Comment
- Share this article Share
- Email article to a friend Email
Increasing chip temperatures are threatening the golden rule of processers
Forty years ago, almost to the day, an article appeared in Electronics magazine that made an incredible prediction.
In an era when commercial mainframes were just appearing, the author, Gordon E Moore, wrote: "Integrated circuits will lead to such wonders as home computers - or at least terminals connected to a central computer - automatic controls for automobiles, and personal portable communications equipment."
The article included a hand-drawn graph plotting the number of transistors that engineers had been able to "cram" onto an integrated circuit over the years 1959 to 1965. Almost unbelievably, it showed consistent, exponential growth, leading Moore to boldly predicted that processor power (i.e. the density of transistors) would double every 18 months for the foreseeable future.
In the four decades since, Moore's law has been the cornerstone of the computer industry, providing the predictability that hardware developers needed to confidently ride the price/performance curve and giving software developers and IT management the assurance that their future projects will find a powerful enough engine to drive them.
Moore, then the director of R&D and a co-founder at Fairchild Semiconductor, went on to co-found the company that 'proved' Moore's law - Intel.
But eventually physics was going to catch up with Moore's law. While Moore talked abut packing 50 transistors onto each integrated circuit, an Intel Itanium 2 chip today crams in 592 million. That, of course, generates a phenomenal amount of processing power - but it also generates a phenomenal amount of heat. And that heat is causing major problems in data centres and computer rooms.
Moore did see this cloud looming, even back in 1965. The photo caption in the article describes him as "one of the new breed of electronic engineers schooled in the physical sciences rather than in electronics", and deep into the piece he poses a question: "Heat problem: Will it be possible to remove the heat generated by tens of thousands of components in a single silicon chip?"
At that time, he reckoned that the flat surface of the chip would provide enough cooling area. But as anyone who has been near a data centre recently knows, that is not the case. Modern chips are being packed so densely and are themselves developing so much heat that data centre's are struggling to keep systems from 'melting' - almost literally.
One of our main features this month, 'Moore and more' examines why processor manufacturers, seeing the physical limitations standing in the way of Moore's law, are racing to find 'faster but cooler' chip designs - and why heat has suddenly become a critical issue not just for them but for IT executives.





