Workflow
新思科技:通过英伟达Grace Blackwell平台将芯片设计加速30倍
NVDANvidia(NVDA) 证券时报网·2025-03-20 05:38

Core Insights - Synopsys and NVIDIA are deepening their collaboration to accelerate chip design by up to 30 times using the NVIDIA Grace Blackwell platform [1][2] - The partnership aims to enhance the efficiency of electronic design automation (EDA) software, significantly reducing simulation times for various applications [2][3] Group 1: Performance Enhancements - Synopsys' PrimeSim™ SPICE simulation workload is expected to achieve a 30-fold acceleration using the NVIDIA Grace Blackwell platform, reducing simulation time from days to hours [2] - The Proteus tool has been optimized for NVIDIA's H100 GPU, achieving a 15-fold speed increase for optical proximity correction (OPC) and a projected 20-fold acceleration for lithography simulations [2] - Early experiments indicate that applying GPU support and NVIDIA's CUDA-X library to Synopsys' Sentaurus™ TCAD solutions can reduce computation time by 10 times [3] Group 2: Future Developments - Synopsys plans to continue advancing its entire product line on NVIDIA platforms, with over 15 cutting-edge EDA solutions optimized for the Grace CPU architecture expected by 2025 [4] - The integration of NVIDIA's NIM microservices with generative AI technology is anticipated to further enhance chip design efficiency, potentially doubling productivity for partners [4]