'Pruning' Down Microchips Makes for Faster, Greener Computers
<p>Researchers at Rice University have trimmed away little-used circuits on microchips to improve energy efficiency and computing speed by sacrificing accuracy.</p>
"Pruning" away the little-used circuits on microchips can double the energy efficiency and computing speed of the chips, while slightly sacrificing accuracy. It's the latest finding from Krishna Palem, the lead researcher on the project at Rice University in Houston.
"I believe this is the first time someone has taken an integrated circuit and said, 'Let's get rid of the part that we don't need,'" Palem said in a news release. "What we've shown is that we can boost performance and cut energy use simultaneously if we prune the unnecessary portions of the digital application-specific integrated circuits that are typically used in hearing aids, cameras and other multimedia devices."
Palem unveiled the new technique in France last week at the Design, Automation and Test in Europe conference.
We previously covered Palem's work on energy-efficient computing in 2009, when he was about to discuss his method for less-accurate but more efficient computing in San Francisco. At the time, Palem was unveiling a probabilistic complementary metal oxide semiconductor chip that uses probabilistic logic rather than conventional Boolean logic to perform calculations. Rather than being accurate nearly 100 percent of the time, as is the case in Boolean logic, probabilistic logic is correct fewer times -- 8 out of 10 or less.
The result of this less-accurate computing process is much greater efficiency. By cleverly managing the probability of errors and by limiting which calculations produce errors, the designers have found they can simultaneously cut energy demands and boost performance.
Palem and his graduate student Avinash Lingamneni talked about the benefits of their "inexact hardware" designs at the DATE conference, saying that it is the key to producing the next generation of energy-efficient microchips.
"Our initial tests indicate that the pruned circuits will be at least two times faster, consume about half the energy and take up about half the space of the traditional circuits," Lingamneni said. He believes the system will perform even better in the final tests, which are still underway.
One of the first potential applications for the pruned chips will be in hearing aids, which can handle error probabilities much higher than the inaccuracies created by Palem's probabilistic computing method. He believes that a hearing aid that runs on these chips will operate four to five times longer on one set of batteries than one that runs on current chips.
