Firmware unlocks target FP32 accumulation on older NVIDIA GPUs
A Reddit user’s technical report says firmware-based unlocks can restore full tensor-core speed with FP32 accumulation on NVIDIA’s Turing, Ampere and Ada GPUs, while consumer Blackwell remains restricted. The author attributes the difference to factory fuse settings and a privileged override register, but the supplied evidence contains no independent hardware verification or NVIDIA statement confirming the mechanism or alleged intent.
The report says FP32 accumulation is common in AI training and inference, so unlocking the reportedly restricted tensor operations could materially change performance on affected consumer GPUs.
Hello guys, hoping you're doing fine. Wall of text next I know, but I want to get this info out of my chest, all hand written lol.
I wanted to make this post explaining some things about GPU unlocks that have been going on lately.
What are these GPU unlocks doing?
NVIDIA on purpose neuters tensor core performance when using FP32 accumulation (FP16, BF16, FP8 and TF32 inputs accumulating in FP32, which is what most AI/LLM/ML tasks use), for market segmentation. Normal FP32 on the CUDA cores (no tensor cores) is NOT affected.
Consumer and prosumer GPUs use the exact same die (like GB202 between the 5090 and 6000 PRO, AD102 between the 4090 and 6000 Ada, or even literally 1:1 with the 3090 Ti vs A6000, both with the same core amounts and 84 SMs).
How it works: the speed of each kind of tensor operation is set by fuses burned at the factory. On consumer cards, the fuse for "FP16/BF16 with FP32 accumulation" says HALF_SPEED, on the pro cards it says FULL_SPEED. These fuses don't cut any circuit, the GPU just reads them at boot and runs at that speed. The GPU also has an override register that can set them back to FULL_SPEED, but it's locked so only NVIDIA's most privileged firmware can write to it, not your PC.
#NVIDIA#GeForce RTX 20 Series#GeForce RTX 30 Series#GeForce RTX 40 Series#GeForce RTX 50 Series#RTX 5090#RTX 6000 PRO