Some people shared their result onnxruntime actually getting faster: https://medium.com/microsoftazure/accelerating-stable-diffusion-inference-with-onnx-runtime-203bd7728540