AI BriefingStable Diffusion, Optimized Support for AMD Radeon™ GPUs and Ryzen™ AI APUs
Key point
In collaboration with AMD, ONNX-optimized versions of Stable Diffusion models were released, optimized for AMD Radeon GPUs and Ryzen AI APUs.
Details
In collaboration with AMD, ONNX-optimized versions of some models in the Stable Diffusion model family are now provided. This allows models to run faster and more efficiently in AMD Radeon™ GPU and Ryzen™ AI APU environments.
The optimized models are available on Hugging Face with an _amdgpu suffix, and users can test them immediately via Amuse 3.0. This work focused on maximizing inference performance while maintaining the models' output quality and open licenses.
The key performance improvements by model are as follows:
- Stable Diffusion 3.5 (SD3.5) Large and Large Turbo: Up to 2.6x faster inference performance compared to the base PyTorch model.
- Stable Diffusion XL (SDXL) 1.0 and Turbo: Up to 3.8x faster inference performance compared to the base PyTorch model.
These models can be immediately integrated into any environment that supports ONNX Runtime, allowing them to be applied directly to existing workflows to boost productivity.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Comfy Launches Comfy Router: A Unified API for Frontier Media Models
Amazon Launches 'CloudWatch Omni' GA for Observability of Generative AI and Agent Workloads