https://www.modular.com/blog/matrix-multiplication-on-blackwell-part-4---breaking-sota
Matrix Multiplication on Blackwell: Part 4 - Breaking SOTA
Ali Taha, Jiexiang Liu, Hengjie Wang, Abdul Dakkak
September 19, 2025
"...how MAX outperforms SOTA implementations of matmul... how the GPU hardware has become more complex with many features required to work in unison to achieve peak TFLOPs."