NVIDIA cuDSS Reaches General Availability for GPU-Accelerated Engineering Simulation

NVIDIA's cuDSS library offloads large sparse system solving to GPUs, targeting engineering and scientific simulation workloads.

IT Solution Journal News Desk

Editorial illustration: A stylized editorial illustration showing a dense irregular network of glowing amber nodes and random connector rods on the left transitioning through a blazing golden energy beam into a perfectly ordered three-dimensional rectangular grid of evenly spaced luminous amber nodes on the right, set against a dark blue background.

AI-generated editorial illustration.

NVIDIA announced general availability of cuDSS, a CUDA-X library designed to accelerate the sparse linear solvers that form a core computational bottleneck in engineering and scientific simulations. According to NVIDIA, the library scales from workstations to computing clusters and supports C, C++, Python, and Fortran interfaces. Several enterprise software vendors report integration: Dassault Systemes claims 8-10x speedups in Abaqus solvers, Applied Materials reports up to 10x acceleration in its Ginestra platform, and COMSOL with ZEISS SMT recorded a 4x speedup in an EUV lithography benchmark. Cadence, Siemens, and Synopsys also cite cuDSS adoption in their respective simulation tools.

Source: NVIDIA News. Original announcement: October 21, 2026.