NVIDIA's Vera CPU, a processor built specifically for agentic AI workloads, has outperformed Intel and AMD's x86 chips in Phoronix benchmarks, delivering a 1.5x performance gain. The results, published this week, mark the first public head-to-head comparison between NVIDIA's server CPU and traditional x86 designs.
Designed for autonomous AI
Unlike general-purpose CPUs, Vera is optimized for agentic AI—systems that can make decisions and take actions without human supervision. That focus shows in the benchmark scores, where Vera handled AI inference tasks far faster than comparable Intel Xeon and AMD EPYC processors. Phoronix tested a range of workloads, including large language model inference and real-time sensor processing, and Vera led in every category.
What the benchmarks measured
The Phoronix test suite included both single-threaded and multi-threaded AI-oriented tasks. Vera's 1.5x advantage held across the board. The results are especially striking because NVIDIA is relatively new to the CPU market, while Intel and AMD have decades of x86 refinement behind them. The Vera architecture relies on custom Arm-based cores and a tightly integrated GPU memory fabric—a combination that Intel and AMD have struggled to match.
NVIDIA already dominates AI accelerators with its GPUs. Vera extends that reach into the CPU space, potentially letting customers run entire AI pipelines on a single vendor's hardware. For Intel and AMD, the benchmark results signal a serious competitive threat in the data center, where they've long held the CPU monopoly. Neither company has an Arm-based server chip that can match Vera's performance in these specific workloads.
What comes next
Phoronix plans to publish additional latency and power-efficiency tests in the coming weeks. NVIDIA has not announced a commercial release date for Vera, but the company is expected to begin sampling the chip with cloud providers later this year.
