Tuesday, November 17

Nvidia Rubin HPC GPU Features Over 50,000 CUDA Cores


Credit: Nvidia

Nvidia has introduced Rubin HPC, a new graphics processing unit designed for high-performance computing, scientific research and advanced data analysis. The chip features more than 50,000 CUDA cores, almost 2.5 times the 21,760 cores found in the GeForce RTX 5090.

Announced at the SC26 supercomputing conference, Rubin HPC is part of Nvidia’s Vera Rubin platform. Unlike consumer graphics cards, it is built primarily for demanding scientific and engineering workloads rather than gaming.

According to Nvidia, the GPU is designed to accelerate complex simulations used in climate modelling, genomic research, engineering and financial analysis. It also features sparse data acceleration technology, intended to reduce delays caused by data movement and keep its large number of computing cores working efficiently.

Another key addition is seventh-generation NVLink, which enables faster communication between GPUs in large computing systems. Rubin HPC will also support Nvidia’s CUDA-X libraries, providing optimized tools for scientific calculations and data processing.

Credit: Nvidia

Despite the impressive core count, Nvidia has not yet revealed several important specifications, including memory capacity, bandwidth, power consumption or pricing. As a result, its real-world performance will need to be assessed once independent benchmarks become available.

Nvidia expects Rubin HPC to arrive in 2028. The new accelerator could become an important component of future supercomputers, particularly as AI-powered research increasingly depends on large-scale simulations and complex scientific calculations.

Leave a Reply

Your email address will not be published. Required fields are marked *