API
< Back to all clusters
[TECHNOLOGY] · United States · 17 sources

started · updated

Nvidia unveils Rubin HPC GPU with over 50,000 CUDA cores

At the Supercomputing 2026 (SC26) conference, Nvidia announced the Rubin HPC, a new GPU within the Vera Rubin platform specifically designed for high-performance computing (HPC), scientific simulations, and large-scale data analytics. Unlike standard Rubin GPUs optimized for AI training and inference, the Rubin HPC features more than 50,000 CUDA cores, a significant increase over previous models.

The architecture focuses on a balanced design that scales compute power alongside data movement capabilities to prevent cores from idling while waiting for data. Key technical features include seventh-generation NVLink interconnect technology and specialized sparse data acceleration to improve efficiency in workloads such as climate modeling, genomic analysis, and financial risk assessment.

Nvidia expects the Rubin HPC to be available in 2028. In related hardware news, the company also showcased the Vera-based rack-scale platform, which utilizes liquid cooling and integrates 192 Vera processors to achieve an aggregate peak memory bandwidth exceeding 200 TB/s. Additionally, Nvidia reported third-quarter fiscal 2027 revenue of $115.8 billion, representing a 103% increase from the previous year.

Entities

Dell Technologies · HPE · Ian Buck · Lenovo · Nvidia · Rubin HPC · SC26 · Supermicro · Vera Rubin · Vera Rubin platform · Vera Rubin platform

Claims

What the coverage asserts, and how many sources carry each claim.

Sources