3 ms·

I think the current generation is more in line with compute costs to the point that this wouldn't make too much sense