Ever left a laptop running on your lap for too long? The heat you feel is the feeling of physics hitting the limit.
The limit, in this case, is Moore's law, which says the number of transistors that can be packed inexpensively on a chip doubles approximately every two years. As chip makers cram up to two billion transistors on a chip, computers have become incredibly powerful, but along the way they face a fundamental challenge in terms of their thermodynamics. The transistors on the chip become "leaky" and dissipate heat even when the computer is turned off.
That leads to the possibility that the doubling of computing power on a chip could come to a grinding halt unless researchers can find alternate way to dissipate the heat. Even better would be if they could find new ways to compute that go beyond the transistor switch.
"Heat dissipation is the most imminent and dangerous limit to Moore's law," says Avik Ghosh, assistant professor at the department of electrical and computer engineering in the University of Virginia. "If we don't do anything about it, at our current pace of miniaturization, in next 10 to 20 years laptops will become as hot as the temperature of the sun. Though of course they would have melted long before that."
Ghosh, his colleague Mircea Stan, and other researchers are trying to solve the problem of the overheating laptops. It's not a question of just getting better batteries or heat sink, says Ghosh, but goes to the heart of how transistor-based switching in computers.
Today computation is done by moving electrons from source to drain inside a transistor switch. But as transistors continue to shrink, the electrons can move spontaneously even when the laptop is turned off. "This will dissipate heat even when you are not doing anything with the machine," says
Ghosh.
Given billions of transistors on a chip, the process can generate enough thermal energy to make laptops burn up.
Tackling this problem will require Ghosh and other researchers to take another look at the Second Law of
Thermodynamics, which states that heat transfers from a hotter unit to a cooler one until both have roughly the same temperature—in this case, heat transfer between electrical computer components.
One way to solve the problem would be to disrupt the flow of heat between the components. Another would be to engineer devices to convert the excess thermal energy into directed motion.
Even better, says Ghosh, would be to find alternate ways to compute by encoding information differently, either in the spin of an electron or its nuclei. These methods tread into the field of quantum computing but Ghosh believes the future depends on it.
Photo: Travellin Librarian/Flickr
