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.
techspot.com
· 2026-10-27
Modders have released firmware exploits that remove Nvidia's artificial half-speed limit on FP32-accumulate matrix math for RTX 20, 30, and 40-series GeForce cards, according to VideoCardz. The unlock lets the RTX 4090 reach double its normal tensor rate and boosts the RTX 3090 Ti's matrix math by up to 80%, though gains vary by card and chatbot token speeds stay roughly the same since those depend on memory bandwidth rather than compute. Three separate projects—gpu-unlock, gpu-unlock-ada, and turing-unlock—provide unofficial Linux kernel modules covering different GPU generations, with experimental Windows builds also in development.
tomshardware.com
· 2026-10-26
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%.
github.com
· 2026-10-24
GPU prices are climbing sharply, with NVIDIA's RTX 50-series up as much as 39 percent and AMD cards rising up to 10 percent, driven by demand from AI companies and individuals building local AI setups. Cards with large VRAM capacities, such as the older NVIDIA GeForce RTX 3090, are commanding strong resale value because they can handle AI workloads despite their age.
engadget.com
· 2026-09-26
Alibaba Cloud has launched Qwen Image 2.1, an open-weight image generation model with only 7 billion parameters, small enough to run on older consumer GPUs like the RTX 3090. The model adds native transparent-background generation and can combine up to 10 reference images into a single consistent output, and Alibaba's own benchmarks claim it outperforms closed models including Google's Nano Banana 2.0. Unlike its predecessor, the new licensing terms bar commercial resale without a separate agreement from Alibaba.
tomshardware.com
· 2026-09-23
GoKawiil