Getting Started with ComfyUI on AMD Radeon™ RX 9000 Series GPUs

2026-03-09 · Source: AMD ROCm Blogs · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, medium

Summary

AMD has released a guide for setting up ComfyUI, a node-based interface for Stable Diffusion and other generative AI models, on AMD Radeon RX 9000 series GPUs. This guide addresses common stability issues like HIP memory errors, slow first-time generation, and VAE decoding failures that previously frustrated AMD GPU users. It outlines hardware requirements, recommending AMD Radeon RX 9000 Series GPUs with 16GB+ VRAM, 32GB+ system RAM, and 100GB+ SSD storage. Software requirements include Ubuntu 22.04/24.04 LTS, ROCm 7.1, Python 3.10-3.12, and PyTorch 2.6.0+ with ROCm support. The guide provides detailed installation instructions for both Docker and native Linux setups, along with specific solutions and recommended launch configurations to optimize performance and stability on AMD Radeon hardware.

Key takeaway

For AI Engineers and content creators using AMD Radeon RX 9000 series GPUs for generative AI, this guide provides essential configurations to overcome historical stability issues. You should prioritize using the latest ComfyUI version and ROCm 7.1, and apply the recommended `--reserve-vram` and `--lowvram` flags based on your GPU's VRAM capacity. This will ensure a more stable and performant experience for image and video generation workflows.

Key insights

ComfyUI is now stable and performant on AMD Radeon RX 9000 series GPUs with ROCm 7.1.

Principles

• VRAM management is critical for AMD GPUs.
• MIOpen compiles kernels once, improving subsequent runs.
• Tiled VAE decoding is a normal memory fallback.

Method

Install ComfyUI via Docker or natively on Ubuntu with ROCm 7.1, then apply specific launch flags like `--reserve-vram` and ensure the latest ComfyUI version for optimal stability and performance.

In practice

• Use `--reserve-vram 3` for 16GB VRAM GPUs.
• Update ComfyUI to disable `torch.backends.cudnn.enabled`.
• Monitor VRAM with `rocm-smi --showmeminfo vram`.

Topics

• ComfyUI
• AMD Radeon GPUs
• ROCm
• Generative AI Workflows
• Stable Diffusion

Code references

• comfyanonymous/ComfyUI
• comfyanonymous/ComfyUI

Best for: Machine Learning Engineer, AI Engineer, Software Engineer

Related on AIssential

• NVIDIA and ComfyUI Streamline Local AI Video Generation for Game Developers and Creators at GDC — NVIDIA's GDC updates enhance local AI video generation with simplified interfaces and significant performance gains on RTX GPUs.
• ComfyUI Update: Stable Video Diffusion on 8GB vram with 25 frames and more. — ComfyUI's latest update enhances video generation, sampling efficiency, and high-resolution image consistency.
• INSANE NEW FREE NSFW IMAGE AI KING IS HERE! LESS THAN 8GB VRAM! — Z Image Turbo is a 6B parameter, uncensored text-to-image model offering high quality and speed on low VRAM.
• ComfyUI Update: What’s new from the last few weeks, SD Turbo, Stable Zero123, Group Nodes, FP8, and more. — ComfyUI updates enhance performance, memory efficiency, and creative capabilities through new models and UI features.
• How to Build, Run, and Scale High-Quality Creator Workflows in ComfyUI — ComfyUI on NVIDIA RTX GPUs enables local, accelerated generative AI workflows for creative professionals.
• Image-to-3D: Incremental Optimizations for VRAM, Multi-Mesh Output, and UI Improvements — Optimizing image-to-3D pipelines involves VRAM reduction, multi-object generation, and UI enhancements for improved workflow.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AMD ROCm Blogs.