How Different are Pre-trained Transformers for Text Ranking?