NVIDIA cuDSS Accelerates Engineering and Scientific Simulations

NVIDIA today announced the general availability of cuDSS, a CUDA-X library that moves a key component of engineering and scientific simulations to GPUs. Complex engineering projects, such as the design of advanced chips and aircraft, rely on simulations to predict the performance of designs before they are realized. As designs become more intricate, running simulations takes increasingly longer, raising computational demands and limiting the number of ideas that can be tested.

With NVIDIA cuDSS, solving large sparse systems, a mathematical step common in engineering and scientific simulations, can be significantly accelerated. This is achieved by leveraging GPUs, reducing the time to obtain results and increasing simulation throughput. This allows end users to explore larger and more realistic models and more accurate representations, with a broader design possibility at lower costs.

Industry leaders like Applied Materials, COMSOL, and Siemens are integrating cuDSS into their tools to transform workflows for their end customers. For example, Applied Materials reports that with cuDSS, the speed of sparse solution processes in their Ginestra platform has been accelerated by up to 10 times. COMSOL has achieved a 4x speedup in their EUV lithography plasma simulation benchmark with cuDSS, while Dassault Systèmes reports up to 10x acceleration when using cuDSS in their Abaqus technology.

With the growing complexity and scale of simulation workloads, cuDSS can easily scale from individual workstations to large computing clusters. Developers have the ability to integrate cuDSS into existing software for engineering, science, and optimization, bringing GPU acceleration to tools already in use by engineers and researchers. It is accessible via interfaces in C, C++, Python, and Fortran.

Learn more about cuDSS by visiting the NVIDIA demo booth and the Dassault Systèmes booth at GTC Berlin.

Read the full article from Nvidia.