Good Job…
May be helpful or not . Latency Reduction at the Core: BIOS Tuning Guide
Arturia_Software_Center__2_10_0_2970
Error code …
Older one used to install and run great with WINE but, now auto update takes over and crashes it.
PS.
Arturia, Please consider changing the auto update option.
Do you want to Update? YES or NO
Same thing with Windows apparently
Hello,
You can re-install the ASC from the newest version.
No problem for me with Xunbuntu 24.10.
Hi ! Back a few months later… 
So the situation has improved since Paul Davis’ Linux kernel patch ([PATCH AUTOSEL 6.14 14/20] ALSA: usb-audio: Fix duplicated name in MIDI substream names - Linux-stable-mirror - lists.linaro.org). The KeyLab (MK3) is automatically detected by the Arturia instruments and its built-in display works 100%.
Also, I put back an old USB audio interface on my PC and noticed I got a lower latency as I can select a rate higher than 48 KHz. For some reason I manage to keep the same number of buffers with no xruns at 96 KHz, so my latency is halved (and sound quality doubled).
ATM with no particular tweaks, I can use 64 buffers @ 96 KHz with some instruments, but some require 128 buffers not to get artifacts.
I still have to check if I get better performance with jack (instead of pipewire).
1 Like
Hi ! I wanted to see if I could slightly improve my configuration.
I figured out using the “Pro Audio” profile for my soundcard (Pipewire) significantly reduced the round-trip latency (at the cost of requiring more buffers).
So I settled with the following settings : Quantum 512, Rate 96 KHz, latency 5.33
→ real round-trip latency : 12.5 milliseconds.
With those settings, I cannot feel the latency and have all Arturia instruments, Bitwig and Pianoteq working perfectly in all conditions.
Before, I used to have 25 milliseconds of round trip latency when using 256 buffers but not using the “Pro audio” profile. (and the DSP load was slightly higher)
I can get to 7.5 milliseconds roundtrip with a quantum of 256 but have a few micro-cuts when using heavy instruments / heavy Bitwig projects.
I think 12.5 milliseconds of round-trip latency is pretty good for both playing & production on a 10+ y-o PC.
More details on my experiments : Cable, an app to change some of Pipewire and Wireplumber settings / Community Contributions / Arch Linux Forums
What’s the specs of your pc? 96kHz is maybe not necessary? Or is it to get half the latency you would have at 48k?
This is an i5-8600K (2017). Okay, only 8 years old actually.
Actually, in my experiments, 96 KHz actually helped me lower the latency without adding xruns / micro-cuts. Actually, in “non pro audio” mode, it enabled me the halve the latency. In pro mode, actually that’s equivalent it seems – but at the benefit of a higher sound resolution.
Edit : hmm. It seems I can slightly improve the situation indeed in “Pro mode” & 48 KHz.
I get no micro-cuts in Pianoteq with 192 buffers. Issue is… I cannot set wineasio to 192 buffers. That’s either 128 or 256.
(256 = equiv latency of 512 in 96 KHz)
So that setting might be slightly better for Pianoteq or Bitwig - but not for running directly my windows instruments without the DAW.
Edit 2 :
No sound cuts & low Bitwig DSP usage
48 KHz, 192 buffers, 4 ms
Measured round trip : 10.5 ms
BUT if I want to be able to use my Arturia instruments directly (not through DAW) I have to use a power of 2 for WINEASIO.
Ain’t really worth losing this possibility for 2 milliseconds !! (also, the DSP load under Bitwig gets slightly “peakier”.)
[Preferred buffersize]
Defaults to 1024, and is one of the sizes returned by GetBufferSize(), see the
ASIO documentation for details. Must be a power of 2.
Really, after making tons of experiments, I really think I figured out the sweet spot for this given hardware configuration Works with everything and plays pretty well.
I’m glad you found a setup that works for you. If you’re happy with it, stick with it. In general, though, I wouldn’t recommend increasing your sample rate only to improve latency, if you don’t have some other reason to work at 96kHz.
When it comes to latency and CPU load, doubling the sample rate while keeping the buffer size the same will have the same effect as halving the buffer size while keeping the sample rate the same. If you double your sample rate and then have to double your buffer size to avoid audio glitches, then you haven’t improved your latency at all.
The biggest difference is what happens with the recorded audio. Changing your buffer size has no effect on the audio recorded (assuming that your CPU isn’t overloaded). On the other hand, changing the sample rate also changes the size of the recorded audio. For a given duration of recording, doubling the sample rate will also approximately double the amount of disk space the recording needs. It also doubles the amount of RAM required to load the recording into memory. Many DAWs don’t load entire recordings into memory, in which case it will halve the amount of audio that can be loaded into a fixed amount of RAM, and double the rate that it needs to read more from the disk, which can be another source of glitches when playing back large projects.
The other difference is that doubling the sample rate allows you to record higher frequencies. However, 48kHz already allows you to record the entire range of human hearing and then some. Unless you have some specific reason for needing higher frequency content (such as if you’re going to then slow the audio down a lot until it’s audible) then it’s probably not going to be worth the extra size.
This is just my lay person opinion, but this seems to be commonly misunderstood, so I thought I’d write it up in case it’s helpful to someone.
3 Likes
Yep, this is indeed what I initially thought. But for some unknown reasons, in regular stereo mode (not in pro mode, where it doesn’t apply), raising my audio rate allowed me to keep the same number of buffers without audio crackling – and hence divide my latency by 2 ! (while I would get crackling when dividing the number of buffers by 2 and keeping the same rate)
I have no idea why, but other people seem to have witnessed the same phenomenon.
This is not the case when using the “pro audio mode” – but using it has a huge benefit on latency so it’s better overall ! (in that case, I kinda wonder if I shouldn’t indeed switch back to 48 KHz to save memory)
Edit : alright. Switched back to 48 KHz with half buffers in “pro mode”. Same latency but should free some resources, which is not to be neglected considering my PC specs.