Round Trip Latency. TWICE as long on Linux. Why is this?

[777funk]

I did some testing today and noticed quite a bit less live monitoring latency on Win XP vs Ubuntu. It was so noticeable that I decided to measure it.

In Linux, I generally am not happy until I reach 128 samples. I was doing some testing in XP and noticed that even 512 was barely noticeable.

When I actually measure, I can see why. 512 in XP is 27ms round trip and nearly twice that in Linux.

Any ideas what could be going wrong to cause this?

[axel_ef]

Did you make a test with ALSA?

[777funk]

No I didn't since it doesn't work well with Firewire. I've had to use Jack.

[Xasman]

Originally Posted by 777funk

No I didn't since it doesn't work well with Firewire. I've had to use Jack.

Jack sits on top of ALSA in the Linux audio stack and as a result adds latency (as does Pulse Audio). If you need minimum latency then using ALSA direct will always yield better results. If your interface doesn't work well with ALSA then it may be time to consider getting a more compatible one.

[777funk]

Originally Posted by Xasman

Jack sits on top of ALSA in the Linux audio stack and as a result adds latency (as does Pulse Audio). If you need minimum latency then using ALSA direct will always yield better results. If your interface doesn't work well with ALSA then it may be time to consider getting a more compatible one.

Interesting. So Jack adds latency. I had no idea.
EDIT: According to Jack themselves, Jack does not add any latency (as in zero):
https://jackaudio.org/faq/no_extra_latency.html

I like Jack since it marks Xruns and I can monitor how well audio is running for hours or days without me sitting there listening or testing the system. I like this since it proves reliability.

Can Pipewire be used without adding latency?

[Xasman]

Originally Posted by 777funk

Interesting. So Jack adds latency. I had no idea.
EDIT: According to Jack themselves, Jack does not add any latency (as in zero):
https://jackaudio.org/faq/no_extra_latency.html

All I can say is that when I last ran a latency check between ALSA and JACK, JACK reported higher latency than ALSA in REAPER (admittedly, that was some time ago on an older machine - I don't even have JACK installed on my current computer). Perhaps the additional latency could have been due to the extra CPU cycles needed by JACK, I don't know. There's also a thread on this forum where others reported similar results: https://forum.cockos.com/showthread.php?t=209268

Originally Posted by 777funk

...Can Pipewire be used without adding latency?

I use pipewire-alsa and that doesn't seem to add any latency here.