A prototype microprocessor has been developed by researchers at Yokohama National University in Japan, using superconductor devices that are around 80 times more energy efficient than the advanced semiconductor devices used in the microprocessors of today’s high-performance computer systems.
By ALANIS HAYAL
The demand for computing resources continues to grow as today’s technology become more and more incorporated into our everyday lives.
As a result of this growth, the consumption of energy for this growing computing capacity is increasing enormously.
For instance, modern data centers consume so much power that some are constructed near rivers so that the flowing water can be used to cool the machines.
Around 10 percent of the world’s energy is currently consumed by the digital communications system that serves the information era we live in today. Studies indicate that in the worst-case scenario, if there are no significant improvements in our communications infrastructure’s underlying technologies, such as the computer hardware in large data centers or the electronics that power communications.
The study of the team, published in the journal: IEEE Journal of Solid-State Circuits, describes an attempt to use superconductors to create a more energy-effective microprocessor architecture – devices that are extremely efficient but require functioning in certain environmental conditions.
To address this energy issue, the team explored the use of the adiabatic quantum flux parametron (AQFP) as a building block for ultra-low-power, high-performance microprocessors and other computing hardware for the next generation of data centers and communications networks, an extremely energy-efficient superconducting digital electronic structure.
In this work, we set out to demonstrate that AQFP is capable of high-speed, realistic energy-efficient computing, and we did so by developing and successfully demonstrating a 4-bit AQFP microprocessor prototype called MANA (Monolithic Adiabatic iNtegration Architecture), the world’s first microprocessor for adiabatic superconductors,”In this work, we set out to prove that AQFP is capable of practical energy-efficient high-speed computing, and we did so by developing and successfully demonstrating a prototype 4-bit AQFP microprocessor called MANA (Monolithic Adiabatic iNtegration Architecture), the world’s first adiabatic superconductor microprocessor,”
The demonstration of our microprocessor prototype shows that all facets of computation can be done by the AQFP, namely data processing and data storage.
We are also demonstrating on a separate chip that the microprocessor’s data-processing portion can run up to a 2.5 GHz clock speed, placing it on par with today’s computing technologies.
If we enhance our design methodology and experimental setup, we also expect this frequency to be increased to 5-10 GHz,” Ayala said.
However, to work effectively, superconductors need extremely cool temperatures. One would think that the power requirements become undesirable when you add in the cooling required for a superconductor microprocessor and outstrip today’s microprocessors.
This, however, was shockingly not the case, according to the research team:
“The AQFP is a superconducting electronic system, which means that to allow the AQFPs to reach the superconducting state, we need additional energy to cool our chips down from room temperature to 4.2 Kelvin.
But even accounting for this cooling requirement, compared to the most advanced semiconductor electronic devices used in high-performance computer chips today, the AQFP is still around 80 times more energy efficient.
Now that the team has proved the principle of this superconductor chip architecture, it aims to optimize the chip and, after optimization, evaluate the chip’s scalability and pace.
“We are now working to improve the technology, including developing more compact AQFP devices, increasing operating speed, and further increasing energy efficiency through reversible computing,” said Ayala. “We’re also scaling our design approach so we can fit as many devices as possible on a single chip and run them all reliably at high clock frequencies.”
The team is also interested in exploring how AQFPs could help with other computing applications, such as neuromorphic computing hardware for artificial intelligence and quantum computing applications, in addition to building standard microprocessors.
Originally published at Brinkwire