In an effort to meet the rising energy demands of data centers, engineers at the University of California San Diego have developed a new chip design that could improve how graphics processing units (GPUs) convert and manage power. The technology demonstrates a more efficient way to perform a critical task in electronics: converting high voltages into lower levels required by computing hardware. In lab tests, a prototype chip performed the type of voltage conversion used in modern data centers with high efficiency.
The advance, published in Nature Communications, could lead to the development of smaller, more energy-efficient systems for advanced computing.
The chip design offers a new approach to improving the performance of a circuit component known as a DC-DC step-down converter, which is found in nearly all electronics. The step-down converter acts as a protective bridge between power sources and sensitive circuits. It transforms a high input voltage into the lower voltage each component in the circuit precisely needs to operate safely. For example, data centers often distribute power at 48 volts, while processors in GPUs need much lower voltages, typically between 1 and 5 volts.
However, converting between these levels efficiently, and within limited space, has become increasingly difficult as computing demands grow.
Traditional step-down converters, for instance, lose efficiency and struggle to deliver enough current when the gap between input and output voltage is large. Most step-down converters rely on magnetic components such as inductors, which, while effective, are approaching their physical performance limits and are growing difficult to scale further. “We’ve gotten so good at designing inductive converters that there’s not really much room left to improve them to meet future needs,” said study senior author Patrick Mercier, professor in the Department of Electrical and Computer Engineering at the UC San Diego Jacobs School of Engineering.
To address this challenge, Mercier and members of his research group, including study first author Jae-Young Ko, an electrical and computer engineering Ph.D. student at UC San Diego, explored a promising alternative: piezoelectric resonators, which are tiny devices that store and transfer energy through mechanical vibrations. Piezoelectric-based converters could potentially be smaller, more energy dense, more efficient and easier to manufacture at scale. “They have a lot of room to grow and have the potential to deliver better performance than anything that’s come before them,” Mercier said.
However, early versions of piezoelectric-based converters have struggled to maintain efficiency and deliver enough power when handling large voltage differences.
In this study, the team developed an improved step-down converter that combines a piezoelectric resonator with small, commercially available capacitors arranged in a strategic way. This new circuit design allows the converter to handle larger voltage conversions more effectively. The team implemented the design in a prototype chip. In tests, it converted 48 volts down to 4.8 volts — a level commonly required in data centers — with a peak efficiency of 96.2 percent. The chip also delivered about four times more output current than earlier piezoelectric-based designs.
Here is an exclusive Tech Briefs interview, edited for length and clarity, with Mercier.
Tech Briefs: What was the biggest technical challenge you faced while developing this new chip design?
Mercier: I think the biggest technical challenge was that this was kind of a brand-new type of converter architecture using piezoelectric resonators instead of inductors. And so the way that you construct the topology, the way that you run the switches, the way that you control things is very different than what we’re used to with inductive-based converters. That's a challenge to start with.
Then, one of the big innovations that we had in this work was adding capacitors — flying capacitors in an interesting, new manner to help reduce the load and the stress on the piezoelectric resonator itself in order to allow for more power to be delivered to the output more efficiently.
Tackling all of those challenges at once was, I would say, the biggest hurdle that we faced.
Tech Briefs: Can you please explain in simple terms how it works?
Mercier: Basically, this is using a piezoelectric resonator, which stores energy between the mechanical and the electrical domains. And it's a resonator, so you put a little bit of energy in and it's going to want to resonate between the mechanical domains. The piezoelectric resonators themselves have a very high quality factor, so you ping them with just a little bit of energy and they’ll resonate or oscillate for quite a long time before eventually decaying.
That's really good. That means there's not a lot of losses within the piezo resonator itself. Basically, what we do is we switch the piezo resonator in a very specific order to put energy in at a certain voltage and take energy out at a different voltage. And because of the high quality factor of the resonator, that allows us to do so with very high efficiency.
Tech Briefs: Professor, you’re quoted in the article I read as saying, ‘We need to continue to improve on multiple areas — materials, circuits and packaging — to make this technology ready for data center applications.’ My question is: Do you have any set plans for further research work, etc.? If not, what are your next steps?
Mercier: We’re continuing to work in this area. We are investigating other materials. We’re investigating making resonators smaller and operate at higher frequencies, which allows for additional power to be processed by them. And we’re also continuing to innovate on topologies as well.
Tech Briefs: Do you have any advice for researchers aiming to bring their ideas to fruition?
Mercier: What I like to try and do is tackle the high-hanging fruit. What I mean by that is the problems that are really difficult and are facing big headwinds in industry in general — and try unique approaches. Right now, piezoelectric resonators have, as we mentioned, quite a few challenges associated with them, but they have great opportunity, right? If we can solve all of those challenges, there may be a path forward where these become kind of the standard for high-efficiency power delivery.
That's the advice that I give to my graduate students and people getting into research: Research is a time to be risky, and you want to go for those big ideas. Sometimes they won't work out, but when they do, it makes a big splash. That's very rewarding.
