Nvidia unveils Rubin HPC GPU with more than 50,000 CUDA cores for science workloads
Nvidia has announced Rubin HPC, a new GPU in its Vera Rubin platform designed to speed up simulation and data processing for science, engineering and analytics, the company said in a blog post at the SC26 supercomputing conference. The chip has more than 50,000 CUDA cores on a single GPU, a record for Nvidia, plus a sparse data acceleration capability and seventh-generation NVLink for GPU-to-GPU connections.
Nvidia pitches the chip at AI agents that help scientists and engineers explore more candidate materials, molecules and designs, which in turn rely on simulations to test those ideas. Rather than adding compute alone, Rubin HPC uses what Nvidia calls a balanced design that scales data movement and compute together, aimed at workloads such as climate models, genomic analysis and financial risk, Data Center Dynamics reports.
"For agents tackling science and engineering, faster simulation means faster discovery," said Ian Buck, Nvidia's vice president of hyperscale and high-performance computing.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Silicon Brief or Daily AI Brief.
Also, consider following us on social media:
More from Data Centers
Nov 16 Sam Altman says he has 'strong empathy' for Jensen Huang but won't say if he should return Nov 16 Blackstone's new data centre REIT buys 75% of AWS joint ventures for about $5bn Nov 16 US hyperscale data centres emitted as much CO2 as Denmark in a year, study finds Nov 16 TensorWave signs six-year deal to provide OpenAI 457MW of AMD-powered AI capacity Nov 16 AMD completes World Labs acquisition, making Fei-Fei Li its chief scientistSilicon Brief
Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.
Whitepaper
Tensordyne Napier: What If One Rack Could Do the Work of Nine?
Tensordyne
This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.
Read moreYou may also like
Supermicro Introduces 256 CPU NVIDIA Vera Rack for AI Agents
NVIDIA Partners Expand European AI Clouds and Agent Controls
Google Cloud Adds NVIDIA Vera Rubin A5X Instances
AM Intelligence Seeks $3 Billion for Nvidia Rubin Chips
Modal Adds Nvidia Vera CPUs for AI Agents
Daily AI Brief: the AI news that matters, in your inbox.