although this is really good news i'm curious ... which sample rate have you been using? at 44100 and 2 * 64 samples it should only be 2.9 ms ... (2.6 with 48000) ... possibly we lose another 64 samples somewhere in the dsp tick (miller?)
I think that's ok since it's measured ADA latency... normally AD and DA conversion lose about 1 ms each. I'll check later with digital loopback.
best greetings, Thomas