DEVOURED
Vera Arrives: NVIDIA's First CPU Built for Agents Lands at Top AI Labs
Nvidia's first custom CPU, Vera, designed for AI agents, has been delivered to major AI labs including Anthropic, OpenAI, and SpaceXAI.
What: Nvidia's new Vera CPU, featuring 88 custom Olympus cores and 1.2 TB/s memory bandwidth with 50% faster per-core performance, was hand-delivered by Nvidia VP Ian Buck to key customers like Anthropic, OpenAI, SpaceXAI, and Oracle. It serves as the host processor for the Vera Rubin NVL72 platform, integrating with Rubin GPUs via NVLink-C2C.
Why it matters: Nvidia is expanding its hardware ecosystem beyond GPUs to include custom CPUs optimized for AI workloads, aiming for deeper integration and performance gains in AI systems, especially for emerging agentic AI applications.
Takeaway: Developers working with advanced AI agent systems at labs like Anthropic or OpenAI may soon be running their workloads on Nvidia's Vera CPU and Rubin GPU platform. Anticipate new performance benchmarks and optimizations that specifically target the Vera-Rubin integrated architecture for AI agent development.
Decoder
- Vera CPU: Nvidia's custom-designed central processing unit, specifically engineered to complement their GPUs for AI and high-performance computing tasks, particularly for AI agents.
- Olympus cores: The custom core architecture developed by Nvidia for their Vera CPU.
- Vera Rubin NVL72: Nvidia's platform that integrates the Vera CPU with Rubin GPUs, designed for high-performance AI and HPC.
- NVLink-C2C: Nvidia's proprietary inter-chip interconnect technology, facilitating high-speed communication between GPUs and CPUs within a single system.
Original article
Full article content is not available for inline reading.
Explore all 43 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi