Using Accelerated Computing to Live-Steer Scientific Experiments at Massive Research Facilities

2026-02-10 · Source: NVIDIA Technical Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Software Development & Engineering · Depth: Advanced, medium

Summary

NVIDIA is collaborating with major scientific research facilities, the NSF-DOE Vera C. Rubin Observatory and SLAC's Linac Coherent Light Source II (LCLS-II), to overcome challenges in processing massive data rates and enabling real-time experiment steering. These facilities, which took a decade to build, generate petabytes of data from observing the entire southern sky and capturing atomic-scale X-ray movies. NVIDIA's accelerated computing, utilizing GPU-accelerated Python libraries like CuPy and cuPyNumeric, has enabled breakthroughs such as reducing data analysis times from nine months to four hours. Specific solutions include Accelerated Space and Time Image Analysis (ASTIA) for the Rubin Observatory's 3.2-billion-pixel camera, which discovers 2,000+ new asteroids nightly, and X-ray Analysis for Nanoscale Imaging (XANI) for LCLS-II, which produces 1 million X-ray bursts per second to map quantum phenomena.

Key takeaway

For research scientists and engineers grappling with petabyte-scale data and the need for real-time experiment steering, adopting NVIDIA's accelerated computing stack, including CuPy and cuPyNumeric, offers a critical path to faster insights. Your team can leverage these GPU-accelerated Python libraries and reference designs like XANI to automate complex data pipelines, reducing analysis times from months to hours and enabling dynamic adjustments to experiments, thereby accelerating scientific discovery.

Key insights

GPU-accelerated computing enables real-time analysis and steering of exascale scientific experiments, drastically reducing data processing times.

Principles

• Modular, parallel pipelines automate data workflows.
• Unified memory simplifies large dataset handling.
• Scalable software runs across diverse hardware.

Method

Develop a NumPy workflow, port to CuPy for single-GPU parallelization, then use cuPyNumeric to distribute computations across multi-GPU nodes for cluster-scale acceleration.

In practice

• Use XANI for pixel-wise, model-fitting X-ray workloads.
• Start development on a laptop, scale to DGX SuperPODs.
• Adopt CUDA Python for live-steering instruments.

Topics

• GPU-Accelerated Computing
• Real-time Experiment Steering
• Astrophysics Data Analysis
• Ultrafast X-ray Imaging
• CuPy and cuPyNumeric Libraries

Code references

• NVIDIA/accelerated-computing-hub

Best for: AI Scientist, Research Scientist, Software Engineer

Related on AIssential

• From Materials Simulation to Experimental Astronomy, New NVIDIA AI Software Unlocks Scientific Discoveries — NVIDIA's new GPU-accelerated software dramatically speeds scientific data processing and simulation across diverse research domains.
• Topping the GPU MODE Kernel Leaderboard with NVIDIA cuda.compute — The `cuda.compute` library enables Python developers to achieve near "speed-of-light" GPU performance with CUB primitives.
• How NVIDIA Turned GPUs into Supercomputers — NVIDIA GPUs, initially for gaming, democratized scientific computing by offering supercomputer-level performance.
• Designing GPU-Accelerated Query Engines with NVIDIA GQE — GPU-accelerated query engines achieve significant speedups by optimizing data movement, compression, and execution on NVIDIA hardware.
• NVIDIA cuPhoton: The Tech Fast-Tracking Space Discovery — NVIDIA cuPhoton accelerates astronomical data processing from months to minutes using GPU-accelerated pipelines.
• A Guide to Understanding GPUs and Maximizing GPU Utilization — GPU bottlenecks often stem from inefficient data pipelines, not compute, requiring optimized data transfer and processing.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by NVIDIA Technical Blog.