GPU Unlock Ada
JustAResearcher/gpu-unlock-adaReviewed by AI
Radar found strong risk signals in this project. It stays visible so you can read the evidence, but it is left out of rankings and recommendations.
What it does
A kernel module for Linux, for people who are comfortable compiling software. It loads before the NVIDIA driver, writes to the card's speed-select registers and lifts the limit NVIDIA puts on FP16, BF16 and FP8 maths with FP32 accumulate: the RTX 4090 figures in the README go from about 165 to about 331 TF/s on instruction rates, while real application gains are reported as much smaller or absent. You build packages from source, add options to the kernel command line with your card's PCI address, and cold-boot. Only the RTX 4090 is treated as supported.
How easy it is to install
E · developers only · Linux
Linux is the only system it runs on and it is a job for developers: clone the repository, run make install for your distribution (kernel headers and a working DKMS toolchain are required), let the build download and generate the pinned NVIDIA firmware inputs, then edit the kernel command line with the exact PCI address from lspci -D and power the machine fully off and on. The Windows folder is a report of one manual boot and ships no binaries or installer.
Coming soon: install in 1 click with the RepoPicks app.
Numbers on GitHub
- Stars
- 4,626on Oct 31, 2026
- Downloads
- 4,4881,542 this week
- Growth
- +790stars in 5 days
Stars over time
Facts
- Runs on
- Linux
- You need to install
- Python, Other tools
- To build from source
- Python
- Disk space
- 100 MB–300 MB · estimate
- Accounts or keys
- None
- Cost
- Free
- License
- GPL-2.0 (Copyleft)
- Last update
- Oct 26, 2026
Risk signals (5)
- Highbypasses vendor firmware limit
This is the project's whole purpose. README: "This Linux kernel module enables the SM speed-select override before the stock NVIDIA driver starts"; the RTX 4090 path "writes SS0 0x88888888 and SS1 0x8, then performs FLR before the unmodified NVIDIA driver binds", and "The payload uses instruction addresses derived from published decrypted debug-booter analysis" of NVIDIA's signed firmware. Removing a limit the maker built into the card's firmware goes against NVIDIA's terms and can count as circumventing a technical protection, which is restricted in some countries.
AI review
- Attentioncan leave the machine or card in a bad state
README, Check and disable: "If the module reports that firmware did not stop and its DMA buffers are being kept, power off before trying again... Do not force-unload it in this state." It also says to "Keep an unmodified stock entry available", that the display stays blank until nvidia-drm takes over, and that an unlocked RTX 4090 "peaked at 346 W against 313 W stock".
AI review
- Attentionanonymous author new repository kernel code
Owner account "JustAResearcher"; the repository was created 2026-10-23 and already has 4276 stars and 556 forks. The code runs inside the Linux kernel with direct hardware access (BAR0, DMA) before the GPU driver loads, and the payload comes from a third-party upload on Hugging Face (TOTORONG/AD103_win_unlocker).
AI review
- Inforesults from single cards
README, Experimental Ada targets: "Results come from one or a few cards per ID; other boards, memory configurations and laptop models remain unverified", and "Laptops differ in VBIOS ROM exposure, embedded-controller power rails and FLR handling, so a result on one model does not carry over to another."
AI review
- Infofirmware downloaded at build time
README, Build and install: "Builds fetch the pinned NVIDIA 610.43.03 source and binary inputs over HTTPS and verify their SHA-256 hashes" and "Prebuilt firmware packages are not distributed by this repository", so NVIDIA material is pulled onto your machine during the build.
AI review
Evidence is shown in its original language.
What Radar could not check
- Whether your exact card, board maker and laptop model unlock at all
- Whether the extra speed shows up in the programs you actually use, beyond the register-level numbers
- Long-term stability, heat and power draw after the change
- Whether the card still works normally after a driver or kernel update
- What the change does to your warranty and to vendor support