2017: Should Linux Benchmarking Still Be Mostly Done With Ubuntu?

[spstarr]

Considering most of the Linux graphics stack is developed by Red Hat/Fedora people, it makes more sense to use Fedora if you want more official. I'm bias though

[waxhead]

First of all I am a huge Debian fan so I am biased a bit. However one should keep in mind that Debian and RedHat today are essentially the mother distros for just about anything that have a penguin calling the shots.

https://upload.wikimedia.org/wikipedia/commons/1/1b/Linux_Distribution_Timeline.svg

Unless a distro is advertising that it does add a significant amount of optimizations on top of for example Debian I don't see the need to test anything but the base system. Now this is awfully close to saying that everybody should use a "operating system" such as Windows since everybody else is using it, but this is not what I am trying to get across.

I think that benchmarks should be done in Debian and RedHat first and if more fine grained benchmark should be needed you can walk the tree and pick the most relevant distros such as Arch, Ubuntu, openSUSE, CentOS etc.. depending on the use case and what they offer (performance wise) on top of the mother distribution.

[mibo]

IMHO *every* distro should be used for benchmarks.
Of course this is not possible, due to time constraints - but that should be the direction to go. (maybe having 2 distros that are used over a month and the next month two different ones. Or one fixed distro (SteamOS?) and a second one per month...)

The point for me is that Michael finds regressions with his testing. So everybody is happy if Michael stumbles across some problems - that would benefit Linux as a gaming OS a lot. Also, it would be visible, how major differences (MIR, desktops, ..) affect the gaming experience - or not.

[leipero]

From performance perspective, it makes little difference, using git versions of mesa and otehr software makes distribution choice even more irrelevant from that standpoint. However, if you already compile kernel, mesa etc., on Ubuntu, it might be easier and faster to test ot on Arch. Enabling test repository would grant you latest stable software, so from that perspective would be easier, multilibs are managed better than on Ubuntu based distributions, and software is not patched as people in this thread already mentioned.

At the end, best option is to use what is the easiest, if that happens to be Ubuntu, it's fine, for me it would be easier to use Arch.

[137ben]

In the limited sample size of people I work with:
--My office's computers are all on Ubuntu GNOME
--I personally use Mint on my personal computers for my Linux needs.
--My colleagues overwhelmingly use either Ubuntu MATE, Kubuntu, Ubuntu GNOME, Mint, or Lubuntu on their personal computers, with a few oddball MacOS, Debian, and Fedora users.

So, for non-graphics related benchmarking, Ubuntu is probably a pretty good indicator of performance for the (admittedly small) group of people I know and work with who use Linux. However, for any benchmark for which Unity/Mir would differ from Wayland or X11, Ubuntu with Unity probably won't provide me with any information I could use. I don't think I actually know anyone in person who uses the Unity version of Ubuntu, but I acknowledge that the group of people I know may not represent the overall Linux community (or the Phoronix reader base).

I should also note that I don't generally play PC games, so the most graphics-heavy thing I do on Linux is usually strreaming video. So, I'm not necessarily a part of the target audience for this website.

[ruthan]

As Windows user i can write, dont compile it, dont use VI for it, just create a poll bro

Otherwise Michael should be definitely Distrowatch bitch! => Mint.

To be honest, i dont believe in present desktop distributions Linux boom at all, so for end user would be maybe more interesting RemixOS (based on Android-x86) + we need to wait for some next big Linux desktop thing.

[darkbasic]

Arch Linux

[bug77]

My 2 cents: since you love benchmarking using defaults so much, try a 50-50 split between Ubuntu (a conservative distro) and Fedora (a more up to date distro).
Because the minute you start: I tested this on Ubuntu with a kernel from this branch and with Mesa from git with a patch for Y, you've lost any casual user.

NB You're doing a great job as is, I can follow it no problem (but I'm the weirdo who still opens up a console). And when it comes to improvements the only thing that springs to mind is making it more accessible to less experienced users.

[Zucca]

I don't use Arch, I use openSUSE, but I agree that using Arch would give a more realistic view of what the performance of Linux is. Not only because of patches, but also because of its rolling nature.

Yeah. The rollingness gives the most up-tp-date results.

[patstew]

I would also +1 Arch, because of its lack of distro patches and up-to-date-ness, but then I use Arch so I'm biased.
I think a good reason to dump Ubuntu is that they're diverging from the rest of desktop linux with Unity+Mir as their primary focus. Even if ubuntu is the biggest it doesn't outweigh the others when the others are closer to each other.