NVIDIA Unveils AI and Quantum Tools at SC26 to Advance Scientific Research




Home

Latest reporting

News


NVIDIA’s Ian Buck announced new AI agent tools, the Rubin HPC GPU, and quantum computing collaborations at SC26 in Chicago.





NVIDIA’s Ian Buck, vice president of hyperscale and high-performance computing, opened the company’s week at SC26 in Chicago with a keynote unveiling new tools and technologies aimed at advancing scientific research. The announcements included Nemotron for Science, the Rubin HPC GPU, and an expanded Vera CPU platform, alongside a new quantum computing collaboration with IBM and Fermi National Accelerator Laboratory.

AI Agents and Nemotron for Science

Buck highlighted the transformative role of AI agents in scientific workflows, stating, “[Agents] have been just a game changer in how we interface with our supercomputers, with our devices, our instruments and how we can do the research.” Nemotron for Science, a centerpiece of the announcements, integrates the Nemotron 3.5 Super VL open model with a new open chemistry dataset and post-training recipes for life sciences and chemistry. Researchers can deploy it locally on NVIDIA DGX Spark systems or their own supercomputers.

At Argonne National Laboratory, an agent built using Nemotron successfully planned and tested hypotheses, ran simulations, and identified two new binding sites from a single prompt. Other examples included an agent repairing 81% of broken GROMACS setups overnight on a DGX Spark and another accelerating a weather model routine by over 5x using NVIDIA GPUs.

Rubin HPC GPU and Vera CPU Platform

NVIDIA introduced the Rubin HPC GPU, designed for scientific, engineering, and AI workloads. The chip features over 50,000 CUDA cores—three times as many as its predecessor, Hopper—and supports precisions ranging from FP64 to FP4. It also includes sparse-data acceleration and seventh-generation NVIDIA NVLink for high-speed connectivity. Rubin HPC is expected to launch in 2028.

The Vera CPU platform, which supports AI agents by running tool calls between model steps, was also expanded. The German Climate Computing Center (DKRZ) selected Vera CPUs for its upcoming HLRE-5 supercomputer, slated for deployment in early 2028.

Quantum Computing Collaboration

NVIDIA announced a quantum computing co-design platform in partnership with IBM and Fermilab. The initiative pairs NVIDIA DGX B200 systems with IBM quantum computers, including those powered by IBM Quantum Nighthawk. Buck also introduced a breakthrough in quantum error correction, with a CUDA-Q decoding algorithm called TCBP-JOBS that reduces computation time for error-correction research from 12 weeks to 45 minutes. Additionally, a simulation on Jülich’s JUPITER supercomputer demonstrated a tenfold reduction in operations required for one logical step using a technique called magic state cultivation.

Advancing Scientific Computing

In networking advancements, NVIDIA showcased its Photonics co-packaged optics switches, which Buck noted are 15x more reliable in terms of bit errors and link flaps. These developments, alongside NVIDIA’s expanded CUDA-X libraries and AI-driven scientific workflows, underscore the company’s commitment to accelerating discovery across disciplines such as astrophysics, materials science, and biology.



Source link

Be the first to comment

Leave a Reply

Your email address will not be published.


*