Sign In

Communications of the ACM

ACM TechNews

To Make Computers Better, Let Them Get Sloppy


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
A less-accurate computer.

Rice University's Krishna Palem wants to address the trade-off between computing performance and energy efficiency.

Credit: Spencer Wilson

Rice University's Krishna Palem says he wants to address the trade-off between computing performance and energy efficiency with a proposal "to alter the computer itself to give you cheaper but slightly less accurate answers."

Palem designed a probabilistic and deliberately unstable version of complementary metal-oxide semiconductor technology that saves power at the cost of precision.

In a test of the concept, Palem's team built a digital video decoder that interpreted the least significant bits imprecisely when rendering pixel data as screen colors; little loss in image quality was noticed by human viewers.

The team is now experimenting with hearing aids, and initial tests found imprecise digital processing can cut power consumption by half while only producing a minor reduction in intelligibility.

University of Oxford researcher Tim Palmer thinks Palem's energy-efficient chips could be integral to the creation of computers that can make more accurate climate change models. "If we can reduce the number of bits that you need to do calculations, that would have an enormous impact on energy consumption," he notes.

To meet the challenge of selecting which bits to downgrade, researchers are developing methods to code specific accuracy thresholds so programmers can say when and where errors are acceptable. Other uses that may benefit from inexact computing include accident-investigation simulation.

From New Scientist
View Full Article - May Require Free Registration

 

Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account