2 ms·

Since their precision is essential to measuring relativistic effects, I'm not sure we're near that limit. For your precise question, it may already be there.