NVIDIA cuDSS Library for GPU-Accelerated Simulations is Now Generally Available

General Availability of NVIDIA cuDSS

NVIDIA cuDSS, a CUDA-X library, has reached general availability. This library is designed to shift a critical component of engineering and scientific simulations onto GPUs.

Accelerating Complex Simulations

Modern engineering and scientific simulations, which are essential for predicting design performance, face increasing computational demands as designs become more complex. NVIDIA cuDSS addresses this by accelerating the process of solving large sparse systems, a computationally intensive step that frequently limits the number of design iterations teams can perform. By leveraging GPUs, cuDSS reduces the time required to obtain results and increases simulation throughput.

This acceleration allows users to explore larger, more realistic models, achieve more accurate representations, and expand the design space, potentially at a reduced cost.

Industry Adoption and Performance

Several industry leaders have already integrated NVIDIA cuDSS into their tools. Applied Materials reported up to a 10x acceleration in sparse solves within its Ginestra platform compared to CPU-only methods. COMSOL, using cuDSS with NVIDIA accelerated computing, saw a 4x speedup in an EUV lithography plasma simulation benchmark with ZEISS SMT.

Cadence uses cuDSS in its Celsius software for accelerating simulations of heat, structural stress, and warpage in advanced chip designs. Dassault Systèmes observed an 8-10x speedup with cuDSS in its Abaqus structural and computer simulation technology. Siemens incorporates cuDSS in Simcenter STAR-CCM+ for fluid-structure interaction workflows, and Synopsys uses it in its Sentaurus Device simulator for advanced chip simulations.

Scalability for Diverse Workloads

NVIDIA cuDSS is designed to scale effectively with growing simulation workloads. It can be deployed from individual workstations to large computing clusters, providing flexibility for various computational needs. Developers can integrate cuDSS into their existing applications.