AMD introduces Stable Diffusion 3.0 Medium AI model generator specialized for XDNA 2 NPUs, allowing for local operation on Ryzen AI laptops within the industry.
AMD and Stability AI have unveiled the Stable Diffusion 3.0 Medium AI model, a text-to-image generator designed to run locally on laptops based on AMD's Ryzen AI laptops. The model leverages the power of AMD's XDNA 2 Neural Processing Units (NPUs) and the BF16 (bfloat16) precision format for efficient and high-quality image generation, without the need for cloud dependency.
Requirements for Usage
To use this model, your system must meet certain requirements:
- The CPU should be an AMD Ryzen AI 300-series or Ryzen AI MAX+ processor, equipped with a built-in XDNA 2 NPU.
- The XDNA 2 NPU must be capable of delivering at least 50 TOPS (trillions of operations per second).
- A minimum of 24GB of system RAM is required, as the model alone uses approximately 9GB during generation.
Installation and Usage Steps
- Platform and Software: The model is available through Amuse 3.1 (or later) by Tensorstack, which serves as the interface to run the Stable Diffusion 3.0 Medium model on Ryzen AI laptops.
- Download & Install Amuse 3.1 Beta: Obtain the Amuse 3.1 Beta client compatible with your Ryzen AI laptop and install it following standard application installation procedures.
- Enable NPU-Optimized Settings: Within Amuse 3.1, switch to HQ mode (High Quality mode) and toggle the setting labeled “XDNA 2 NPU Stable Diffusion 3” to enable the AMD optimized pipeline leveraging the NPU.
- Run The Model Locally: Input text prompts; the model generates images at 1024×1024 resolution. Images are then upscaled via an NPU-powered two-stage pipeline to 2048×2048 (4MP) resolution, suitable for professional and print use.
- Offline Operation: The whole system runs locally, requiring no internet or cloud GPU use. This allows fast, private, and customizable image generation directly on the device.
Summary
| Requirement | Details | |-----------------------------|----------------------------------------------------------| | CPU | AMD Ryzen AI 300-series or Ryzen AI MAX+ processors | | NPU | XDNA 2 NPU with ≥50 TOPS capability | | RAM | Minimum 24GB system RAM (model uses ~9GB during runs) | | Software | Amuse 3.1 Beta for running the Stable Diffusion 3.0 Medium model | | Image Output | 1024×1024 base images, upscaled to 2048×2048 (4MP) |
The combination of optimized BF16 precision, XDNA 2 NPU acceleration, and local execution enables efficient text-to-image generation directly on high-end Ryzen AI laptops without relying on cloud resources. The model runs entirely on-device, enabling fast, offline image generation without the need for Internet access or cloud services.
The Stable Diffusion 3.0 Medium AI model is available now via Amuse 3.1. The usage of the model is subject to the Stability AI Community License. AMD provides examples of prompts for the model, such as a prompt to draw a toucan. It is important to note that Amuse is still in beta, so its stability or performance may vary.
The model is aimed at content creators and designers who need customizable images and supports advanced prompting features for fine control over image composition. During generation, the model uses 9GB of system memory. The model interprets written prompts and produces 1024×1024 images, then upscales them to 2048×2048 resolution, resulting in 4MP outputs.
The model is free for individuals and small businesses with less than $1 million in annual revenue, but licensing terms may change eventually. To use the Stable Diffusion 3.0 Medium AI model, users need to install the latest AMD Adrenalin Edition drivers and the Amuse 3.1 Beta software from Tensorstack. The model is suitable for generating customizable stock-quality visuals for design and marketing applications.
The AMD Ryzen AI laptop's local execution of the Stable Diffusion 3.0 Medium AI model, driven by advanced technology like data-and-cloud-computing, artificial-intelligence, smartphones, and gadgets, ensures efficient text-to-image generation without cloud dependency. This technology empowers content creators and designers to generate customizable, high-quality images directly on their devices, without relying on internet access or cloud services.
