Efficiency through s-MTJs Energy-efficient computer combines CMOS with stochastic nanomagnets

Related Vendors

FAULHABER Drive Systems

Wuxi InfiMotion Technology Co., Ltd.

Researchers from Tohoku University and the University of California, Santa Barbara, have introduced the prototype of a probabilistic computer.

A team of researchers from Tohoku University and the University of California devoted themselves to a particularly exciting project, namely a computer prototype that can be made with near-future technology. It combines a CMOS circuit with a limited number of stochastic nanomagnets, resulting in a heterogeneous probabilistic computer.

The development of computers that are capable of efficiently executing probabilistic algorithms, which are often used in artificial intelligence and machine learning, is a challenge that scientists have long been trying to overcome.

The approach presented in this work represents a promising and viable solution to this problem, with researchers confirming that the superior computational power and energy efficiency surpasses the current CMOS technology. The details of this breakthrough were published in Nature Communications.

Probabilistic algorithms for probability calculation

Artificial intelligence and machine learning have brought about a profound change in society in recent times. These technologies employ probabilistic algorithms to solve problems that are fraught with uncertainty or where an exact solution is computationally unfeasible.

These operations follow specific instructions within CMOS circuits, but sometimes there are discrepancies between the interplay of software (instructions) and hardware (circuits), leading to different results. As the role of artificial intelligence and machine learning increases, there is a great need for a new computer paradigm that compensifies for this discrepancy by achieving higher complexity while significantly reducing energy consumption.

Easy to manufacture in the future

In the study "Researchers Develop Energy-Efficient Computer by Combining CMOS with Stochastic Nanomagnet", doctoral student Keito Kobayashi and Professor Shunsuke Fukami from Tohoku University, together with Dr. Kerem Camsari from the University of California, Santa Barbara and colleagues, developed a futuristic, heterogeneous version of a probabilistic computer, tailored to the execution of probabilistic algorithms and simple manufacturing.

"Our constructed prototype has demonstrated that excellent computing power can be achieved by operating pseudorandom number generators in a deterministic CMOS circuit with physical random numbers generated by a limited number of stochastic nanomagnets," Fukami emphasized. "More specifically, a limited number of probabilistic bits (p-bits) with a stochastic magnetic tunnel junction (s-MTJ) should be manufacturable with a future integration technology."

The final form of the spintronic probabilistic computer, which is mainly made up of s-MTJs, is expected to be much more efficient at executing probabilistic algorithms compared to current CMOS circuits, according to the research. They are talking about a reduction in area by a four-digit order of magnitude and a reduction in energy consumption by a three-digit order of magnitude.

Ultimately, Fukami and his colleagues' prototype pushes the boundaries of current deterministic CMOS circuits for artificial intelligence and machine learning. "We anticipate that future research and development work will lead to the introduction of innovative computer hardware characterized by exceptional computing power and energy-saving capabilities into society," Fukami added.

Literature: "CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning".

Nature Communications: DOI: 10.1038/s41467-024-46645-6 (sb)

Semiconductor technology

World's first n-channel Diamond FET

Link: Tohoku University