Ian Buck discloses Rubin HPC GPU and Nemotron for Science at SC26 keynote
NVIDIA Debuts Rubin HPC GPU and Sweeps MLPerf v6.1 at SC26
The Rubin HPC GPU and Vera Rubin platform extend NVIDIA's position in both scientific HPC and AI training benchmarks. The DGX Spark kit and Nemotron for Science lower the barrier to running large models and autonomous science agents outside shared cluster environments.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
At SC26, NVIDIA VP Ian Buck announced the Rubin HPC GPU, a science-and-engineering processor with more than 50,000 CUDA cores (roughly three times the Hopper count), FP64-to-FP4 precision range, sparse-data acceleration, seventh-generation NVLink, and an expected 2028 ship date. Concurrently, NVIDIA's Vera Rubin NVL72 platform swept MLPerf Training v6.1 as the only system with results across every benchmark, delivering up to 2x more training performance per GPU than GB300 NVL72, and led the inaugural MLPerf Endpoints agentic benchmark with more than 10x throughput per watt of GB300 NVL72 at equivalent response speeds. The DGX Spark HPC Developer Kit, a four-node ring-connected desktop cluster requiring no external switch, provides 512GB combined memory and 20TB NVMe storage and can run models up to 700B parameters; UC Berkeley's Rubin Lab reported results arriving 15x sooner compared to an 11-hour shared cluster queue. NVIDIA also released Nemotron for Science, pairing the Nemotron 3.5 Super VL model with a chemistry dataset and post-training recipes for life sciences and chemistry, enabling local deployment on DGX Spark or researchers' own hardware with full data and model ownership.
How it developed
NVIDIA releases CUDA-X science library updates including cuPhoton, cuPotential, Daviator, and cuSpa
Sources
Related
- Related storyNvidia Vera CPU deployment contracts
- NVIDIA Vera Rubin and NVLink Fusion platform
- Google Cloud Vera Rubin GPU instances
- CoreWeave's Vera Rubin and Forge at FullyConnected26
- RTX Spark and MXC on Windows
BackgroundCompute: the chips, clouds and data centers behind AI
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.