NVIDIA Officially Reveals the Architecture Behind Its Future Feynman Platform

NVIDIA has shared the first details about its next-generation computing solutions. In a recent post on the company's official blog, it revealed the name and architectural design of the CPU cores that will power its upcoming generation of proprietary superchips.

The Evolution of Codenames

According to NVIDIA's updated roadmap, the current-generation Vera processors, designed to work alongside Rubin GPUs, feature 88 custom CPU cores codenamed Olympus. These processors are built specifically for agentic AI workloads, where CPUs are responsible for code execution, tool orchestration, data management, and controlling isolated environments.

The next major step in NVIDIA's infrastructure roadmap will be the Feynman architecture. This platform will introduce the Rosa CPU, powered by an entirely new CPU core design known as Rigel.

What to Expect from Rigel

The new Rigel CPU core will be based on the Arm v9.2 architecture, offering a significant increase in per-core performance while maintaining the same physical die size.

According to NVIDIA, Rigel will feature several key improvements, including a more efficient instruction delivery mechanism, a larger L2 cache, and an optimized memory subsystem designed to improve overall computing efficiency.

Release Timeline and Remaining Unknowns

Although NVIDIA has unveiled the core architecture, the company has yet to disclose detailed specifications for the Rosa processor. The final core count, clock speeds, cache sizes, and official launch date remain undisclosed.

The Feynman platform is currently expected to reach the market around 2028.