5 ms·

efficiency per intelligence is the benchmark i look at the most, as that allows the most use by most people.