I've heard screen readers going way faster (I don't know quantitatively) than what it sounds like when I listen to videos at 2x.

I also think it's important that the screen readers use a "old school" synthesized voice, instead of a voice that sound natural, because each of the phonemes is articulated distinctly.