NVIDIA unveils the Hopper H100
AIMade by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made
Unveiled the Hopper H100 — the Transformer Engine dynamically selects FP8 or 16-bit per layer and rescales. NVLink 4 delivers 900 GB/s
Why it matters
The turning point that made FP8 training a hardware-supported feature. The first case of software and silicon bound together as one
Source — Primary sources first; secondary reports are labelled “Press”
Officialdeveloper.nvidia.comhttps://developer.nvidia.com/blog/nvidia-hopper-architecture-in-depth/Added in Week 35, 2026 · See that week
On the map — ⑱ AI Infrastructure Software line
- Previous stopZeRO-Offload
- This record · stopHopper H100
- Next stopZeRO++