method · first seen 12 Nov 2026
FP8 and FP4 block-scaled matrix multiplication
Also: block-scaled MMA
Low-precision tensor core matrix multiplication using per-block scaling factors in FP8 and FP4 formats.
Mentions per week
PapersCodeDiscussion
Earliest sources
- GC-MoE on SM120: Overlapping Expert-Parallel Token Dispatch with Warp-Level FP8xFP4 Block-Scaled MMA on Consumer BlackwellPapers · 12 Nov 2026 · Yong Li, Haoyu Wang, Kan Liu, Tao Lan
Latest sources
- GC-MoE on SM120: Overlapping Expert-Parallel Token Dispatch with Warp-Level FP8xFP4 Block-Scaled MMA on Consumer BlackwellPapers · 12 Nov 2026 · Yong Li, Haoyu Wang, Kan Liu, Tao Lan