logo
#

Latest news with #MaxwellLabs

Sandia Labs director: Nation needs to overcome energy barriers, embrace AI
Sandia Labs director: Nation needs to overcome energy barriers, embrace AI

Yahoo

time17-05-2025

  • Business
  • Yahoo

Sandia Labs director: Nation needs to overcome energy barriers, embrace AI

May 17—When you think of a laser, do you picture it heating something up? That's not the case for an experimental laser-based photonic cooling system that Sandia National Laboratories is helping test. It's one of a few methods new Labs Director Laura McGill envisions will help the nation address intensive energy demands required for artificial intelligence and data centers. McGill, who previously worked in nuclear defense for the lab and stepped into her new role at the beginning of May, said in a Wednesday interview that the nation needs to embrace AI, a rapidly advancing field of computer science allowing for the creation of products like personalized chatbots. "That's one of the advantages of being at Sandia because we work both sides of the problem," McGill said. "We're working on the exquisite AI algorithms, but we also have people who are working on how do we make the energy more efficient?" She said there are three ways to solve the energy demand barriers: algorithm designs, high power computing infrastructure and neuromorphic processing. On algorithms, McGill said there are plenty of experts at Sandia experimenting with design techniques to make AI energy usage more efficient. Similarly, computing professionals with access to state-of-the-art resources at the lab are working on the infrastructure side of things. That's where she brought up the laser technology Sandia is helping Minnesota-based Maxwell Labs Inc. develop to cool computers. Sandia announced the work last month, explaining that a particular light frequency matched with a specific element causes lasers to cool rather than heat things. "AI is still kind of new," McGill said. "I think there's a lot of opportunity in that space to see how we're pulling on all that energy." Neuromorphic processing isn't really directly applicable to AI work, McGill acknowledged, but there are a lot of computing areas that replicate the way the brain works and takes less power to run — something experts can study for potential applications in the AI field. On top of the intensive energy demands, high-tech systems like large training models or data centers take a lot of water. It's an especially prominent challenge in a desert like the Southwest. McGill said it's important to build these large data warehouses and storage facilities in places where there's water, but added that's not an area Sandia has as much expertise in. "But we certainly have people that pay attention to the use of the natural resources and how we preserve those as best we can," she said. Looking at the big picture, McGill said she wants her tenure to be defined as Sandia very quickly responding to the nation's needs. "I think we lost a little bit of that during the Cold War, when the need for the nuclear deterrent went away. So we went more into more of what we call the stewardship of the stockpile," she said. "So now we're in a mode where we need to pick up our ability to accelerate. "And that's where my background helps because I worked in (the) commercial industry, and I'm familiar with the digital tools and different ways that you can align the workforce to be able to work more efficiently, and especially in delivering really complex systems."

US scientists toy with lasers to cool computer chips, bring down data center costs
US scientists toy with lasers to cool computer chips, bring down data center costs

Yahoo

time14-04-2025

  • Science
  • Yahoo

US scientists toy with lasers to cool computer chips, bring down data center costs

A high amount of the energy used by data centers is for cooling purposes. Data centers handle everything from basic tasks like email processing to complex operations like AI training. However, these systems face a major challenge with the heat generated by the servers. The current air and water-based cooling systems are struggling to keep up. As Sandia National Laboratories physicist Raktim Sarma explains, "About 30 to 40 percent of the energy data centers use is spent on cooling. In some communities, the amount of water needed can strain local resources." Now, a Minnesota-based startup, Maxwell Labs, is teaming up with Sandia and the University of New Mexico to explore laser-based photonic cooling. This new technology seeks to control computer chip temperatures, lower energy use, and improve efficiency compared to current methods. 'A successful project will not only address the immediate need for energy savings but also pave the way for processors to operate at performance levels that were previously thought impossible,' said Mike Karpe, Maxwell Co-Founder and Chief Growth Officer. Photonics, the science of harnessing light, is already being used for data processing, national security, and communication. But the team at Sandia and Maxwell says they are the first to explore its potential for cooling computers. Despite their common use for heating, lasers can also produce a cooling effect in specific situations. This happens when a precise light frequency interacts with a highly pure target of a specific element. The team describes it as using lasers to hold individual atoms at incredibly low temperatures in quantum computers. The laser cooling method might work for GPUs if the cooling light targets small, hot areas. 'We really only have to cool down spots that are on the order of hundreds of microns, about the size of a speck of dust,' Sarma noted. Maxwell Labs envisions a photonic cold plate, a light-based alternative or complement to existing water and air cooling systems. A key element in cold plate design is gallium arsenide, a semiconductor similar to silicon. For the laser cooling to work effectively, the gallium arsenide needs to be incredibly pure and grown in extremely thin, crystalline layers known as epitaxial layers. This innovative design utilizes materials and microscopic features, thousands of times smaller than a human hair. The technique allows to channel cooling laser light directly to those critical hot spots. According to the company's models, laser cooling can achieve lower chip temperatures than water cooling. 'This will enable novel energy-recovery paradigms not possible with traditional cooling technology,' said Maxwell CEO Jacob Balma. If successful, this technology could unlock significant benefits. Cooler chips can operate at higher speeds without overheating, leading to improved performance and greater power efficiency. Furthermore, it might be possible to recapture the light generated as heat and transform it back into electrical power. The bright idea of using lasers to cool computers might just be the key to unlocking the next generation of high-performance, energy-conscious computing.

DOWNLOAD THE APP

Get Started Now: Download the App

Ready to dive into the world of global news and events? Download our app today from your preferred app store and start exploring.
app-storeplay-store