NVIDIA cuDSS Generally Available to Accelerate Engineering Simulations on GPUs

NVIDIA cuDSS, a new CUDA-X library, shifts the computation of large sparse linear systems — a critical step in engineering and scientific simulations — onto GPUs. This enables significant speedups for complex designs in chips, aircraft, and industrial systems, allowing teams to test more ideas faster. Early adopters including Applied Materials, COMSOL, and Dassault Systèmes report speedups ranging from 4x to 10x compared to CPU-only approaches, reducing costs and expanding the design space.