NVIDIA unveils Rubin HPC GPU for science and engineering workloads
Reported by DatacenterDynamics, VideoCardz.com, NVIDIA and 6 more outlets11 sources ↓
NVIDIA announced Rubin HPC at SC26 on Nov. 16, describing it as a GPU for science, engineering and analytics. The company says it has more than 50,000 CUDA cores and expects it to be available in 2028.
Key points
- NVIDIA says Rubin HPC has more than 50,000 CUDA cores on one GPU
- It is designed for simulation and data processing, including sparse calculations
- The GPU uses seventh-generation NVLink for GPU-to-GPU connections
- Availability is expected in 2028
- NVIDIA has not disclosed memory capacity, bandwidth, FP64 performance or price
Sources
- DatacenterDynamicsRead ↗Nvidia unveils Rubin HPC GPU for agentic AI-powered science and engineering workloads
- VideoCardz.comRead ↗NVIDIA announces Rubin HPC GPU with over 50,000 CUDA cores and NVLink 7, launching in 2028
- NVIDIARead ↗NVIDIA Announces Rubin HPC — A New GPU to Accelerate Agents for Science, Engineering and Analytics
- Traders UnionRead ↗NVIDIA unveils Rubin HPC GPU for science and engineering workloads
- The DefiantRead ↗Nvidia Unveils Rubin HPC GPU for Scientific Computing
Details
NVIDIA describes Rubin HPC as part of its Vera Rubin platform, with sparse-data acceleration intended to keep its cores supplied with data. The workloads it names include sparse matrix operations, stencil computations, sorting and reductions, used in areas such as climate modeling, engineering simulation, genomic analysis and financial risk.
Tom’s Hardware reports that NVIDIA did not disclose the GPU’s physical layout or memory capacity and bandwidth. The company also has not published FP64 performance or pricing; the announcement gives no measured results for the targeted workloads.
Written from coverage as of
Images
- NVIDIA
- Tom's Hardware
- Unite.AI
- Tech Times
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh