Skip to content
Christmas & Holiday Lights 2026Explore the community map

Scientific Computing & Quantum HPC

NVIDIA Marks 20 Years of CUDA with New CUDA-X and Quantum Slurm Tools

NVIDIA celebrated the 20th anniversary of CUDA across 11 million developers at SC26, introduced cuPhoton, cuPotential, and cuSpa libraries with up to 900x speedups, and expanded Slurm 26.11 to orchestrate quantum processors.

By Lamorinda Post · Syndicated from Diablo Valley News
Syndicated reporting from Diablo Valley News · Diablo Valley News
Published · 4 min read

Candid 35mm documentary photograph of a modern scientific supercomputing data center aisle with dark server racks and fiber-optic cabling
Rows of high-performance server cabinets and structured fiber-optic interconnects power scientific simulations and hybrid quantum-classical workloads inside a modern supercomputing center. (Photo: Diablo Valley News) · Diablo Valley News Visual Desk
Your listening Queue & catch up
Unheard stories first, timed at 1×. Your selections stay first.

Add a story or build a catch-up queue to start listening.

Scientific computing and national laboratory workflows reached new milestones on Monday at SC26 as NVIDIA celebrated the 20th anniversary of its CUDA parallel computing platform, introduced specialized CUDA-X libraries used by researchers at Lawrence Berkeley National Laboratory and Pacific Fusion, and expanded Slurm 26.11 to schedule quantum processors alongside GPU supercomputers.

NVIDIA Marks 20 Years of CUDA Across 11 Million Developers

Twenty years after introducing CUDA in November 2006 at SC06, NVIDIA honored the researchers and engineers who built the platform into a foundation used by more than 11 million developers and supported by more than 1,000 CUDA-X libraries, according to the NVIDIA Blog. Highlights across the two decades include biophysicist Klaus Schulten's team at the University of Illinois Urbana-Champaign, which collaborated with the University of Pittsburgh and Vanderbilt University in 2013 on the Blue Waters supercomputer to resolve the atomic structure of the HIV capsid; University of Toronto researchers Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton training AlexNet on 2 NVIDIA GPUs to win the 2012 ImageNet competition; and Demis Hassabis and John Jumper sharing the 2024 Nobel Prize in Chemistry for protein structure prediction with CUDA-accelerated AlphaFold 2.

Expanded CUDA-X Scientific Libraries Accelerate Berkeley Lab, Fusion, and Biology Research

To speed scientific discovery across astrophysics, fusion energy, materials science, and biology, NVIDIA expanded its CUDA-X portfolio of scientific libraries with cuPhoton, cuPotential, Daviator, DAQIRI, and cuSpa, the NVIDIA Blog reported. Using the cuPhoton.xBeam module on multi-GPU systems or an NVIDIA DGX Spark system, researchers reduced Event Horizon Telescope black hole observation correlations—including jets J1924 and 3C279 and Messier 87*—from more than a month to 8 minutes, while Pacific Fusion researchers paired the cuPhoton.xLens module with a local Nemotron agent that steers plasma simulations in real time.

In molecular dynamics, cuPotential—developed from work at the Simons Foundation's Flatiron Institute and D. E. Shaw Research—accelerates simulations by up to 3x across Schrödinger, AMBER, NAMD, GROMACS, LAMMPS, and Open Molecular Software Foundation workflows for adopters including Cadence, Fathom Therapeutics, Topos Bio, Genesis Molecular AI, and Terray Therapeutics. For materials science, Daviator brings batched GPU kernels to NVIDIA ALCHEMI to run Vienna Ab initio Simulation Package quantum-mechanical calculations up to 22x faster at up to 75% lower cost, while at Lawrence Berkeley National Laboratory (Berkeley Lab), the NVIDIA DAQIRI networking library links high-speed electron microscopy detectors directly to GPUs for real-time atomic-resolution analysis. In biology, open-source cuSpa accelerates cell polygon calculations by 900x over CPUs at Memorial Sloan Kettering Cancer Center, rapids-singlecell—codeveloped with scverse and integrated into 10x Genomics' Sentira platform—enabled Therna Biosciences to process single-cell data 200x faster, and GPU-accelerated HMMER ran protein sequence searches up to 84x faster at Basecamp Research.

Slurm 26.11 and CUDA-Q Unify Quantum and GPU Supercomputing

With quantum processor deployments at supercomputing centers growing nearly sevenfold over the past 5 years to nearly 30 supercomputing centers deploying more than 50 quantum processors, NVIDIA announced that Slurm 26.11—available later this month—expands support for the QRMI and QDMI interfaces so quantum processing unit builders including Alice & Bob, IBM, IQM, Pasqal, Quantinuum, and MQSC can integrate hardware through a single configuration file, according to the NVIDIA Blog. Complementing Slurm, updates to the open CUDA-Q platform include CUDA-Q QEC with a TCBP-JOBS error-correction decoder that is 13,000x faster than the industry standard, fault-tolerant primitives that enabled Fermilab to complete 5-month studies in 3 weeks, CUDA-Q Algorithms, and CUDA-Q Pulse, alongside Quantum Motion's QUICHE integration, an Altera and Oxford Quantum Circuits FPGA controller using NVQLink, QunaSys simulating more than 300 million quantum operations for a 4-hydrogen-atom model on 1,024 NVIDIA Hopper GPUs on ABCI-Q, and NVIDIA researchers simulating 104 noisy qubits on the JUPITER supercomputer.

Sources

Comments (0)

No comments yet. Be the first to share a thought on this story.