On Fri, 12 Oct 2001, Ivica Bukvic wrote: > 2) I am interested in what is the current latency achievable not only for > DSP stuff, but also for a MIDI aspect as well, since the last up-to-date > benchmarking of this sorts I was able to locate on the net is over a year > and a half old and not very comprehensive.
using benno's mididelay, i get about 1.12 ms average, max 1.135966 ms on my m-audio audiophile using using an at-keyboard cable (and not a quality midi cable). i get the same results when using two cables and routing through my yamaha sy85 keyboard. sometime soon i'll try it with my roommate's high-quality midi cables. benno says : Tn theory since the MIDI interface does run at 31kbps we assume that one byte takes about 0.3ms to transmit, therefore a full note-on =3 bytes should not take more than 1 - 1.5ms well, i'm at the low end of that range. however, he's done some bad rounding in his math. 31kbps = 1 byte per .252 ms, so .756 ms minimum for three bytes. he says that the awe64 (isa) sucks as a midi card (it sucks as an audio card as well - tons of white noise [turn yours up and listen]). you could probably route your midi cables through another computer (say, running windows) and test it's latency as well. i'd do it but my roommate has more problems with all his proprietary pro hardware and software than i do with getting alpha and cvs programs to work reliably! btw, i'm using alsa cvs from october first on an ll-patched 2.4.8 kernel. -dave
