Does the desktop environment you use affect audio latency in any way? Do different environments cause more or less latency? This is what I have been trying to find out.

This is a companion article to a video experiment. Please watch the video as well as reading the article.

Audio latency is the natural born enemy of all recording musicians. It is the delay that can occur between your playing and the actual recording on your computer. Audio engineers, software developers and hardware manufacturers do everything they can to reduce audio latency as much as possible.

I use Ubuntu Studio to help with my work as a musician. It is set up to be great for audio work, from the software that comes preinstalled to the Low Latency Linux Kernel, it feels like everything on it has been tailored to give you the best experience. This got me thinking however. It uses Xfce for its desktop environment, which is light on resource usage making it a good choice for running other creative programs such as video editors or 3D renderers, but does it have any effect on audio latency? And if so, how does it compare to other desktop environments?

Set up

I chose to test the majority of popular Linux desktop environments on my computer, which is running Ubuntu Studio 17.10 with the Low Latency Linux Kernel 4.13.0-45. I used Ardour 5’s audio latency test on each desktop environment, testing them five times each for both ALSA and Jack. I logged out and back in to every session after each test to try and give the fairest results.

Unfortunately, some of the environments refused to run Ardour, these were:

• Unity
• Cinnamon
• Budgie
• LXDE

I was however able to test these desktop environments:

• Kde Plasma
• Gnome
• MATE
• LXQt
• Xfce
• i3
• OpenBox

Each desktop environment was set to its default config, I disabled the conky I had running, and I only started testing after I had installed all of the desktop environments. I did this so if there was any risk of bloat from the environments negatively affecting results, this will be on every test, thus making its effects redundant.

The Test

Other than the issue with some desktop environments refusing to run Ardour 5, the testing itself went quite smoothly. For each environment you may find screenshots of every test for both ALSA and Jack, these results will be elaborated on in the Results section of this article (as well as in the video).

Heavy Desktop Environments

Gnome

Gnome is one of the most popular desktop environments for Linux distributions, it is very modern in design, but can be a bit resource heavy. For this reason, you don’t see it offered as standard on many audio production Linux distributions.

The results for Gnome are:

Gnome ALSA 1

Gnome ALSA 2

Gnome ALSA 3

Gnome ALSA 4

Gnome ALSA 5

Gnome Jack 1

Gnome Jack 2

Gnome Jack 3

Gnome Jack 4

Gnome Jack 5

Kde Plasma 5

Plasma is the desktop environment from the Kde community, it utilises the Qt framework, and is already being used by the audio production Linux distribution KXStudio. Despite its reputation for being heavy it can be stripped down to be lighter.

The results for Kde Plasma are:

Plasma ALSA 1

Plasma ALSA 2

Plasma ALSA 3

Plasma ALSA 4

Plasma ALSA 5

Plasma Jack 1

Plasma Jack 2

Plasma Jack 3

Plasma Jack 4

Plasma Jack 5

Lightweight Desktop Environments

MATE

MATE is the continuation of the old Gnome 2 desktop environment. It is lighter than Gnome 3 and quite customisable. It is a popular choice for a lot of Linux users, but doesn’t seem to be offered as standard on many audio production distributions.

The results for MATE are:

MATE ALSA 1

MATE ALSA 2

MATE ALSA 3

MATE ALSA 4

MATE ALSA 5

MATE Jack 1

MATE Jack 2

MATE Jack 3

MATE Jack 4

MATE Jack 5

Xfce

Xfce is the default desktop environment for multiple audio focused Linux distributions including Ubuntu Studio and AV Linux. It is light weight, customisable and stable.

The results for Xfce are:

Xfce ALSA 1

Xfce ALSA 2

Xfce ALSA 3

Xfce ALSA 4

Xfce ALSA 5

Xfce Jack 1

Xfce Jack 2

Xfce Jack 3

Xfce Jack 4

Xfce Jack 5

LXQt

LXQt is the evolution of LXDE (which failed to run Ardour and so has been omitted from this experiment), it has done away with the GTK framework which LXDE used and is focusing on using the Qt framework (which Kde Plasma also uses). It is still under development and so has yet to be incorporated into audio production Linux distributions.

The results for LXQt are:

LXQt ALSA 1

LXQt ALSA 2

LXQt ALSA 3

LXQt ALSA 4

LXQt ALSA 5

LXQt Jack 1

LXQt Jack 2

LXQt Jack 3

LXQt Jack 4

LXQt Jack 5

Super Lightweight Desktop Environments

i3

i3 is a very interesting desktop environment. It is run almost entirely from the user’s keyboard and uses a tiling window manager. It is a top choice for people who want to squeeze the most performance out of their computers. However, it isn’t really offered as standard with audio production Linux distributions.

The results for i3 are:

i3 ALSA 1

i3 ALSA 2

i3 ALSA 3

i3 ALSA 4

i3 ALSA 5

i3 Jack 1

i3 Jack 2

i3 Jack 3

i3 Jack 4

i3 Jack 5

OpenBox

OpenBox is an incredibly light weight desktop environment, it lacks panels which at first might seem like a bad thing, but the environment works very well by right clicking the desktop to access all of your programs. Negating a panel means less resources are used on the desktop itself and more can be allocated to other things. OpenBox comes as one of the desktop environments offered on the audio production Linux distribution Musix.

The results for OpenBox are:

OpenBox ALSA 1

OpenBox ALSA 2

OpenBox ALSA 3

OpenBox ALSA 4

OpenBox ALSA 5

OpenBox Jack 1

OpenBox Jack 2

OpenBox Jack 3

OpenBox Jack 4

OpenBox Jack 5

Results

I was honestly surprised by the results. Going into this I was doubtful that there would even be a difference between desktop environments, and I believed that if there was a difference it would be negligible. However, there does seem to be a difference between the desktop environments. For the most part the hypothesis that heavier desktop environments would cause more latency has held true, at least for ALSA. Indeed, the results are quite different between ALSA and Jack.

The latency test covers two different kinds of latency; Detected Roundtrip Latency and Systemic Latency. I chose to take my results from the Systemic Latency test, which is split into Milliseconds and Samples. This has been done for both ALSA and Jack.

Below are the graphs for all of the results:

When it comes to ALSA i3 won overall. Seemingly proving (although more testing would be preferred) the hypothesis that a lighter desktop environment will cause less audio latency than a heavier one. i3 won when testing audio latency in both milliseconds and samples. OpenBox came a close second, followed by Xfce. Gnome, Kde Plasma, MATE and LXQt all did a lot worse in this test, with LXQt fairing the worst. This is most likely due to LXQt still being in development.

Jack however, is a rather different story. Kde Plasma came out on top, with overall improved results compared to ALSA. Kde Plasma is significantly ahead of the other major desktop environments, and even comfortably ahead of the super light weights. i3 and OpenBox follow, but with OpenBox leading ahead of i3 this time around. LXQt did better, MATE and Gnome came last, and Xfce slid down the ranks from third best, to third worst.

Conclusion

These results are very interesting. They seem to prove for the most part (once again more testing would be needed to completely confirm this) the hypothesis that lighter weight desktop environments cause less audio latency to occur than heavier desktop environments.

Even with the evidence of desktop environments such as i3 and OpenBox doing so well, I still believe distributions such as Ubuntu Studio choosing to use Xfce as their default desktop environment is a good choice. Although both i3 and OpenBox are great desktop environments that offer lightning fast speed and minimal resource usage for the user, they aren’t what one would consider to be a ‘standard’ desktop environment. By this I mean Xfce follows a desktop paradigm that most users are familiar with. On the other hand, i3 and OpenBox follow more unique paradigms which can be confusing for the average user, or to a recording engineer who isn’t interested in learning a new paradigm, and just wants to get on with recording, mixing and mastering.

There are rumours that Ubuntu Studio are considering implementing a second default desktop environment to go alongside their Xfce based distribution. This potential desktop environment is Kde Plasma. When I first heard this, I was sceptical as Kde Plasma is seen as being quite system intensive when compared to other desktop environments. However, with the results of the Jack portion of this experiment it seems as though, for audio work at least, Kde Plasma would make a good companion for Ubuntu Studio as a second default desktop environment. Just on a side note Kde Plasma would suit Ubuntu Studio for a lot of non-audio work cases. For example, a reasonable amount of the software bundled with Ubuntu Studio are either Qt based, or a full part of Kde itself, for example the video editor Kdenlive.

In conclusion, I would recommend that anyone looking to squeeze as much performance for their audio production work as possible, out of their computer, should consider either i3 or OpenBox if they mostly work using ALSA (although their performance under Jack was also very good), or Kde Plasma if they plan on working mostly using Jack.

For now, I plan on continuing using Xfce as my desktop environment and doing more research into this topic. If you have any words of advice for me, or comments and criticisms about my experiment please comment, either on this article or on the video. If you would like to attempt this experiment yourself, a similar experiment, or you find a way of running the experiment on the desktop environments I couldn’t, then please publish your work and link it to me, I am fascinated to see what other people find. With any luck we will be able to come to a definitive answer, as to whether desktop environments affect audio latency, and which is the best overall desktop environment to use for audio production.

Sources

Here are links to articles I read in preparation for this experiment. Some are quite old, but they came in handy.

https://flexion.org/posts/2014-03-memory-consumption-of-linux-desktop-environments/

https://www.phoronix.com/scan.php?page=article&item=linux_desktop_vitals&num=1

http://libremusicproduction.com/articles/advantages-choosing-audio-orientated-linux-distribution

https://pmdelgado.wordpress.com/2018/03/25/implementing-low-latency-audio-processing-on-low-power-linux-devices-with-pure-data/