Generating 512*512 or 768*768 images using SDXL text to image model

Hi, from the document I find that SDXL1.0 text to image model was trained using images with resolution 10241024. In order to save the generation time and the memory usage in my GPU, I want to generate images using the resolution of 512 512 or 768768. Will there be any side effect when generating images with resolution smaller than 1024 1024?