AI BriefingNvidia Vera CPU Benchmarks: Olympus Cores Show Outstanding Performance
Key point
NVIDIA's next-generation CPU for AI data centers, Vera, has signaled strong performance with its self-designed Olympus cores.
Details
NVIDIA's next-generation data center CPU, Vera, is an ARM-based processor optimized for agentic AI and modern data center workloads.
The biggest feature of Vera is that it uses NVIDIA's self-designed Olympus cores instead of the Arm Neoverse-V2 used by the previous Grace CPU.
Key Technical Specifications:
- Core Configuration: 88 Olympus cores (2x performance and higher energy efficiency compared to the previous generation)
- Threads and Precision: Armv9.2 ISA support, FP8 precision support, total of 176 threads via spatial multithreading
- Memory and Bandwidth: Up to 1.2TB/s bandwidth via LPDDR5X memory
- Cache and Connectivity: 2MB L2 cache per core, 164MB L3 cache, PCIe Gen 6 and CXL 3.1 support
Vera is planned to be used as the host CPU in NVIDIA's NVL72 Vera Rubin rack, or supplied as a standalone CPU rack, with shipments targeted for the second half of this year. On Linux, it will receive stable support on major ARM64 distributions such as Ubuntu and Fedora via the Linux 7.1+ kernel.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Stanford and NVIDIA Release 'Contrastive Language Model' CLM with 9x Faster InferenceFinnish AI Cloud Startup Verda Raises $189 Million in Series B
NVIDIA Releases Open-Source LLM Router 'Switchyard'… Up to 30% Cost Savings Compared to Opus