Running ComfyUI with ZLUDA on AMD GPUs: A Seamless Experience

Running ComfyUI with ZLUDA on AMD GPUs: A Seamless Experience

Running ComfyUI with ZLUDA on AMD GPUs: A Seamless Experience

If you’re an AI enthusiast or creative who uses ComfyUI for Stable Diffusion workflows but own an AMD GPU, you might have felt limited by the lack of native CUDA support. The good news? Thanks to ZLUDA, you can now run ComfyUI efficiently on AMD cards like the Radeon RX 6800 XT — and I’ve tested it myself with great results.

What Is ZLUDA and Why Does It Matter?

ZLUDA is a compatibility layer that translates CUDA API calls to work on AMD GPUs. This opens up the rich ecosystem of CUDA-dependent AI tools — such as ComfyUI — to AMD users, who traditionally had to rely on slower CPU runs or less optimized frameworks.

My Setup and Installation

Following the excellent guide from the Stable Diffusion Info Wiki, I configured ComfyUI to run with ZLUDA on my AMD Radeon RX 6800 XT. The process was straightforward:

  • Installed ZLUDA as the CUDA compatibility layer
  • Disabled incompatible options like xformers in ComfyUI
  • Set environment variables as recommended
  • Launched ComfyUI and verified GPU usage

How It Performed

  • Stable GPU acceleration: ComfyUI ran smoothly, leveraging my AMD GPU’s compute power.
  • Good speed: Image generation was fast and consistent, comparable to what I expected from an NVIDIA setup.
  • No major errors or crashes: The setup felt reliable for both casual use and batch generation.

Why You Should Try This

  • If you own an AMD GPU, this is a practical way to get full GPU acceleration for ComfyUI without switching hardware.
  • It unlocks a powerful, user-friendly interface for Stable Diffusion on AMD.
  • You can experiment with complex pipelines and attention mechanisms without performance penalties.

Tips for a Smooth Experience

  • Follow the installation guide carefully—small tweaks like disabling xformers or adjusting batch sizes can make a difference.
  • Use the latest ZLUDA builds for best compatibility.
  • Keep an eye on updates from the ComfyUI and ZLUDA communities—they’re actively improving AMD support.

Final Thoughts

Running ComfyUI with ZLUDA on an AMD 6800 XT GPU worked better than I expected. For AMD GPU owners wanting to dive into AI art generation without compromise, this is a real breakthrough.

Want to get started? Check out the official guide here, and let me know if you have questions or want tips on optimizing your setup.

 

Here are some of the images I created using Free Diffusion Models like Dreamshaper, Flux Dev/S:

Back to blog