Skip to main content
Back to timeline
NVIDIA BlogSource publication:

NVIDIA unveils Rubin HPC GPU: over 50,000 CUDA cores per card to speed simulations for science agents

Synopsis

NVIDIA announced Rubin HPC at SC26, a new GPU in the Vera Rubin platform aimed at simulation and data processing for science, engineering and analytics workloads. The card carries more than 50,000 CUDA cores on a single GPU plus a sparse data acceleration capability that keeps those cores fed, and NVIDIA says the design scales data movement and compute together rather than raising compute alone. The product is expected to be available in 2028.

Source-provided article image: NVIDIA Announces Rubin HPC — A New GPU to Accelerate Agents for Science, Engineering and Analytics
NVIDIA Research

Interpretation

NVIDIA is offering a GPU purpose-built for scientific computing, treating simulation speed as the bottleneck when agents explore materials, molecules and designs. Scientific computing workloads previously had to balance massive data reads and computation on the same GPU, so more compute did not necessarily yield proportional simulation speedup. NVIDIA's stated comparison is a balanced design that scales data movement and compute together instead of increasing compute alone, so cores spend time computing rather than waiting.

Rubin HPC feeds its compute units through more than 50,000 CUDA cores on a single GPU plus a sparse data acceleration capability. Sparse data acceleration is the newly added capability NVIDIA describes for keeping those cores supplied with data. NVIDIA names sparse matrix operations, stencil computations, sorting and reductions as recurring patterns across climate models, engineering simulations, genomic analysis and financial risk.

Perspective

The report covers the product positioning, hardware features and expected 2028 availability NVIDIA announced at SC26, for research and engineering readers tracking scientific computing hardware.

NVIDIA did not publish measured speedups against existing GPUs or detail how sparse data acceleration is implemented; the 2028 availability is an expectation, not a commitment.