You are correct for ALSA == the kernel driver model. But using "alsa" as a userspace audio stack has its own latency, dmix in particular. A highly optimized .a

You are correct for ALSA == the kernel driver model. But using "alsa" as a userspace audio stack has its own latency, dmix in particular. A highly optimized .asoundrc might be competitive. This feels to me like background knowledge from long ago, but I can't provide any specific source other than anecdotes like https://forum.audacityteam.org/viewtopic.php?p=234235#p234235 https://forum.audacityteam.org/viewtopic.php?p=234235#p23423... .