Zyphra open-sources FP8 attention kernel ZYFA for AMD MI355X
Understand this story
Zyphra has open-sourced ZYFA, a set of FP8/BF16 prefill attention kernels for AMD Instinct MI355X, under the Apache 2.0 license. The kernels support only MI355X (CDNA4, gfx950), run on ROCm 7.2.3 and PyTorch 2.12, and cover only prefill, not decode or paged KV. According to MarkTechPost, ZYFA delivers up to 1.76x speedup.
AI overview based on source reports · Updated 2 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- MarkTechPostZyphra ZYFA: FP8 FlashAttention for AMD MI355X, Up to 1.76x Faster
Zyphra released ZYFA, an Apache 2.0 open-source library of FP8 and BF16 prefill attention kernels for AMD Instinct MI355X GPUs (ROCm 7.2.3, PyTorch 2.12) with a drop-in flash_attn_func interface. Its FP8 kernels run 1.16-1.76x faster than AMD's AITER on real model activations, reaching 2,672 TFLOP/s at 32K tokens (53% of FP8 peak), and show 62-78% lower FP8 error than AITER at head dimension 128. ZYFA covers causal prefill forward only, with no decode or paged KV support.
Activity for this story
Not enough continuous observations to show a trend yet.