GoKawiilOpen-source kernel mods unlock full-speed FP32-accumulate AI math on GeForce RTX 20/30/40 GPUs
Developers have released Linux kernel modules—gpu-unlock, gpu-unlock-ada, and turing-unlock—that rewrite speed-select registers on Nvidia RTX 20-, 30-, and 40-series GPUs before the driver loads, removing a hardware-level cap on FP32-accumulate tensor math that Nvidia normally reserves for workstation cards. Benchmarks show gains ranging from roughly 12% to nearly 2x depending on card and workload, with an RTX 3090 Ti jumping from 87.5 to 157.4 BF16 TFLOPS and an RTX 4090 doubling its raw FP16 instruction rate, though real-world matrix workload gains are smaller.
GoKawiil's interpretation of the reporting above, not reported fact.
These mods could make used consumer GPUs, already popular for home AI work due to cheap VRAM, significantly more competitive with costly workstation cards like the RTX A6000 for training and inference tasks. Because the unlock works by rewriting hardware registers outside Nvidia's official driver, it raises questions about stability, warranty, and whether Nvidia could patch out the workaround in future driver or firmware updates. The gap between raw instruction-rate benchmarks and real matrix-multiply performance suggests actual practical speedups will vary widely by workload.
- Multiple open-source Linux kernel modules now unlock full-speed FP32-accumulate tensor math on RTX 20-, 30-, and 40-series GPUs.
- Benchmarked gains range from about 12% to nearly 2x depending on GPU generation and workload, with raw instruction-rate figures overstating real-world matrix math improvements.
- The unlock works by rewriting GPU registers before Nvidia's driver loads, requiring no VBIOS flash or driver patch, but its long-term compatibility and risk profile remain unclear.
NVIDIA GeForce RTX 3090 24GB GPU — The RTX 3090 is the card at the center of this story, prized by AI hobbyists for its generous 24GB VRAM at a fraction of workstation-card cost. If you're experimenting with local AI models or these new tensor-core unlock mods, this is the GPU that makes it all possible.
See NVIDIA GeForce RTX 3090 24GB GPU on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: techspot.com — Zo Ahmed, 2026-10-27
Published there as: “Modders unlock hidden AI power on Nvidia RTX 20-, 30-, and 40-series GPUs, nearly doubling some benchmarks”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.