Loading...
Loading...
Early Phoronix benchmarks reveal Nvidia’s Vera data-center CPU, built on Arm v9.2 with 88 in-house Olympus cores, delivers a substantial performance leap—about 63% faster than Nvidia’s prior 72-core Grace and roughly 10% ahead of a 64-core AMD EPYC 9575F, while outpacing Intel’s 128-core Xeon by about 55%. Positioned for agentic AI and reinforcement-learning workloads, Nvidia claims Vera is overall 1.5× faster than x86 alternatives and compiles the Linux kernel twice as fast. These initial results signal Nvidia’s push to challenge traditional x86 server CPUs with custom Arm designs optimized for demanding AI tasks.
Nvidia's Vera demonstrates that custom Arm server CPUs can outpace established x86 datacenter processors, affecting server architecture choices and procurement. Tech professionals should reassess CPU selection, software portability, and optimization priorities for AI and RL workloads.
Dossier last updated: 2026-05-27 10:16:22
Nvidia Vera CPU Benchmarks: Olympus Cores Delivering Great Performance
Nvidia Vera CPU Benchmarks: Olympus Cores Delivering Great Performance
Phoronix benchmark results for NVIDIA's new Vera data-center CPU show it is 63% faster on geometric mean than NVIDIA's prior 72-core Grace processor. Announced at GTC 2026 as a CPU optimized for agentic AI and reinforcement learning, Vera uses Arm v9.2 and 88 NVIDIA-designed Olympus cores. The preliminary suite also shows Vera outperforming a 64-core 5 GHz AMD EPYC 9575F by about 10% and beating Intel's 128-core Xeon 6980P by roughly 55%. NVIDIA highlighted additional claims that Vera is 1.5× faster than x86 processors overall and compiles the Linux kernel twice as fast, positioning the chip for AI workloads demanding high single-thread and multi-thread CPU performance.
Michael Larabel / Phoronix : Initial benchmarks show Nvidia's Vera CPU, which features 88 in-house-designed Olympus cores, packs a heavy-hitting punch, beating Intel's and AMD's x86_64 CPUs — NVIDIA's Vera data center CPU isn't ramping up until later this year but I recently had the opportunity to try out this new ARM-based CPU designed for agentic AI workloads.