The Next Generation: NERSC-10

The NERSC-10 project is designed to deliver a next-generation supercomputer in the 2026 timeframe for the DOE Office of Science (SC) research community.

A simulation of a white dwarf merger. (Credit: Max P. Katz et al)

With a new era for high performance computing (HPC) upon us, along with the convergence of simulations and modeling, advances in artificial intelligence (AI), and an explosion of experimental and observational data, there is a critical need to integrate these capabilities to enable new modes of scientific discovery.

Funded by the DOE SC’s Office of Advanced Scientific Computing Research (ASCR), NERSC is the mission HPC facility for the Office of Science, uniquely supporting the needs of science across the entire office. ASCR’s mission is to discover, develop, and deploy computational and networking capabilities to analyze, model, simulate, and predict complex phenomena important to the DOE.

The NERSC community is currently experiencing a transition with the emergence of energy-efficient GPUs, specialty processors, and other components. At the same time, new software technologies are emerging, inspired in part by the needs of data-driven workflows that couple HPC and experimental facilities in real time. These technology changes, along with new science use cases requiring more flexible, programmable, and resilient HPC systems, present new opportunities for mapping computational tasks and workflow components to hardware that might be more efficient and faster for a given problem.

The NERSC-10 system is intended to leverage these new technologies and support the emerging needs in AI and experimental and observational science by accelerating end-to-end DOE SC workflows and enabling new modes of scientific discovery through the integration of experiment, data analysis, and simulation.

We expect the NERSC-10 supercomputing system to push the boundaries of scientific computing by being

• The first configurable and programmable supercomputer that allows users to assemble and compose the set of capabilities that work best for their research project,
• The first supercomputing system designed to support urgent and interactive workflows from DOE SC’s experimental and observational facilities, and
• The first HPC system to provide seamless performance across converged networks, accelerators, and storage components.

The NERSC-10 design is also expected to

• Feature world-leading AI capabilities and support the integration of simulations, AI, and data analysis, including the interactive exploration of very large models,
• Leverage container technologies to create a highly customized environment for a wide variety of science use cases, and
• Enable new compute, storage, and software to be easily integrated into a comprehensive, extensible HPC system for science.