Decision time? Read Gizmag's latest product comparisons

'Pruned' microchips are leaner and meaner

By

March 17, 2011

Researchers are designing chips that are faster and more energy-efficient than conventiona...

Researchers are designing chips that are faster and more energy-efficient than conventional chips, by 'pruning' out the portions of circuits that are used the least (Photo: Rice University)

If you had to use a commuting bicycle in a race, you would probably set about removing the kickstand, fenders, racks and lights to make the thing as fast and efficient as possible. When engineers at Houston's Rice University are developing small, fast, energy-efficient chips for use in devices like hearing aids, it turns out they do pretty much the same thing. The removal of portions of circuits that aren't essential to the task at hand is known as "probabilistic pruning," and it results in chips that are twice as fast, use half the power, and are half the size of conventional chips.

"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,'" said principal investigator Krishna Palem, a Professor of Computing at Rice. "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."

There is a cost for those improvements, however – the pruned chips gain an error magnitude of 8 percent. The researchers are OK with that, though. The chips are designed to allow for the probability of errors, and to limit which calculations cause them. Even with that additional 8 percent error magnitude, they should reportedly still be fine for hearing aids and similar devices, which the team states can tolerate error magnitudes as high as 10 percent.

Pruned and traditional chips that were created together will be presented in a side-by-side comparison at this week's DATE11 microelectronics conference in Grenoble, France. While the pruned chip is expected to perform at twice the speed and using half the power of the regular chip, Palem has even higher hopes for a hearing aid-specific chip that he is about to design – four to five times more run time on a set of batteries.

Rice University is collaborating on the research project with Nanyang Technological University in Singapore and Switzerland's Center for Electronics and Microtechnology.

Probabilistic pruning stands in sharp contrast to the self-monitoring, self-repairing chips that we just featured in Gizmag. In that case, researchers have added cores to chips, so that tasks allocated to cores that are found to be defective can be reassigned to functioning ones.

About the Author
Ben Coxworth An experienced freelance writer, videographer and television producer, Ben's interest in all forms of innovation is particularly fanatical when it comes to human-powered transportation, film-making gear, environmentally-friendly technologies and anything that's designed to go underwater. He lives in Edmonton, Alberta, where he spends a lot of time going over the handlebars of his mountain bike, hanging out in off-leash parks, and wishing the Pacific Ocean wasn't so far away.   All articles by Ben Coxworth
3 Comments

But how can simple hardware prune away the unecssary parts is there any "intelligent software " or someone working to prune it off?

Josyula Krishna
17th March, 2011 @ 06:40 pm PDT

They might as well design and produce ASIC and mass produce it. It would be higher efficiency and lower power because that is by design. Plus, 0 errors.

Simon Tan
19th March, 2011 @ 10:34 am PDT

"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,'" said principal investigator Krishna Palem, a Professor of Computing at Rice

Dear professor has never heard of the RISC idea/architecture?

Michał Ḃorsuk
21st May, 2012 @ 01:55 am PDT
Post a Comment

Login with your gizmag account:

Or Login with Facebook:


Related Articles
Looking for something? Search our 28,719 articles