Advancing Semiconductor Innovation Across Materials Engineering and Manufacturing | NVIDIA Technical Blog

This post details the integration of NVIDIA CUDA-X libraries (cuDSS, cuEST, PhysicsNeMo) into Applied Materials' Ginestra and ACE+ platforms to create an end-to-end digital development model for semiconductor innovation. It highlights specific speedups: up to 10x with cuDSS for Ginestra's sparse linear algebra, ~55x with cuEST for DFT simulations on B200 systems, and up to 35x faster topography simulations with ACE+ and PhysicsNeMo. The post also mentions the use of Omniverse for fab digital twins and emphasizes the shift from physical experimentation to simulation for materials discovery, process engineering, and factory optimization.