GoKawiilLinux kernel module unlocks full FP/BF16 throughput on Nvidia RTX 3000 GPUs
A new open-source Linux kernel module called GPU-unlock removes Nvidia's artificial cap on FP16/BF16 instruction issue rates for GeForce RTX 3000-series cards based on the GA102, GA103, GA104 and GA106 chips, bringing them up to the rate normally reserved for professional cards. It works by writing specific speed-selector registers and performing a function-level reset before the stock Nvidia driver loads, avoiding the need for manual driver patching after each update. Community-reported benchmarks on cards like the RTX 3090 Ti, 3090, 3080 and 3070 show BF16 GEMM throughput gains ranging roughly from 25% to over 80%.
GoKawiil's interpretation of the reporting above, not reported fact.
This effectively lets gamers and hobbyists access compute performance that Nvidia otherwise segments into pricier professional and datacenter products, which could pressure the company's product-tiering strategy if adoption spreads. Because the tool operates as a kernel module that intercepts initialization before the official driver binds, it may prove more durable than earlier UEFI-based unlock methods that required re-patching with every driver release. However, results vary significantly by card and are self-reported by users, so real-world reliability and Nvidia's potential response remain uncertain.
- The module doubles FP/BF16 FMA-to-FP32 instruction issue rate on GA102/103/104/106-based RTX 3000 GPUs.
- Unlike prior UEFI-based unlocks, it persists across Nvidia driver updates without manual repatching.
- Community benchmarks report BF16 GEMM throughput gains from about 25% to over 80%, though some cards like the 3080 Ti 20GB OEM remain only partially unlocked.
NVIDIA GeForce RTX 3090 Ti — This is the exact card benchmarked in the article, showing massive BF16/TF32 throughput gains with GA102 unlock tricks. If you're tinkering with GPU compute unlocks for deep learning workloads, the 3090 Ti's GA102 silicon and 24GB VRAM make it a prime candidate to replicate these results.
See NVIDIA GeForce RTX 3090 Ti 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: github.com, 2026-10-24
Published there as: “GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards”
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.