Hierarchos-Native v0.2.5 ships P3 FP32-merge for RDNA3 WMMA GEMM, up to 27% faster
PhysicsDisastrous462 · reddit · 2026-11-22
Duplicate post of the Hierarchos-Native v0.2.5 release: P3 sums the eight WMMA FP16 partials in FP32 with one final rounding, passing 37/37 numerical gates and improving max error to 0.02594; per-dispatch GEMM gains of 1.02x–1.27x; portable Windows x64 package ships with statically-linked executables, Vulkan driver required; qualified only on RDNA3 Phoenix.
Related event: Hierarchos-Native v0.2.5 Boosts RDNA3 FP16 GEMM by 27%(2 posts)→
More from Infra
- LBNL report: data center load growth did not drive up residential electricity prices (2019-2025) — Eliv_nurotic · 2026-11-22
- WSJ: Associate of Jensen Huang's Partner Accused of Funneling Nvidia Chips to China — GaryMarcus · 2026-11-22
- Hierarchos-Native v0.2.5 ships P3 FP32-merge for RDNA3 WMMA GEMM, up to 27% faster — PhysicsDisastrous462 · 2026-11-22
- prime-rl 0.11.0 ships: 1.5x faster training, new pretraining entrypoint, Mega MoE dispatcher — samsja19 · 2026-11-22
- Math suggests 7.7M GPUs would serve everyone: fuel for the AI overbuild hypothesis — bookwormengr · 2026-11-22
- Chinese NAND maker YMTC promotes ZHITAI SSDs at Korea's G-Star game show — terryyuezhuo · 2026-11-22