🚀 ComfyUI X AMD (Windows)

A streamlined PowerShell script to set up ComfyUI with ROCm support for AMD RDNA graphics cards on Windows. Uses UV for blazing-fast package management.

✨ Features

• 🎮 Full AMD GPU Support: RDNA 3, RDNA 3.5, and RDNA 4
• ⚡ Lightning Fast: Uses UV for 10-100x faster package installation
• 🔄 Automatic Setup: One command to rule them all
• 💾 Persistent Configuration: Remembers your GPU settings
• 🎨 Beautiful CLI: Colorful interface with helpful emojis
• 🔧 Zero Config: Handles virtual environments, dependencies, and ROCm PyTorch automatically

🎯 Supported GPUs

Generation	GPUs	Architecture
🔴 RDNA 3	RX 7900 XTX, RX 7900 XT, RX 7800 XT, RX 7700 XT, RX 7600	gfx1100
🟣 RDNA 3.5	Ryzen AI (Strix Point/Halo), Radeon 890M, 8060S	gfx1151
🔵 RDNA 4	RX 9070 XT, RX 9070	gfx1200

📋 Prerequisites

• Windows 10/11 (64-bit)
• Git (Download)
• PowerShell 5.1+ (included in Windows)

⚡ Quick Install

Open PowerShell and run this one-liner:

irm https://raw.githubusercontent.com/SystemVll/ComfyUI-AMD/main/ComfyUI-AMD.ps1 | iex

Alternative: Manual Installation

1. Download the script:

   git clone https://github.com/SystemVll/ComfyUI-AMD.git
   cd YOUR_REPO

2. Run the setup:

   .\ComfyUI-AMD.ps1

🎬 What It Does

1. ✅ Installs UV (if not present)
2. ✅ Clones ComfyUI repository
3. ✅ Creates Python 3.12 virtual environment
4. ✅ Installs ROCm PyTorch for your specific GPU
5. ✅ Installs all ComfyUI dependencies
6. ✅ Configures GPU settings (HIP_VISIBLE_DEVICES)
7. ✅ Launches ComfyUI web interface

🔧 Configuration

GPU Selection

On first run, you'll be prompted to select your GPU generation:

• 1 - RDNA 3 (RX 7000 series)
• 2 - RDNA 3.5 (Ryzen AI)
• 3 - RDNA 4 (RX 9000 series)

Your choice is saved and reused on subsequent runs.

Multi-GPU Systems

If you have multiple GPUs, the script will ask you to configure HIP_VISIBLE_DEVICES:

• 0 - First GPU
• 1 - Second GPU
• 0,1 - Use both GPUs

📁 Directory Structure

your-folder/
├── setup.ps1                 # This setup script
├── .gpu_config.json         # Saved GPU configuration
└── ComfyUI/                 # ComfyUI installation
    ├── .venv/               # Virtual environment
    ├── models/              # Place your models here
    ├── custom_nodes/        # Custom nodes
    └── output/              # Generated images

🎨 Usage

First Run

.\setup.ps1

Select your GPU, configure settings, and ComfyUI will launch automatically.

Subsequent Runs

.\setup.ps1

Just hit Enter through the prompts - your settings are remembered!

Access ComfyUI

Once launched, open your browser to:

http://127.0.0.1:8188

🔄 Updating

To update ComfyUI to the latest version:

cd ComfyUI
git pull
cd ..
.\setup.ps1

The script will handle any new dependencies automatically.

📦 What's UV?

UV is a modern, extremely fast Python package manager written in Rust. Benefits:

• ⚡ 10-100x faster than pip
• 🎯 Better dependency resolution
• 💾 Efficient caching
• 🔒 More reliable installations

🐛 Troubleshooting

"UV not found" after installation

• Close and reopen your PowerShell terminal
• Run the script again

"Execution policy" error

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

GPU not detected

• Ensure your AMD GPU drivers are up to date
• Check that ROCm is supported on your GPU model
• Try setting HIP_VISIBLE_DEVICES=0 manually

ComfyUI won't start

• Check that no other instance is running on port 8188
• Look for error messages in the console
• Try deleting .venv folder and running setup again

🔗 Useful Links

• ComfyUI Documentation
• ROCm Documentation
• UV Documentation
• Model Downloads

📝 Notes

• First installation may take 5-10 minutes depending on your internet speed
• ROCm PyTorch builds are large (~2-3GB)
• Models are not included - download them separately to ComfyUI/models/

🤝 Contributing

Found a bug? Have a suggestion? Open an issue or submit a PR!

📄 License

This setup script is provided as-is. ComfyUI has its own license - please refer to the ComfyUI repository for details.

⭐ Credits

• ComfyUI - The amazing UI for Stable Diffusion
• UV - Modern Python package management
• AMD ROCm - GPU compute platform

---

Made with ❤️ for the AMD + AI community