How to use GPU Unlock: what you can do with it

3 use cases

• Speed up local LLM prompt processing on an RTX 30 card

  Install the gpu_unlock module so FP16 and BF16 tensor math runs at full rate for vLLM, PyTorch and Transformers.

  • REPO Zagitta/gpu-unlock
  • Security and science
  • Advanced
  View prompt

  Walk me through installing Zagitta/gpu-unlock on this <distro> machine with an <RTX card>: build the module, add gpu_unlock.targets=all to the kernel command line for my bootloader, and give me a before and after benchmark to run.

  Share Email me this

• Run BF16 LoRA fine-tuning faster on a GeForce card

  Unlock full tensor rate before a long LoRA or full fine-tune so training hours drop.

  • REPO Zagitta/gpu-unlock
  • Security and science
  • Advanced
  View prompt

  Help me use Zagitta/gpu-unlock before a BF16 LoRA run on my <card>: confirm my card is supported, install it, and write a short timing script to compare steps per second before and after.

  Share Email me this

• Generate SDXL and ComfyUI images faster on RTX cards

  Remove the FP16 tensor cap so diffusion models in ComfyUI render faster.

  • REPO Zagitta/gpu-unlock
  • Security and science
  • Advanced
  View prompt

  Help me install Zagitta/gpu-unlock and measure ComfyUI SDXL render time on my <card> before and after with the same seed and workflow.

  Share Email me this

Quick start with GPU Unlock

1. Check you have what it needs: Linux (Windows is not supported; the README points to a third party port); An Ampere RTX 30 card (stable) or an Ada RTX 40 or Turing RTX 20 card (experimental); RTX 50 is not supported; Root access to install a kernel module and edit the kernel command line.
2. Open Zagitta/gpu-unlock on GitHub and read the README and license.
3. Run git clone https://github.com/Zagitta/gpu-unlock, cd gpu-unlock and make install
4. Append gpu_unlock.targets=all to your kernel command line (GRUB, systemd-boot or UKI, as the README shows)
5. On Debian or Ubuntu run sudo update-grub, then reboot
6. Rerun your usual vLLM or Transformers benchmark and compare prefill speed

These steps come from our top use case, Speed up local LLM prompt processing on an RTX 30 card, which also has a copy-paste agent prompt.

Using the Deduped MCP, get the setup steps for Zagitta/gpu-unlock

claude mcp add --transport http deduped https://dedupedai.com/mcp

Best AI repos for Security and science

About this repo

• Who it is forLinux users running local AI training, fine-tuning or image generation on RTX 20, 30 or 40 series GeForce cards.
• Why it is hot14k stars, launched 13 days ago, talked about in 2 AI sources we track
• LanguageC
• Sourcegithub.com/Zagitta/gpu-unlock · Back to Hot repos

Requirements

• Linux (Windows is not supported; the README points to a third party port)
• An Ampere RTX 30 card (stable) or an Ada RTX 40 or Turing RTX 20 card (experimental); RTX 50 is not supported
• Root access to install a kernel module and edit the kernel command line

Related repos

• Trended earlier REPO JustAResearcher/gpu-unlock-ada GPU Unlock Ada GPU Unlock for Ada is a Linux kernel module that runs at boot, before the stock NVIDIA driver loads, and switches supported RTX 40 cards to full-rate FP16 and BF16 tensor math with FP32 accumulation. The authors measured about 2x in instruction benchmarks on an RTX 4090 (165 to 331 TFLOPS) and smaller gains in real GEMM workloads.