Uncle Sam's next big supercomputer might use something more exotic than GPUs

Nine of the top 10 most powerful supercomputers rely on GPUs, but scientific computing needs ultra-precise floating point performance, especially FP64. Chipmakers prioritize AI-focused throughput, pushing National Labs to explore new architectures that better support FP64 calculations. NextSilicon’s Maverick-2 is a dataflow processor designed for 64-bit floating point mathematics used in Department of Energy simulations. The Department of Energy runs major public supercomputers supporting work ranging from nuclear weapons physics and bioweapons defense to public health and safety. Sandia’s Spectra supercomputer, built with Penguin Solutions and NextSilicon, uses 64 nodes and 128 Maverick-2 runtime-configurable accelerators as a test bed. Sandia reported system acceptance requirements were met, enabling future deployment in larger systems. Maverick-2 uses a reconfigurable dataflow architecture rather than the von Neumann compute model used in most CPUs and GPUs.

"Of the world's most powerful supercomputers, nine of the top 10 are powered by GPUs, but that might not be the case for much longer. As chipmakers like Nvidia prioritize AI FLOPS over the ultra-precise floating point calculations used in scientific computing, US National Labs are turning to new chip architectures to get their FP64 fix."

"Among the candidates is NextSilicon's Maverick-2, a dataflow processor designed explicitly with the 64-bit floating point mathematics that dominate the Department of Energy's most important simulations. Despite its name, the Department of Energy is concerned with far more than the US' power grid. It operates some of the largest publicly known supercomputers in the world, which are responsible for everything from simulating the physics of nuclear weapons at the moment of criticality and bioweapons defense to public health and safety."

"But that's not the case for Sandia National Laboratory's new Spectra supercomputer, which was built in collaboration with Penguin Solutions and NextSilicon. Compared to exascale systems like Frontier or El Capitan, Spectra is tiny. The machine counts 64 nodes and 128 of NextSilicon's "runtime-configurable" accelerators. But scale isn't the point. Spectra is a test bed for NextSilicon's Maverick-2."

"Despite some similarities to Nvidia's B200, Maverick-2 is a very different beast. Instead of the standard von Neumann compute architecture that underpins most CPUs and GPUs today, NextSilicon's chips employ a reconfigurable dataflow architecture. The processor's two compute dies comprise a grid of arithmetic logic units interconn"

Read at theregister