From NVIDIA: 2024-11-18 13:30:39

Scientists facing the challenge of sifting through massive amounts of data now have a solution with the NVIDIA cuPyNumeric accelerated computing library. This tool allows researchers to easily run their Python code on CPU-based laptops, GPU-accelerated workstations, or supercomputers, enabling faster data analysis and decision-making. The latest version of cuPyNumeric offers support for the NVIDIA GH200 Grace Hopper Superchip, automatic resource configuration, and improved memory scaling, benefiting institutions like SLAC National Accelerator Laboratory and Los Alamos National Laboratory.

Python, the go-to programming language for data science and numerical computing, is widely used in scientific fields like astronomy and drug discovery. The NumPy library, with over 300 million downloads last month, is a critical component of many applications. By implementing cuPyNumeric, researchers can scale their programs to run on multiple GPUs, significantly increasing throughput for analyzing large datasets from instruments like electron microscopes and radio telescopes.

At SLAC National Accelerator Laboratory, cuPyNumeric has accelerated X-ray experiments at the Linac Coherent Light Source, reducing run time from minutes to seconds and enabling parallel analyses. Similarly, other institutions like Australia National University and Los Alamos National Laboratory are utilizing cuPyNumeric to scale their algorithms for climate models and machine learning, showcasing its potential to accelerate scientific discovery and streamline data analysis processes.

By leveraging cuPyNumeric, institutions like UMass Boston and the National Payments Corporation of India have seen significant improvements in their data analysis workflows. For example, NPCI was able to speed up matrix multiplication by 50x on NVIDIA DGX systems, allowing for faster processing of transaction data and more efficient detection of money laundering activities. With the potential for limitless GPU scalability and zero code changes, cuPyNumeric is revolutionizing the way researchers handle big data challenges and accelerate scientific breakthroughs.



Read more at NVIDIA: cuPyNumeric Enables Scientists to Harness GPU Acceleration at Cluster Scale