Daytona Benchmarks NVIDIA Vera CPU on Agentic Workloads, Cited in Official Blog
mattturck · x · 2026-09-17
Daytona spent five days at NVIDIA HQ testing the new Vera CPU on agentic workloads, earning a citation in NVIDIA's official blog alongside Perplexity, ClickHouse and DeepInfra.
- Vera shows serious gains for agentic execution, with memory bandwidth above traditional server CPUs
- Wide SVE2 vector units with native FP8 make NumPy-heavy code fly
- Agent sessions accumulate hundreds of thousands of input tokens — stateful, long-running, CPU-heavy — matching Vera's design
Context from the AI Infra Summit (8,000+ attendees, up from 3,500): Amazon Annapurna's NVHBM custom HBM, d-Matrix pairing Vera CPUs with Raptor XPUs via NVLink Fusion, Lambda's 23% perf-per-watt gain with DSX MaxLPS, and Pinterest using Blackwell + Dynamo for visual discovery — all under the theme of optimizing tokens per watt for AI factories.
More from Infra
- Baseten Adds Server-Side Web Search for Open-Weight Models via One-Line Tool Config — PolarBearby · 2026-09-17
- The new AI stack starts below models: energy, grid and data centers — ingliguori · 2026-09-17
- US House Passes Bipartisan Bill Making Data Centers Cover More Grid Costs — Polymarket · 2026-09-17
- Dev trains a 2B Apache-licensed model on 24GB VRAM using OLMo embeddings and a 1B Engram table — NineThreeTilNow · 2026-09-17
- Applied Materials to invest $5bn in India as country boosts chip support — pstAsiatech · 2026-09-17
- Cloud AI Costs 10-20x More Than On-Premises, Expert Warns of Repeat of Early Cloud Mistakes — DavidLinthicum · 2026-09-17