SINGAPORE, 10 FEBRUARY 2009 - Scientists from Singapore’s Nanyang Technological University (NTU) and US-based Rice University have successfully created a microchip that uses 30 times less electricity while running seven times faster than today's technology.
In layman’s terms, this means a mobile phone powered by the new proof-of-concept microchip will significantly improve the device’s battery life to as much as two weeks without recharging.
The technology, dubbed ‘probabilistic complementary metal-oxide semiconductor’ (PCMOS) was invented by Professor Krishna Palem of Rice University and director of NTU’s institute for sustainable nanoelectronics.
Team member Dr Natalie Kong Zhi Hui, teaching fellow, NTU, said: “Our technology is a significant contributor towards environmental-friendliness—green computing, or probabilistic computing, with an extremely energy-aware attribute. This is due to the fact that, unlike conventional designs that view noise as a nuisance, our design concept embraces noise as a ‘gem’—this novel technology recycles noise.”
Dealing with noise
While today’s silicon transistors become increasingly 'noisy' as they get smaller, engineers have historically dealt with this by boosting the operating voltage to overpower the noise to ensure accurate calculations, leading to higher energy consumption levels.
“With this PCMOS technology, noise/parameter variations are part of the overall design and are managed as a resource to achieve significant energy savings. The success of this project will go a long way in promoting the advent of a new generation of ‘green’ IT at lower costs to consumers,” says Associate Professor Yeo Kiat Seng. He is NTU’s head of division of circuits and systems, school of electrical and electronic engineering, college of engineering.
The team hopes to realise a new generation of probabilistic-based nanoelectronics with diverse applications in media, biomedical and consumer electronics. The team envisions that PCMOS technology will enter the consumer computing market in as little as four years and may present itself as a parallel to mainstream CMOS technology in the near future.
