Zyphra releases FP8 FlashAttention kernel ZYFA for AMD MI355X, claims up to 1.76x speed boost
Zyphra released ZYFA, an open-source kernel library for AMD Instinct MI355X GPUs, on November 6. The FP8 and BF16 prefill attention kernels improve performance by 1.16–1.76x over AMD’s default AITER library on real model activations. ZYFA’s FP8 error is 62–78% lower than AITER’s at a head dimension of 128, according to Zyphra’s benchmarks, which include Mistral-7B, Qwen3-8B, and DeepSeek-V2-Lite…
Key points
- ZYFA delivers **1.16–1.76x FP8 speedup** over AMD’s AITER on real model activations like Mistral-7B and Qwen3-8B
- FP8 error drops **62–78% lower** than AITER at head dimension 128, with FP64 comparisons showing **3.4×10⁻³ KL divergence** for Mistral-7B
- Library runs on **AMD MI355X GPUs only**, requires ROCm 7.2.3 and PyTorch 2.12, and is **open-source under Apache 2.0**
Zyphra ZYFA: FP8 FlashAttention for AMD MI355X, Up to 1.76x Faster
MarkTechPost · 6 November 2026
Loading the full article…
This text was published by MarkTechPost and written by Michal Sutter. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- IREN’s Texas delays reportedly cut Microsoft’s AI capex by $5B · 7 src
- Liontrust highlights NVIDIA’s AI infrastructure role in Q3 2026 letter · 3 src
- OpenAI tests demand response in Georgia hub with 1,000MW cut · 1 src
- TSMC may enter 3D DRAM production on Nvidia’s request amid AI-driven shortages · 2 src
- Google buys Crusoe’s Wyoming AI data-center site after pre-development work · 1 src
Comments
via GitHub Discussions