Nvidia releases cuPhoton for GPU-accelerated scientific image analysis

Nvidia has launched cuPhoton, an open-source CUDA-X toolkit that moves high-throughput scientific image processing from CPU pipelines to GPUs, targeting astronomy, X-ray analysis and laser experiments. The company claims speed-ups of up to 14,900× for image loading and 14,550× for signal processing versus an x86 CPU baseline.

• Version 0.1.3 supports Python 3.12–3.14 and CUDA 13 on Linux
• Modules xDataReader, xRep, xPois, xFit, xScan and xRay handle FITS and HDF5 data
• Claimed speed-ups of up to 14,900× for loading and 14,550× for signal processing
• Released under Apache 2.0 with wheels for x86-64 and Arm64