Evolution Distros over time
ProtonDB is mostly used to track games, but there is another way to use it: to track distros used by gamers over time. Turns out we have now a great dataset since 2018 of what distributions were used to make reports of ProtonDB, and we can exploit that observe trends. The usual caveats apply (for the methodology, see at the end of this article for more details): this may not be representative of the Linux Gaming market at large, there are variations month after month so we won’t care about a few percents ups and down, and so on. Enough said, you already know all that. Still, I would argue that people who contribute to ProtonDB are avid and active Linux Gamers, and probably at the forefront of larger trends. You can expect ProtonDB users to feature more tinkerers as well, so seeing Arch over-represented is not surprising. But the point is that the sample is probably consistent over time, and we are interested in how the choice of ProtonDB users is evolving since 2018.
And my oh my, the tide has turned in 2021: (click to expand the picture, you will need it)
Shocking! Ubuntu has collapsed.
Ubuntu used to be a juggernaut, a staple in the world of Linux Gaming. Ubuntu was the default go-to distro for many users for probably a decade. It’s still an important distro at almost 20% share in our graph, but it’s not the first one anymore. It has not had any kind of mindshare for a while now, with the popularity of rolling release distros and new entrants like Pop!_OS doing a better job at being Ubuntu than Ubuntu itself. It may not be a coincidence that suddenly we see Canonical looking to hire folks to help around the gaming aspects of the distro. Too little too late, probably.
It’s an Arch, Arch world
Not only has Arch Linux gained progressively some share to be solidly at 20% now for several months, but the whole landscape looks a lot like Arch derivatives now. Manjaro is almost as big as Arch (but seems to be stagnating in adoption below 20%) and Garuda Linux and EndeavourOS add 6-7% to the mix. In effect, almost half of the gamers on ProtonDB are running some form of Arch.
The New kids on the block
What’s the most surprising is that it’s still possible to create new distros and get major follow-up, without much marketing apart from the word of mouth. Pop!_OS is one of the best examples of that. It was almost nothing 3 years ago, and it has been growing steadily to represent something like 12.5% of the Linux gaming market now. EndeavourOS followed suit not too long after, going from minuscule share levels to almost 5% now. Garuda Linux has emerged much more recently but seems to follow the same pattern, gaining a few percents over the last year.
We have talked about Ubuntu already, no need to rub salt on the wound. First, Linux Mint, which is a distro I usually recommend without hesitation to beginners, has been falling out of fashion for the past 3 years. It’s not going to disappear anytime soon, but it looks as if people who liked Ubuntu and Mint have progressively moved to Pop!_OS, while the truth is probably a lot more complex. Pop!_OS has been getting a lot of good reviews, System76 is courting MacOS users with it, and at the same time Linux Mint has been pretty quiet. Note that Mint has never been a project that was marketing itself too much: the kind of silent submarine, well-funded by its community and steady in its progress and releases. Pop!_OS has made itself a lot more noticeable, first by being part of the System76 offering, but also by explaining what they did differently better, and staying in the headlines often enough to be noticed.
And then there’s Debian, the odd one of the bunch. Who uses Debian for gaming? There’s HexDSL, that’s true. It’s an odd choice because Debian is known for its focus on stability leading to outdated packages. Recent packages are typically required for anything gaming-related - if you want to use an AMD GPU, you’d better have a recent Mesa driver version and recent kernel to play along. Of course, you could always switch to Debian testing or Debian unstable to be closer to the edge, but then if you go there, there’s very little point in using Debian vs anything else: in the DEB space, Pop!_OS or even Linux Mint would probably be the better choice in user-friendliness.
Fedora is not very big when it comes to ProtonDB reports. I’m fairly confident that Fedora is much bigger than that out there on the desktop, but maybe when it comes to gaming it’s not as big… or else, ProtonDB contributors may be from places where Arch or Ubuntu were more often the choice than not. Fedora’s user base is thought to be bigger in the US than in Europe, so that may be a factor.
How consistent is it with our previous Linux Gamers Survey?
Since the Linux Gamers Survey we last conducted last year in April 2021, we can check how well it matches the ProtonDB sample.
The order is almost the same, but Ubuntu is ever lower than in ProtonDB. Arch is much bigger than Manjaro in the Linux Gamers survey as well, so there are certainly samples biases in one way or another. Fedora is at a more respectable place in the survey, which feels more appropriate. Debian and EndeavourOS are about at the same place as where they were in April 2021 in the ProtonDB chart.
The limitation of this survey is that it’s a one-time thing, with some deviation based on whoever saw the survey and replied. The ProtonDB analysis is an indirect method, tracking an actual reporting activity where hardware/software information is directly extracted from the Steam client info. Purely in terms of information reliability, I would say that the ProtonDB measurement is more accurate to detect to detect variations of distro choice over time as well, while the bias of whoever makes reports on ProtonDB exists.
What to make of this?
Well, first I did not expect to see much changes within 2021 when I prepared this graph. Seems like Valve’s move to use Arch for the Steam Deck may have come at a time when Arch and its derivatives are enjoying peak popularity as well. I was also quite surprised to see new distros making such a dent: the distro market is more alive than a few years back, or at least it feels this way: new derivatives, specific purpose distros like ChimeraOS (ex-GamerOS), and avant-garde distro designs such as NixOS and GNU Guix. Interesting times, and 2022 may see this landscape change at an even faster pace.
Methodology and caveats
In case you were wondering how to extract individual distro choice information from ProtonDB, since there is no individual user identifier, we proceed this way:
- Extract OS version and hardware information (CPU and GPU) from each report
- Make a unique footprint based on such extracted info
- We assume that reports who share the exact same footprint in the same month are from the same person, so if that same footprint made 10 reports in the month, we only count it as being from one configuration.
- This way we are able to fairly reliably identify how many actual users produce reports on ProtonDB and what distribution they use.
Note that there is of course the remote possibility that two different users share the exact same configuration in the exact same month, but in all likelihood this should be an exception, not the rule, when considering PC gamers. On the other hand, there is also the distinct possibility that someone changes from one distro to another in the same month, or changed their hardware while staying on the same distro, in which case they would be counted as different users and counted twice. Against, the number of people for whom such changes would occur should be a small percentage of the total reporters.