Mission supercomputer adds NVIDIA Vera CPUs with 12x performance

In a significant advancement for high-performance computing, Los Alamos National Laboratory has completed the installation of NVIDIA's latest Vera CPUs, which are based on the innovative Grace Hopper architecture. This upgrade, finalized in the fourth quarter of 2023, replaces the existing processors across multiple compute nodes, marking a substantial enhancement in computational capabilities. While specific performance metrics such as floating-point operations per second (FLOPS) and memory bandwidth were not disclosed, laboratory officials confirmed that the new processors deliver a remarkable 12-fold increase in performance when benchmarked against typical mission workloads.

Details of the CPU Upgrade

The Vera CPUs are engineered to handle the most demanding computational tasks, including large-scale simulations essential for nuclear weapons physics and climate modeling. The Grace Hopper architecture is notable for its integration of high-bandwidth memory and a unified memory space, which facilitates faster data transfers between processors and accelerators. This design choice aligns with an industry-wide shift towards heterogeneous computing, where central processing units (CPUs) and graphics processing units (GPUs) collaborate to maximize throughput. The laboratory officials have validated the impressive 12x performance gain through a series of benchmark tests that reflect the laboratory's typical operational workloads.

In addition to performance enhancements, the new Vera processors also boast improved energy efficiency, allowing for more computations per watt compared to the previous system. This upgrade is part of Los Alamos's ongoing strategy of periodic hardware refreshes, a practice that has historically involved collaborations with leading chip manufacturers to maintain its position at the forefront of computational science.

Historical Context of Supercomputing at Los Alamos

Los Alamos has a long-standing history of operating high-performance computing systems, primarily for nuclear stockpile stewardship simulations under the U.S. Department of Energy's Advanced Simulation and Computing program. The laboratory's previous supercomputers, such as the Roadrunner, which was the first to achieve petaflop performance in 2008, and Trinity, a Cray-built machine that was among the fastest when it was introduced in 2016, underscore its commitment to maintaining cutting-edge technology. The Vera CPU upgrade continues this tradition of approximately four-to-five-year refresh cycles, ensuring that the laboratory remains at the leading edge of computational science.

Interestingly, the naming of the Vera CPU may pay homage to Vera Rubin, the astronomer known for her pivotal work on galaxy rotation curves that provided critical evidence for dark matter. While the laboratory has not officially confirmed any connection, this aligns with NVIDIA's trend of naming processors after notable scientific figures.

As global competition in supercomputing heats up, highlighted by China's LineShine supercomputer being named the fastest in the world in June 2026, the need for sustained upgrades at U.S. national laboratories becomes increasingly critical. This competitive landscape emphasizes the importance of continual innovation to maintain technological leadership.

Impact on Research and National Security

The upgraded supercomputer at Los Alamos will serve a variety of classified and unclassified applications, ranging from weapon physics simulations to materials science, climate modeling, and the development of advanced energy technologies. A spokesperson from the laboratory noted that this upgrade allows them to tackle problems that were previously deemed intractable, although specific details were not disclosed. The 12x performance gain is anticipated to significantly reduce simulation times from weeks to mere days, thereby accelerating research timelines for crucial national security projects.

This upgrade aligns with broader federal initiatives aimed at enhancing AI and computing capabilities, such as the "Genesis Mission" AI initiative launched by President Trump in November 2025, which was likened to the historical Manhattan Project. Furthermore, NVIDIA has been advocating for the renewal of the National Quantum Initiative, emphasizing the importance of quantum-classical convergence for future supercomputing advancements. The Vera CPUs at Los Alamos are expected to provide a platform for integrating quantum acceleration as these technologies evolve.

Conclusion: Next Steps in Supercomputing

The installation of the Vera CPUs positions Los Alamos to meet the increasing computational demands that are paving the way toward exascale performance. While the laboratory has not disclosed any plans for further system expansions or additional upgrades, the integration of Vera CPUs signifies an ongoing collaboration between Los Alamos and NVIDIA that is likely to continue. Looking ahead, NVIDIA's introduction of NVQLink, an open system architecture designed to connect quantum processors directly with GPU-based supercomputers, could significantly influence future upgrades at national laboratories. As the fields of artificial intelligence and high-performance computing converge, the necessity for systems capable of managing both extensive simulations and AI workloads will grow, ensuring that Los Alamos retains a competitive edge in the evolving landscape of supercomputing.