3 ms·

IMO if the trade off is cheaper, faster hardware iteration then Nvidia’s strategy makes a lot of sense.