Interesting links, thanks!
I liked the blueprint for a home-made MIDI->audio cable that allows testing latency compared to incoming MIDI signal in the first article.
For excessive latency of various sound interfaces, i find this
link very useful - it shows roundtrip latencies measured for many popular sound interfaces.
Even though the roundtrip latency (delay between sending a signal from a program to a soundcard and receiving it back, using a loopback cable plugged between inputs and outputs on the soundcard) is not exactly the latency relevant for virtual pianos - because of the complete ("out"+"in") roundtrip latency only the "out" part is part of the overall latency when we use VSTs, it still IMHO shows very well how well a particular sound interface is designed and what potential delay it can cause.
Also, It's easy to measure the roundtrip latency of one's sound interface at home using this
What I'm personally concerned with more, is the inherent slowness of MIDI - 1 event takes 1ms to transfer. Does it mean that if I press e.g. 8 keys at the same time, it will take 8 ms to get the last note-on event to the other side? That sounds quite disconcerting to me, more than the more often discussed quantization of MIDI velocities to values 0-127...
I mean, how great is it to get the latency in the computer to extemely low levels, if even the midi cable sees the last note played in a chord only after e.g. 8ms?
I always wonder if instruments connected through USB have the same limitation, because if not, that would IMHO make USB a better interface than MIDI.