TellWorldAI World Archive
Ask中文

Home›Compute

Chip

NVIDIA Rubin HPC

NVIDIA Rubin HPC is a GPU in the Vera Rubin platform, announced at SC26 for simulation and data processing in science and engineering, with more than 50,000 CUDA cores and an expected 2028 availability.

Type
GPU
Developer
NVIDIA

What it is

Rubin HPC is a graphics processing unit (GPU) that NVIDIA announced at the SC26 conference. It is part of the Vera Rubin platform and is designed for simulation and data processing in science, engineering and analytics.[1] HPC in its name stands for high-performance computing, the field that runs large simulations on supercomputers.

Origin

NVIDIA presented Rubin HPC as a new GPU for the kinds of workloads that supercomputers run, with a focus on accelerating agents that work in science and engineering.[1] It belongs to the same Vera Rubin family that NVIDIA is building for data centers, but it is aimed at scientific computing in particular.

Main features

NVIDIA says the Rubin HPC GPU has more than 50,000 CUDA cores on a single GPU. The company calls this a record number of CUDA cores.[1] CUDA cores are the parallel processing units inside an NVIDIA GPU that carry out many calculations at the same time.

The GPU includes sparse data acceleration, a feature that speeds up work on data where most values are zero. It also provides GPU-to-GPU connectivity through seventh-generation NVLink, the high-speed link that lets GPUs exchange data quickly.[1]

NVIDIA says Rubin HPC is designed to scale data movement and compute together. Its target workloads include sparse matrix operations, stencil computations, sorting and reductions.[1] A stencil computation updates each point in a grid using the values of its neighbors, which is common in physics simulations. A reduction combines many values into a single result, such as a sum or a maximum.

Many scientific programs are limited less by arithmetic than by waiting for data to arrive from memory. Pairing compute with data movement in one GPU design targets exactly this kind of bottleneck, which is why NVIDIA describes the two as scaling together.

Place in development

NVIDIA says Rubin HPC is expected to be available in 2028.[1] Since the product is not yet available, the figures above are NVIDIA's design targets, not measured results from deployed systems.

CapabilitiesScience

Ecosystem & regionHardware & compute

20261

  1. ReleaseNVIDIA announces Rubin HPC GPU for science workloads at SC26

    On November 16, 2026, NVIDIA announced the Rubin HPC GPU at the SC26 conference, positioning it for the simulations that AI agents use to test ideas in science and engineering.

Developed by1

NVIDIA

Cite this page: NVIDIA Rubin HPC. TellWorld, https://www.tellworld.net/compute/nvidia-rubin-hpc. Updated 2026-11-18