Nvidia Deploys Vera CPU for Chip Development

Nvidia expands its engineering Agent Toolkit with the Arm-based Vera CPU and new CUDA-X libraries to accelerate semiconductor design and EDA.

Nvidia is utilizing its Arm-based Vera CPU architecture as a high-performance computing platform for internal Electronic Design Automation (EDA). Deployed in a Portland datacenter cluster, Vera combines high single-thread performance with significant memory bandwidth. The company also expanded its Agent Toolkit with PhysicsNeMo and new CUDA-X libraries, including cuISS for sparse linear equations, cuDSS for multi-node numerical calculations, and cuEST for quantum chemistry. These tools, alongside the Nemotron 3 Ultra model for agentic RTL coding, are being adopted by industry leaders. Samsung uses cuLitho and PhysicsNeMo for lithography and thermal stress analysis, while Cadence, Siemens, and Synopsys integrate the technologies into simulation and verification workflows. Partners include TSMC, Silvaco, and Keysight.