US Labs Test Brain-Like Chips for New Supercomputers

US national labs are exploring new computer chips inspired by the human brain. This is a big change from the computer chips they usually use.
By Newsroom | Technology, United States |

A Shift From Familiar Architectures

US national labs are exploring unconventional computing architectures for their next-generation supercomputers, potentially moving beyond the prevalent Graphics Processing Units (GPUs) that have dominated high-performance computing. This exploration signifies a move toward specialized, potentially neuromorphic, hardware for advanced computational tasks.

A startup named NextSilicon has received endorsement from Sandia National Laboratories for its high-performance computing accelerators. Details are sparse, but this collaboration points towards the kind of "exotic" technology the government is considering.

Brain-Inspired Computing Emerges

Sandia National Laboratories, located in New Mexico, has quietly activated a new computing system inspired by the human brain. This system, known as SpiNNaker 2, operates without traditional operating systems or disk storage.

  • The SpiNNaker 2 architecture relies on event-driven computation and parallel processing, aiming to mimic the human brain's structure.

  • Each chip boasts 152 cores and specialized accelerators.

  • A fully configured system can house a massive amount of DRAM—up to 138,240 terabytes.

  • This approach contrasts sharply with conventional supercomputers that depend on GPUs and centralized storage.

The long-term efficacy and real-world application of such brain-inspired systems for critical national security projects remain subjects of ongoing observation.

Background

The pursuit of ever-increasing computational power is a constant in scientific research and national security. Traditional supercomputing has relied heavily on the parallel processing capabilities of GPUs. However, the push for more efficient and specialized architectures is leading institutions like Sandia to investigate novel designs that deviate from the norm. The potential advantages of these new architectures, such as those seen in neuromorphic computing, include energy efficiency and the ability to handle specific types of complex problems in ways that traditional hardware struggles with.

Read More: Computer Security Expert Peter Neumann Dies at 93 in California

Frequently Asked Questions

Q: Why are US national labs looking at new kinds of computer chips?
US national labs want to build better supercomputers. They are testing chips inspired by the human brain, which are different from the usual computer chips called GPUs.
Q: What is special about the new brain-inspired chips?
These chips use a design that works like the human brain, with many small parts working together. One system, SpiNNaker 2, has 152 cores per chip and can store a lot of data, unlike normal computers.
Q: Who is working with these new chips?
Sandia National Laboratories in New Mexico is testing these new chips. A company called NextSilicon has also been approved by Sandia for its new computer accelerators.
Q: What is the goal of using these brain-inspired chips?
The goal is to create more powerful and efficient supercomputers for scientific research and national security tasks. These new chips might be better at solving certain complex problems than current technology.
Q: When will these new chips be used in supercomputers?
The article does not state a specific date for when these new chips will be fully used in supercomputers. It mentions that the labs are exploring these options and that their long-term use is still being studied.