[osxmidi]

Did some more latency testing, this time with a basic Edirol UA-1EX USB device and a M-Audio Transit usb device on a AMD quad core (instead of using the inbuilt motherboard sound).

FM8 and the Mercuriall Spark amp sim get no xruns at 48000Hz/64 samples (16 bit) 2.6ms latency on Debian 9 with the realtime kernel.

M-Audio Transit usb gets 2.6ms latency at 48000hz (duplex).

The M-Audio Transit can either do 96000hz playback or 96000hz record (not duplex) and jackd needs the -P or -C options.

jackd --realtime -d alsa -P -d hw:1 -p256 -n2 -r96000 // playback

jackd --realtime -d alsa -C -d hw:1 -p256 -n2 -r96000 // record

The Edirol UA-1EX USB device can get 2.6ms latency at 96000hz and probably could go lower, I had to work the driver out as it uses a non standard 24 bits in it's advanced mode.

The Edirol isn't duplex at 96000hz either.

For anyone with the Edirol, setting it up for 96000hz

~/.asoundrc

pcm.!default
{
type plug
slave sl3
}

ctl.!default
{
type hw
card 1
}

pcm_slave.sl3
{
pcm "hw:1,0"
format S24_3LE
channels 2
rate 96000
}

pcm.96000
{
type plug
slave sl3
}

then

jackd --realtime -d alsa -d hw:1 -p128 -n2 -r96000


Onto one of my Intel quads with M-Audio PCI cards, Audiophile 2496 and Revolution 7.1.

I can get 1.3ms latency at 48000Hz/32 samples and 0.6ms at 96000Hz/32 samples with no xruns with Ubuntu Studio 16.10 with the real time kernel.

Obviously the hardware matters for latency, but so does the distro/kernel.

I'm usng the LinVst-rt version btw.