NVIDIA cuDSS Now Generally Available to Accelerate Engineering and Scientific Simulations

The CUDA-X direct sparse solver library targets engineering and scientific simulation workloads, scaling from single workstations to compute clusters.
Key points
- Accelerates solving large sparse linear systems by offloading calculations from CPUs to GPUs
- Provides developer interfaces for C, C++, Python, and Fortran
- Integrated into simulation software including Cadence Celsius, Dassault Systèmes Abaqus, Synopsys Sentaurus Device, and Siemens Simcenter STAR-CCM+
Why it matters
Simulation software developers can offload sparse linear math routines to NVIDIA GPUs without building custom solver backends.
Sources
- NVIDIA cuDSS Now Generally Available to Accelerate Engineering and Scientific Simulations (NVIDIA AI, Oct 21, 2026, 11:18 AM)