I quote that figure off the top of my head.
1.2ms A-to-D is about as fast as anyone is getting it done... to the best of my knowledge.
Add another 1.2ms D-to-A and that is the other half of the 2.4ms.
I'll assume that the dsp process can actually happen faster than real time so I didn't add any for that and called it a "good day".
As always, I welcome corrections.
When mixing I don't think latency, even 46ms, is an issue until your start turning a knob and "use your ears" to hear stuff that doesn't even happen for, well.. let's say another 46ms.
Watching guys do that while they are acting like they are hearing the subtlety of a small knob twist cracks me up.
Good times!
all the best,
mike