Nvidia's Scarcity Is a Firmware Setting

A GitHub developer just proved that Nvidia's RTX 3090 carries the same tensor core hardware as a professional A6000 — the company simply throttles FP16 and BF16 instruction issue rates in firmware. The open-source gpu-unlock kernel module, released on October 24, 2026, lifts these artificial limits on every Ampere GeForce card from the RTX 3060 12GB to the RTX 3090 Ti. The result: BF16 throughput on an RTX 3090 Ti jumps from 87.5 to 157.4 TFLOPS — an 80% gain from flipping two bits in firmware.

📖 Read the full version with charts and embedded sources on ComputeLeap →

This is not a security vulnerability. It is not a driver hack.…