Dell Is Building Next DOE Supercomputer, Powered by Nvidia

The U.S. Department of Energy has commissioned Dell to deliver its next supercomputer, expected to come online in 2026. Referred to as Doudna, in honor of the Nobel Prize-winning biochemist Jennifer Doudna, it is also known as NERSC-10 for its home at the DOE’s National Energy Research Scientific Computing Center at Lawrence Berkeley National Laboratory in Berkeley, California. Powered by Nvidia’s new Vera Rubin platform, Doudna will be optimized for AI workloads and aims to deliver a greater than tenfold speed boost over NERSC’s current flagship machine, Perlmutter, while using only 2-3x the power.

Doudna is notable “for its technology choices, which indicate the growing desire for government labs to adopt more technologies from commercial AI systems,” writes The New York Times, pointing out that “Nvidia chips, though widely used by big cloud companies as well as in supercomputers, were passed over by the Energy Department for three previous record-setting machines.”

Dell, NYT notes, “has hardly been a player in the highest end of the supercomputer market, but it has had success in large commercial AI installations.” Hewlett Packard Enterprise and its Cray subsidiary have been major players in the DOE space, working with Intel and AMD.

The new supercomputer “will harness Nvidia’s cutting-edge Vera Rubin chips, paired with Dell’s advanced liquid-cooled server technology,” writes WebProNews, noting the combination “promises unprecedented computational power, tailored to meet the DOE’s demanding scientific and AI-driven workloads, ranging from climate modeling to drug discovery.”

“Doudna is tailored for real-time discovery across the Department of Energy’s most urgent scientific missions,” Nvidia explains in a blog post.

While The Register acknowledged “we still don’t know much” about the next-gen Vera Rubin superchips Nvidia CEO Jensen Huang teased at GTC in March, “we do know Nvidia’s latest batch of Blackwell Ultra accelerators sacrificed double-precision performance, long considered essential for scientific computing, in favor of increased use of 4-bit precision formats tailored to AI workloads.”

Earlier this year, Fast Company wrote that the architecture would combine “Nvidia’s first custom-designed CPU, ‘Vera,’ and two new ‘Rubin’ GPUs” together delivering “more than double the inference performance of Blackwell” to meet the demands of reasoning models.

“AI is the Manhattan Project of our time, and Doudna will help ensure America’s scientists have the tools they need to win the global race for AI dominance,” Secretary of Energy Chris Wright said in a DOE announcement.

No Comments Yet

You can be the first to comment!

Leave a comment

You must be logged in to post a comment.