Nvidia Accelerates Chip Engineering with AI Agents
3 hours ago
- Nvidia is using AI agents to address the growing complexity of chip engineering, which by 2030 is expected to produce 2 trillion chips and approach quadrillions of transistors.
- AI and accelerated computing are transitioning from productivity tools to foundational engineering infrastructure, enabling faster simulation, verification, and optimization.
- Nvidia's Vera CV100 CPU, with 88 custom Olympus cores, delivers 1.5x performance over AMD Epyc in EDA workflows, and is being used with Cadence and Synopsys to design future chips like 'Rosa'.
- Cadence's AI Super Agent and Synopsys' Fully Autonomous Design Verification Workflow achieve 40x faster RTL validation and 50x faster validated RTL, respectively, using agentic AI.
- Nvidia expanded its Agent Toolkit with updated CUDA-X libraries (including new cuISS for sparse linear systems) and a rearchitected PhysicsNeMo for AI physics models in engineering workflows.