What’s the best way to tease better audio performance out of Ardour? (lower DSP, fewer xruns)?
Here’s the recap:
I just started making music in Linux (been a long-time Cakewalk user). I normally make music at 192kHz/24-bit, and I only ran into audio problems in Cakewalk/Windows when I would get silly with high-CPU plugins/instruments/whatevers. So I know the computer can handle it. (On a side note, I also recently upgraded to a beefier computer.)
I’m currently running Ardour on Linux Mint with the low latency kernel, using Jack. I start Jack first, apart from the DAW, set it to 96kHz, and set the buffer/frames to 2048/2 (the latency is a reported 42.7ms). Then, for comparison, I created two similar projects, one in Ardour, and one in Reaper - same synths, same effects, playing (generally) the same thing. What’s weird is that I don’t get dropouts in Reaper, but I still get significant dropouts in Ardour. So I know my hardware is powerful enough, and Jack is working, but I’m wondering if anyone’s had a similar experience, or what tips/tricks people have to squeeze every last drop out of Ardour. (I prefer Ardour’s workflow. And looks. And flexibility…)
My testing looked like this:
-Open DAW #1. Play something. Close DAW #1.
-Open DAW #2. Play the similar project. Close DAW #2.
-Start Jack again, this time, load the DAWs in reverse order.
-Find that, both times, Ardour had dropouts when “other DAW” did not.
-Tried restarting Jack between opening the DAWs. Same result.
-Using ALSA doesn’t work as well for me.
-Using the open source Nvidia driver isn’t an option for me, because I need the proprietary driver for my video editing.
tl;dr - how come Ardour drops out, and other apps don’t? wut?
Depending on interface you are using,I would try to use alsa (without jack) and 3 periods instead of 2. On an apogee firewire ensemble it makes a big difference.Moving from windows to linux can present a few challenges and it takes some experimentation to get it dailed in.Also, reaper performance will be tough for any DAW to match.
Yeah, my ALSA experiments left me with the conclusion above. The long version is that it tended to have slightly better performance, but was less stable overall. Jack was more consistent more of the time. And I also played with the periods before, but two vs three periods made no difference in terms of the end result.
Using the Nouveau driver definitely improved performance, but I need me some CUDA for Davinci Resolve… heh.
And I did do some searching and got the system running as close to realtime as it can. I mentioned above that I’m using the low latency kernel, but I also make sure to change the system limits for access to the audio group, made sure my user is in the audio group, tried changing the app/group priority, and probably some other stuff I can’t remember. Oh, and I’m using the Steinberg UR22mk2.