GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards github.com

Zagitta’s GPU-unlock 1.0.0 is a Linux kernel module that unlocks the FP/BF16 instruction issue rate on Nvidia GeForce RTX 3000 cards based on GA102, GA103, GA104 and GA106 chips, without patching the Nvidia driver. User reports show BF16 dense matrix-multiplication gains of roughly 25–80%; on an RTX 3090 Ti, the project reports BF16 throughput rising from 87.5 to 157.41 TFLOPS.