Chip at bottom of perf list: new GPUs pack up to 72x H100 inference capacity, per Epoch AI
bookwormengr · x · 2026-11-10
Pushing back on claims that selling a certain chip amounts to giving up compute advantage, the author notes the chip in question ranks second from the bottom — the latest Nvidia and AMD chips are over 50x more powerful. Epoch AI's chart shows per-chip inference capacity relative to GB300=1: AMD MI455X at 2.89, Nvidia VR200 at 2.18, versus 0.04 for H100 and 0.12 for H200 — up to 72x H100. Whether a sale violates law is for governments to decide, he argues; remote cloud access to top chips is a separate question worth its own debate.
More from Infra
- Maryland activists with 25K signatures block $5B AI data center on abandoned mall site — beffjezos · 2026-11-10
- Chrome 153 ships native tab recorder: 120fps 4K AV1 capture, no ffmpeg needed — Vjeux · 2026-11-10
- AWS EKS now offers per-Pod hardware isolation with Nitro Sandbox — tekbog · 2026-11-10
- AI capex needs to grow 5-10x before oversupply, argues Rob LeClerc — robleclerc · 2026-11-10
- simdjson disables 256-bit LASX on Loongson 3A6000: 128-bit LSX is 6.7% faster — lemire · 2026-11-10
- Vera launches gyms, benchmarks and agents for cloud infrastructure operations — tianyin_xu · 2026-11-10