Computer chip to run faster, use less power, thrive on errors

In the first real-world test of a revolutionary type of computing that thrives on random errors, scientists have created a microchip that uses 30 times less electricity while running seven times faster than today’s best technology.

The U.S.-Singapore team developing the technology, dubbed PCMOS [pronounced “pee-cee-moss”], revealed the results here today at the International Solid-State Circuits Conference (ISSCC), the world’s premier forum for engineers working at the cutting edge of integrated-circuit design.

Conceived by Rice University Professor Krishna Palem, PCMOS piggybacks on the “complementary metal-oxide semiconductor” technology, or CMOS, that chipmakers already use. That means chipmakers won’t have to buy new equipment to support PCMOS, or “probabilistic” CMOS. Although PCMOS runs on standard silicon, it breaks with computing’s past by abandoning the set of mathematical rules — called Boolean logic — that have thus far been used in all digital computers. PCMOS instead uses probabilistic logic, a new form of logic developed by Palem and his doctoral student, Lakshmi Chakrapani.

“A significant achievement here is the validation of Rice’s probabilistic analogue to Boolean logic using PCMOS,” said Shekhar Borkar, an Intel Fellow and director of Intel’s Microprocessor Technology Lab. “Coupled with the significant energy and speed advantages that PCMOS offers, this logic will prove extremely important because basic physics dictates that future transistor-based logic will need probabilistic methods.”

Silicon transistors become increasingly ‘noisy’ as they get smaller, but engineers have historically dealt with this by boosting the operating voltage to overpower the noise and ensure accurate calculations. Chips with more and smaller transistors are consequently more power-hungry.

“PCMOS is fundamentally different,” Palem said. “We lower the voltage dramatically and deal with the resulting computational errors by embracing the errors and uncertainties through probabilistic logic.”

PCMOS was jointly validated by Rice and Nanyang Technological University (NTU) in Singapore via a joint institute that Palem founded in 2007, the Institute for Sustainable Nanoelectronics (ISNE). Directed by Palem, ISNE is based at NTU, where the first prototype PCMOS chips were manufactured last year in collaboration with Professor Yeo Kiat Seng and his team.

The prototypes were application-specific integrated circuits, or ASICs, that were designed solely for encryption. Unlike the general-purpose microprocessors that power PCs and laptops, ASICs are designed for a specific purpose, and they are “embedded” by the millions each year in a growing constellation of products like automobiles, cell phones, MRI scanners and electronic toys.

The Rice-NTU team plans to follow its proof-of-concept work on encryption with proof-of-concept tests on microchips for cell phones, graphics cards and medical implants.

Palem said PCMOS is ideally suited for encryption, a process that relies on generating random numbers. It’s equally well-suited for graphics, but for different reasons. In a streaming video application on a cell phone, for example, it is unnecessary to conduct precise calculations. The small screen, combined with the human brain’s ability to process less-than-perfect pictures, results in a case where the picture looks just as good with a calculation that’s only approximately correct.

“The key is to consider the value that the computed information has for the user,” said Palem, who directs Rice’s Value of Information-based Sustainable Embedded Nanocomputing Center, or VISEN. “Our goal is green computing. We’re looking for applications where PCMOS can deliver as well as or better than existing technology but with a fraction of the energy.”

If PCMOS can slash energy use for embedded ASICs in key devices, the implications are enormous. For consumers, it could mean the difference between charging a cell phone every few weeks instead of every few days. Globally, that would help reduce the information technology industry’s carbon footprint.

“Based on our findings, we view PCMOS as a path to help IT become more ‘green’ even as it keeps pace with Moore’s Law,” said Palem, the Ken and Audrey Kennedy Professor of Computing, professor of computer science, professor of electrical and computer engineering, and professor of statistics.

Palem said he hopes PCMOS technology will enter the embedded computing market in as little as four years.

Palem’s PCMOS research was funded by the Defense Advanced Research Projects Agency and Intel Corp.

Substack subscription form sign up
The material in this press release comes from the originating research organization. Content may be edited for style and length. Want more? Sign up for our daily email.