Radar ·
FP/BF16 issue-rate unlock on the RTX 3090 with gpu-unlock
A post reports that the gpu-unlock project raised measured FP/BF16 throughput in PyTorch on an RTX 3090 from 75T to 135T, with a 360 W limit. The repository describes unlocking the issue rate of these instructions up to the level of professional cards.
Source: briefing published on 2026-10-23. According to the post, the gpu-unlock project raised measured PyTorch throughput on an RTX 3090 from 75T to 135T, with a 360 W limit. The GitHub repository describes unlocking the issue rate of FP/BF16 instructions up to the equivalent of professional cards.
The relevant point is that FP/BF16 throughput of consumer Nvidia GPUs can differ from professional cards, which matters for ML tuning. The figures come from the post and the repository and have not been independently verified here.
To check, read the README, the license and the terms of use of the original repository, and repeat the measurement with your own PyTorch version and your own hardware. Changing GPU behavior is outside the scope of Rota Nacional.
Original source: Desbloqueio da taxa de issue FP/BF16 na RTX 3090 com gpu-unlock