If you could remove the layers of circuitry in your computer and touch the main processor while it’s running a video, you would feel its blistering heat, which can exceed 100 °C. Such heat, a natural by-product of shuttling electrons through transistors, can impede performance and even damage the processor in the long run. Traditionally, engineers have used simple copper plates to pull away the heat, and fans or liquid-based cooling systems. But these systems are bulky and can sap energy.
Now researchers at Intel, RTI International of North Carolina, and Arizona State University have shown that it’s possible to build an efficient microrefrigerator that can target hot spots on chips, saving power and space, and more effectively cooling the entire system. Their work also demonstrates, for the first time, that it is possible to integrate thermoelectric material into chip packaging, making the technology more practical than ever before. A paper detailing the research was just published in Nature Nanotechnology.
The fundamental technology used to chill the chip, a thermoelectric cooler, isn’t new, explains Rama Venkatasubramanian, senior research director at the Center for Solid State Energetics at RTI International. In a Nature paper from 2001, he and his team showed that a material called a nanostructured thin-film superlattice has superior thermal properties to other types of thin thermoelectric materials: the superlattice conducts electricity well but impedes the flow of heat. When an electric current zips through the material, its temperature can drop to about 55 °C.
“People have been talking about using high-efficiency thermoelectric materials for cooling hot spots on chips for years,” says Intel manager Ravi Prasher. He says that part of the reason he and his colleagues were able to succeed is because they used a material that has shown exceptional thermal properties, and they relied on Intel’s knowledge of chip packaging to build an integrated thermoelectric system that was engineered to fit within the confines of a chip’s housing.
To put the microrefrigerator in the chip package, the engineers integrated the cooler onto a square of copper, just like the type that’s already used in chip packaging to disperse heat. Usually this piece of copper is in close contact with the chip, but the researchers put the 0.4-millimeter-square cooler in between the chip and the copper. When the microrefrigerator was turned on, it cooled a localized region on the chip by about 15 °C. This is significant, says Venkatasubramanian, because generally speaking, for each five-degree increase in chip temperature, there is a marked decrease in reliability and performance of a chip. In the demonstration, the researchers only used one microrefrigerating unit but foresee using three or four per chip, to cover the hottest areas.
However, the performance wasn’t even close to the maximum amount of cooling that the microrefrigerator is capable of when it’s not confined to the chip housing. “We’ve found good performance,” Venkatasubramanian says, “but there are still a lot of challenges.” When engineers put the cooler inside the package, there are a number of additional contact points where the cooler is connected to the copper plate and packaging electronics, he says. Prasher explains that the thermal characteristics of these contacts play a significant role in reducing the cooler’s efficiency: “By itself, [reducing resistance of thermal contacts] is a significant research area.” People are exploring different types of solder and even carbon nanotubes to reduce the resistance at the interface, he says, but the problem still has to be resolved.
Regardless, Ali Shakouri, a professor of electrical engineering at the University of California, Santa Cruz, is impressed by the work so far. “This is a good achievement,” he says. “The idea [that] there’s an uneven distribution of temperature in a microprocessor, and that by selectively cooling certain locations you can do a better job and save power, has been around for a while, but it hadn’t been demonstrated on a chip before.”
Shakouri notes that as the microprocessor industry moves toward using multiple cores or processing centers on a chip, the problem of hot spots will get worse, because workloads are shifted from core to core, creating more transient hot spots. Fans, used in many computers today, don’t respond quickly or effectively. “If you could selectively have microrefrigerators throughout a multicore chip,” he says, “you could lower power and increase performance.”
The researchers don’t have a timeline for commercialization. Right now, even though the cooler could be incorporated into traditional chip packaging, it would still be prohibitively expensive. After all, says Venkatasubramanian, adding a cooler is essentially adding a completely new layer of electronics to a chip. He says that if the cost and scalability of these coolers can be addressed, then he’s confident that they’ll find a market.
These weird virtual creatures evolve their bodies to solve problems
They show how intelligence and body plans are closely linked—and could unlock AI for robots.
A horrifying new AI app swaps women into porn videos with a click
Deepfake researchers have long feared the day this would arrive.
Chinese hackers disguised themselves as Iran to target Israel
But they left a few clues that gave them away.
DeepMind says it will release the structure of every protein known to science
The company has already used its protein-folding AI, AlphaFold, to generate structures for the human proteome, as well as yeast, fruit flies, mice, and more.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.