· Desk editor: Daniel Mercer · · Single source ·Updated

Nvidia unveils Rubin HPC GPU at Supercomputing 2026

Nvidia announced the new Rubin HPC GPU at the Supercomputing 2026 conference, stating it will arrive in 2028.

Nvidia announces Rubin HPC GPU with 'record number of CUDA cores' at Supercomputing 2026
File photo Nvidia announces Rubin HPC GPU with 'record number of CUDA cores' at Supercomputing 2026 Photo: Tom's Hardware

Chip features core count

The company revealed a chip with over 50,000 CUDA cores designed to accelerate agentic workflows. This specific model contains 224 streaming multiprocessors and 28,672 CUDA cores on the package. The high core count implies a layout similar to the four-die Rubin Ultra GPU configuration. Nvidia did not share exact physical layout details or HBM capacity figures alongside the announcement.

Interconnect and bandwidth specs

Data movement between these GPUs will use the NVLink 7 interconnect technology for the system. Past disclosures indicate this link retains 3.6 TB/s of per-GPU bandwidth found in the previous generation. The company did not share new details regarding the physical layer or scale-up domain size within the rack. Analysts note that the chip targets scientists and engineers using large data sets for simulation.

Targeted applications and use

Nvidia expects faster simulation performance to help researchers in climate research, genomic analysis, and financial simulation fields. Users can leverage CUDA-X domain-specific libraries to tighten the reasoning loop from hypothesis to refinement. The balanced design aims to minimize idle compute resources while feeding massive data sets typical of those workloads.

Reported by one outlet

Only one outlet has published this. Nothing here has been checked against a second report, so read it as that outlet's account and follow the link below for the original.

Reported by

1 independent outlet. Headline as published. Links open the original report.