NVIDIA claims Vera CPU beats x86 on AI agents: 1.6x more agents per watt, 2x faster
NVIDIA Blog · rss · 2026-11-06
NVIDIA published agentic workload benchmarks for its Vera CPU, claiming that in a 100-megawatt data center comparison a single Vera beats both the fastest x86 CPU and the most energy-efficient one: agent speed above the fastest x86 while supporting 1.6x more agents per watt, and 2x faster with leading efficiency versus the most efficient x86.
The argument: agents shuttle between GPU reasoning and CPU execution for search, DB queries, code testing, and actions, and power-limited AI factories need one CPU balancing speed and density.
Early adopter results: Perplexity ran agentic coding 1.5x faster; DeepInfra hit up to 2.2x best x86 on agent orchestration; ClickHouse scaled near-linearly on 176 cores topping ClickBench with no tuning; DuckLabs reached 1.5x Intel Xeon 6 median TPC-H SF1000; Daytona spun up 1,000 agent sandboxes in 10.8s vs 88.4s on x86; Modal got up to 2x sandbox throughput per core.
More from Infra
- Qwen3.8-27B hits 381 tok/s on a single RTX 3090 via inference software alone — Redo · 2026-11-06
- Dolphin Networks details 3.8× throughput gains running DeepSeek-V4.1 Flash on PCIe GPUs — AlpinDale · 2026-11-06
- Lawyer runs DeepSeek-V4.1-Flash on 8x 2018 V100s: 215 tok/s on code, 524K context fully in VRAM — TumbleweedNew6515 · 2026-11-06
- Hyperkernel generates SOTA megakernels across 11 domains, now on GitHub — MillionInt · 2026-11-06
- Musk: space will be the cheapest place to run superintelligence in 2-3 years — XFreeze · 2026-11-06
- Ex-Microsoft engineer: Azure runs 1M+ VM services, adds a datacenter every 4 days — viksit · 2026-11-06