Does Stable Diffusion Work With AMD? What You Need to Know (2024)

Stable Diffusion, a revolutionary artificial intelligence (AI) image generation model, has been making waves in the tech community recently. This open-source machine learning (ML) model can use a text prompt to generate astounding images. However, by default, it requires Nvidia’s CUDA framework and an Nvidia GPU. But does Stable Diffusion work with AMD? Let’s find out!

This article will guide you through the process of running Stable Diffusion using an AMD GPU, specifically on a Windows PC.

Understanding Stable Diffusion

Stable Diffusion is an innovative neural model capable of creating high-quality images from input data. It’s a diffusion model, a generative model that mimics the behavior of variables over time using a stochastic process.

The model is trained on large text and image datasets, allowing it to generate visually coherent and semantically meaningful images.

Does Stable Diffusion work with AMD?

Yes, Stable Diffusion can work with AMD GPUs. In fact, AMD has released driver support for a metacommand implementation that can improve performance and reduce the time it takes to generate output from the model. Using the recently released AMD Software: Adrenalin Edition 23.5.2, we can observe an average 2X performance improvement with Stable Diffusion 1.5 on an AMD Radeon™ RX 7900 XTX graphics card.

However, there are some caveats to using Stable Diffusion with AMD GPUs. First, you will need to use a specific version of the model that has been compiled for AMD GPUs. Second, you may need to use a specific version of the Stable Diffusion WebUI that is compatible with AMD GPUs.

Looking to craft the best Stable Diffusion prompts? Check out our detailed blog to be updated on the dos and don’ts.

Compatibility of Stable Diffusion with AMD

According to the Stable Diffusion website, most Nvidia and AMD GPUs with 6GB or more of video memory should be able to run Stable Diffusion without issues. So, if you have an AMD graphics card, you should be able to run Stable Diffusion seamlessly. Specifically, the following AMD GPUs are known to run Stable Diffusion effectively:

• RX 470
• RX 570
• RX 6600 XT
• RX 6650 XT
• RX 6600
• RX 6700 XT
• RX 6750 XT
• RX 6800
• RX 6800 XT
• RX 6900 XT

Minimum system requirements

To run Stable Diffusion effectively, your system should meet the following minimum requirements:

• A 64-bit operating system.
• At least 8 GB RAM.
• At least 10 to 12 GB of hard drive or solid-state drive (SSD) space.

Check out to know whether you can access Stable Diffusion for free!

Preparing the workspace

Before you start the process of setting up Stable Diffusion to work with your AMD GPU, ensure that you have the following:

• A powerful AMD GPU with at least 6GB of video memory.
• A working Python installation, at least version 3.7.
• The willingness to download around six gigabytes of ML model data.
• A Hugging Face account.
• A working installation of Git for storing login credentials.

Steps to run Stable Diffusion on AMD GPU

Once you have all the necessary requirements, follow the steps below to run Stable Diffusion on your AMD GPU.

Step 1: Install dependencies

Firstly, create a new folder (let’s call it stable-diffusion), and open a command line in the folder. We are going to create a virtual environment to install some packages. Use the venv package to create a virtual environment named virtualenv.

Activate your virtual environment using the following commands (depending on your shell):

For PowerShell:

./virtualenv/Scripts/Activate.ps1

For cmd.exe:

virtualenv\Scripts\activate.bat

Next, install the necessary Python packages into your virtual environment using the following commands:

• pip install diffusers==0.3.0
• pip install transformers
• pip install onnxruntime

Looking to use Stable Diffusion on your device but don’t know how? Check out our extensive blog to find out how to run Stable Diffusion locally on your PC.

Step 2: Download and install DirectML Onnx runtime

Next, download a build of Microsoft’s DirectML Onnx runtime. You can get it from here. Download the package that corresponds to your installed Python version. Once downloaded, use pip to install the package. Make sure you use the –force-reinstall flag to override some previously installed dependencies.

Step 3: Download and convert the Stable Diffusion model

Download a utility script from here, which will automatically download the Stable Diffusion model, convert it to Onnx format, and put it somewhere useful.

Authenticate your shell with your Hugging Face API key by running the following command:

huggingface-cli.exe login

And paste in your token when prompted.

Once authenticated, run the utility script to download and convert the Stable Diffusion model into Onnx format. The script will also place the converted model in a specified output path.

python convert_stable_diffusion_checkpoint_to_onnx.py –model_path=”CompVis/stable-diffusion-v1-4″ –output_path=”./stable_diffusion_onnx”

Step 4: Run Stable Diffusion with AMD

Finally, create a new Python file named text2img.py and write a script that runs the Stable Diffusion model.

from diffusers import StableDiffusionOnnxPipeline

pipe = StableDiffusionOnnxPipeline.from_pretrained(“./stable_diffusion_onnx”, provider=”DmlExecutionProvider”)

prompt = “A happy celebrating robot on a mountaintop, happy, landscape, dramatic lighting, art by artgerm greg rutkowski alphonse mucha, 4k uhd'”

image = pipe(prompt).images[0]

image.save(“output.png”)

Run the script with the command python .\text2img.py, and it will generate an image named output.png based on the text prompt you provided.

Unlock the mesmerizing world of AI-generated artistry and unleash your creativity with a brushstroke of innovation by exploring our curated list of cutting-edgeAI art tools.

Frequently asked questions(FAQs)

Q: Can Stable Diffusion run without a GPU?

A: Yes, Stable Diffusion can run without a GPU, relying solely on CPU power. OpenVINO, a tool developed by Intel, can be used to run Stable Diffusion without a GPU.

Q: What are some of the alternatives to Stable Diffusion?

A: If for any reason you can’t use Stable Diffusion, there are many alternatives available such as Midjourney, DreamStudio, Playground AI, and DRAI, among others.

Wrapping up

Though Stable Diffusion was initially designed for Nvidia GPUs, with a bit of tweaking, it’s possible to run it on AMD GPU hardware. However, remember that your AMD GPU should have at least 6GB of video memory for an optimal experience.

Also, now that you know how Stable Diffusion works with AMD GPU, ensure that you meet the minimum system requirements before running Stable Diffusion on your system.