Researchers from Japan based Yokohama National University (YNU) have reportedly developed a novel prototype microprocessor utilizing superconductor devices that are more energy efficient than the present state-of-the-art semiconductor devices used in the microprocessors of high-performance computing systems of today.
Speaking of the innovation, Christopher Ayala, associate professor, Yokohama National University and lead author of the study stated that the present digital communications infrastructure uses nearly 10% of the world’s electricity. Subsequently, studies have suggested that in case of no fundamental change is made in the underlying technology pertaining with communications infrastructure such as computing hardware across large data centers or electronics driving the communication networks, this electricity usage may increase to around 50% of global electricity by 2030.
As a solution to the problem, the team reportedly explored the application of a superconductor digital electronic structure having extremely high energy efficiency, called AQFP (adiabatic quantum-flux-parametron), as a fundamental unit for high-performance, ultra-low power microprocessors, and various computing hardware for future communication networks and data centers.
Ayala stated that the team intended to prove the potential use of the AQFP in high-speed and energy-efficient computing, and they did this through the development and demonstration of a prototype AQFP microprocessor of 4-bit named Monolithic Adiabatic Integration Architecture (MANA), the first ever adiabatic superconductor microprocessor.
Ayala also added that the demonstration of the prototype microprocessor indicates that the AQFP is suitable for all aspects of computing, namely: data storage and data processing. The team also showed that the data processing of this microprocessor can work up to 2.5 GHz of clock frequency making it as capable as present computing technologies. The team further expects the frequency to reach 5-10 GHz as they enhance the design methodology and experimental setup.
Commenting on the energy efficiency of AQFP, Ayala stated that the as it is a superconductor device, it requires additional power for cooling the chips to 4.2 Kelvin from room temperature in order to enter the superconducting state. However, taking this cooling overhead into consideration, the AQFP is still about 80 times more energy efficient than the state-of-the-art semiconductor devices found in the present high-performance computer chips.
As the team has proven this superconductor chip architecture concept, they further plan to optimize the chip and understand the chip’s scalability and speed after optimization. Moreover, the team is also examining the future use of AQFPs in other computing applications including quantum computing applications as well as neuromorphic computing hardware in artificial intelligence.