THE CRUNCH

NVIDIA has announced Rubin HPC, a new GPU in the Vera Rubin platform aimed at accelerating simulation and data processing for science, engineering and analytics. Unveiled at SC26, the chip is pitched at a changing research workflow: as AI agents help scientists and engineers explore more candidate materials, molecules and designs, those ideas still need to be tested through GPU-accelerated simulation before anyone commits to experiments or physical prototypes.

On the specifications front, Rubin HPC carries what NVIDIA calls a record number of CUDA cores, more than 50,000 on a single GPU, alongside a sparse data acceleration capability designed to keep those cores fed with data. GPU-to-GPU connectivity runs over seventh-generation NVLink interconnect technology.

The design philosophy is balance rather than raw compute. NVIDIA says the GPU scales data movement and compute together, because scientific workloads, from climate models to genomic analysis and financial risk, tend to reach into vast datasets, gather scattered values and repeat calculations at enormous scale. Keeping the cores supplied with data is meant to accelerate recurring patterns such as sparse matrix operations, stencil computations, sorting and reductions.

NVIDIA's CUDA-X libraries are positioned as the software counterpart, offering optimised implementations of simulation and analytics algorithms that agents can invoke. Faster agent-driven simulation, the argument goes, shortens the loop of hypothesis, simulation and refinement, letting researchers point lab work at the most promising candidates sooner. Ian Buck, NVIDIA's vice president of hyperscale and high-performance computing, framed it as faster simulation meaning faster discovery. Rubin HPC is expected to be available in 2028.