Working with Intel, Google and colleagues at Dartmouth College, Princeton researchers have completely re-engineered high-performance computer power technology, resulting in systems that increase power delivery 10 times beyond the current state of the art.
Energy demand ultimately comes from IT demand. As designers pack more circuitry into microscopic sections of microchips, the power delivered to tiny circuitry clusters is approaching equivalent levels inside a nuclear reactor. Engineers call the amount of energy delivered to an area power density. The power density of supercomputers reaches nuclear levels not primarily because of the computing requirements, but because of the tiny spaces of the computer chips.
This highly controlled power demand has proven to be a limitation for chip designers and a major challenge for power electronics engineers. Sending large amounts of power into tiny areas generates heat, which is not only inefficient, but can also be fatal to computer components.
“It must be extremely efficient with very low noise, precisely controlled in a very small area,” said Minjie Chen, assistant professor of electrical and computer engineering and leader of the Princeton research team. “If the efficiency isn’t high enough, you’ll overheat. If the components overheat, you won’t be able to deliver the power.”
Among other areas, Chen said his research team has focused on developing “smaller, smarter and more efficient power electronics for emerging and important applications.” Although new generations of chips and circuits are attracting the most attention, power delivery is an increasingly critical part of computer system design. As Chen points out, better power supply is needed for small smartphones and other devices, more efficient server centers and advanced processors to support increasingly sophisticated artificial intelligence systems. The group is also looking beyond computer systems to design programmable solar panels, the smart grid and other critical infrastructure.
In the most recent project, presented in an article in the IEEE Transactions on Power Electronicsthe researchers demonstrated a strategy for achieving industry goals in a way that can be applied to small systems or scaled up to meet the needs of massive data centers.
To meet the demands of new computer systems, the team had to overcome three challenges: supplying power to smaller areas to allow the microprocessors to sit closer and closer to each other; operate at high efficiency both to reduce costs and to avoid overheating; and switch power between components at lightning speed to meet the demands of microprocessors.
“Google and Intel originally asked, ‘How can you deliver 10 times more power to a square millimeter of silicon without sacrificing speed or efficiency?'” Chen said.
The research team accomplished all three by redesigning everything from power distribution components to their architecture and control. They used capacitors instead of the traditional method of magnetism to process energy, and they built the systems vertically instead of traditional horizontal construction. Both features introduced significant design challenges, but once the researchers resolved them, they were able to deliver a superior system.
“We tested efficiency up to full power, and we tested dynamics,” Chen said. “It’s a fully functional system 10 times smaller than the best on the market.”
The article, “Vertically Stacked LEGO-PoL CPU Voltage Regulator”, was posted online on December 14, 2021.
Researchers seek to disrupt the energy and healthcare industries with a miniature pulsed power system
Jaeil Baek et al, Vertical Stacked LEGO-PoL CPU Voltage Regulator, IEEE Transactions on Power Electronics (2021). DOI: 10.1109/TPEL.2021.3135386
Provided by Princeton University
Quote: Vertical Thinking Broke the Bottleneck in Powering High-Performance Computers (May 2, 2022) Retrieved May 3, 2022 from https://techxplore.com/news/2022-05-vertical-broke-bottleneck- power-high-performance.html
This document is subject to copyright. Except for fair use for purposes of private study or research, no part may be reproduced without written permission. The content is provided for information only.