The National Renewable Energy Laboratory has installed 132 graphics processing unit nodes to its Kestrel supercomputer, completing the system and boosting its computing power to 44 petaflops.
The additional Nvidia GPUs will combine their power with the existing 2,314 central processing unit nodes, powered by Intel Sapphire Rapids processors, to advance ongoing research efforts on renewable energy, NREL said.
The GPU nodes will enable artificial intelligence and machine learning capabilities to support the Office of Energy Efficiency and Renewable Energy’s research programs, the laboratory added.
With the Hewlett Packard Enterprise-built Kestrel completed, numerous projects started using the system’s GPU muscle, which Derek Vigil-Fowler, principal investigator of the Beyond-DFT Electrochemistry with Accelerated and Solvated Techniques project at NREL, described as “extremely powerful.”
Vigil-Fowler noted that the GPUs have demonstrated excellent efficiency and scaling for simulating electrocatalytic systems, which aims to design better catalysts for water electrolysis, fuel cells and carbon dioxide reduction.
According to Kristin Munch, Kestrel’s project manager, the high-performance machine gives researchers the computing capabilities to perform research at a pace and scale that could support the envisioned energy transition for the United States.
NREL said Kestrel delivers more than five times the computing power and over two times the efficiency of its previous supercomputer, Eagle.