Friday, September 17, 2010

Best Linux Distro for 3D Performance

There is one question all new Linux users ask themselves at one point or another:

Which Linux distro do I want to use?

The answer to the question is different for everyone and varies depending on what you are doing with your Linux box. One thing I think many users fail to consider (or perhaps don't care about) is the level of 3D performance their distro gets. You may think that your choice of distribution does not matter in this area, but you will soon see it very much does.

The Hardware: While my hardware is not fastest in the world it is decently quick. Processor - Intel p9700 2.8ghz Dual Core, RAM - 4gigs of DDR3 1066, Video Card: nVidia 260m with 1gig DDR3 dedicated memory (Running the latest stable nVidia driver 256.53).

The Software: I used my favorite OpenGL benchmarks made by Unigine Company. The distros I am going to compare are:
All of the tests where run on clean, fully updated install of each distribution. They use the stock kernel each of the distributions provides. Desktop effects where turned off in all cases.

The Results:
Well, being a math guy, I firmly believe the numbers speak for themselves. So here are the results of the three benchmarks (higher is better - click on image to enlarge):

As you can see the scores of most of the Linux distros are fairly close (within 3% of the number one), with one exception: Ubuntu. It appears that of all the wonderful improvements Canonical has been making to Ubuntu, 3D performance is not one of them (10.10 scored higher than 10.04, but only by a small amount). It is hard to get the exact numbers by reading a histogram so here are the scores in numerical form:

Across all three tests Chakra scored the highest (With PCLinuxOS and Sabayon in close second and third). Ubuntu 10.04 was at the very bottom (over 10% behind Chakra). While I think Ubuntu is a great distro it appears that if you are a Linux Gamer, you are better off using a non-Ubuntu distro.

~Jeff Hoogland
Please note while these benchmark scores presented are accurate to the best of my abilities, they only represent my personal hardware and software configurations. Your results on your own system(s) may vary (and if they do, please share them!).


  1. Using chakra and i'm happy it comes out the highest. Didn't expected that ubuntu was that low.

  2. I would think the distro you choose is not relevant to the raw performance you get from specific drivers and configurations. Also, your bar graphs are pretty misleading; they simply start off at 2700 without a breaking symbol and makes it look like Ubuntu performs half as well as the other distros.

  3. @Sumant Those are the default graphs OpenOffice Calc made for the data I gave it. Sorry if you feel they are miss leading, but the y axis is clearly labeled and I state that Ubuntu is only 10%ish lower than the others.

  4. How about you compare Linux 3D to Windows? No offense but comparing a bunch of Linux Distros 3D performance is like seeing which kid in the Special Olympics is the fastest.

  5. @HouseOfBugs
    Because we don't care about Windows.

  6. The reason is that Ubuntu has compiz (3d desktop effects) enabled by default.

    If you disable them and re-ran the test they would be similar to the others.

    I can safely say that most distros are roughly the same speed having just benchmarked various games (via phoronix-test-suite) on ubuntu,opensuse,arch and gentoo yesterday

    Re : Houseofbugs - if you see any 'workstation' (i.e power user) benchmarks Linux generally beats Windows..

    i.e : -

  7. @Yossarian UK "All of the tests where run on clean, fully updated install of each distribution. They use the stock kernel each of the distributions provides. Desktop effects where turned off in all cases."

  8. @houseofbugs i think he just made one special for you ;-)

  9. Just a question - rather than comparing benchmarks, are the real games that you yourself play still *playable* on Ubuntu, given that it is 10 percent slower than the fastest distro compared?

    i.e. if you're playing a game on Chakra that runs at 66 frames per second, does it really matter that it is running at 60 frames per second on Ubuntu, if Chakra is much harder to use otherwise?

    (I don't know if Chakra is harder to use than Ubuntu, I was just using this as an example.)

    My point being there is sometimes more to it than raw speed, such as ease of configuration, availability of drivers, support etc.

  10. I belong to the category that you mentioned parenthetically at the beginning of your article: I really don't care much about the 3-D performance of my Linux distro.

    Nice article, nevertheless.

  11. Interesting that Mint which is a derivative of Ubuntu (last I heard) was so much faster (relatively) than it's parent.

    Also, worth mentioning is that different distros have multiple kernels available which might affect the numbers - for example openSuse provides 3 of them: default, desktop and pae.

    Finally I would think that desktop used might have an effect on performance and that you should have kept that consistent across platforms.

    Ok, one more thing - what about a test using one (or more) distro's with diff de's (Gnome, Xfce, KDE, Lxde)

  12. @Annymous Note the Linux Mint version I used was built on Debian, not Ubuntu. As for the DE, that is on my list of benchmarks to run, expect it within the week.

  13. Ubuntu Lucid uses older Mesa than all other distros it has been benchmarked against here but I expected Maverick with newer MESA to do better so thats a surprise, tests done at Phoronix show Lucid to be at par or slightly slower than most other distros in terms of 3D performance.

  14. What exactly are you benchmarking? You give numbers, but without a (y-axis) label, so how do we know that these numbers are even relevant?

    I don't mean to be rude, but I am a bit picky when it comes to making sure everything is in order.

  15. @malachi1990 The Y axis is the score the distros recieve from the benchmark I linked to. Fairly standard for benchmarking.

  16. PCLinuxOS FTW ;--P

    Well, I'm pleasantly impressed but not entirely surprised that PCLOS is right up there.
    Video and 3D run really well on my E8400 box.

  17. Nice Article, but I would like to know if you can really feel the difference (lets say, between Ubuntu and Chakra) while gaming. Those 300 points mean 1 fps of difference or 50 fps?


  18. Right about 10%, meaning for every 10fps you see under Chakra you would only see 9 under Ubuntu

  19. If that's right, I don't think you can really feel the difference while gaming.

    But normally different games throw different results, so I don't think we would get that 10% difference all the time.

    Still, this is very interesting for people who work with 3D rendering and all that, but I don't think a gamer would feel a lot of difference.


  20. Huh... Nice to see some benchmarks actually comparing different linux distro's instead of just 'this one is faster because...'

    ...because, well, it's part of their philosophy, 'mmkay?

    Anyways, I'd love to see something about how Slackware compares (picking the faster desktop environments that come with it and such, as Slackware doesn't 'default' as much as other linux distros).

    Maybe I'll rerun some of the tests you made to see if hardware makes a difference when comparing various distro's 3D performance, since I went with AMD and ATI (2.8 x6 processor, IGP Radeon 4290) instead of Intel and Nvidia.

    You actually think... What?

  21. I have been meaning to try out slackware... Maybe at some point in the next couple weeks I can give it a try :)

  22. I quote:
    "Canonical has been making to Ubuntu, 3D performance is not one of them (10.10 scored higher than 10.04, but only by a small amount)"

    I don't have much experience with Linux's guts but i don't see how performance can be improved by Canonical. X as far as i know is not developed by them, maybe they just contribute, plus the video drivers also not made by them. How can they work on this ? The only thing they can probably do is optimize compilation or remove unused stuff.
    In my experience proprietary drivers from Nvidia for linux are the best out there. Comparing with ATI which are a piece of crap I can say these guys are doing a good job.

  23. Im come Ubuntu 10.10 smashes fedora 13 for me, is it because i have an I7 processor and kernel 2.6.35 works better with it? fedora 14 has that kernel but i cant get broadcom and nvidia to work right on it

  24. ummmm...didn't see it anywhere. Were these 32 or 64 bit distros?

  25. did you use the same kernel version on all of them? on my laptop, i get poor performance with kernel 2.6.35 with ubuntu, about 45 fps average in tremulous, but by using any other kernel, specifically kernel, i get 60 fps average in tremulous and really noticeable performance increase in compiz and every single opengl app and games. not the 2d games though.

  26. please! for next benchkmark 64 bits :)

  27. I would very much like to do away with windows and try something new - I'm a hardcore gamer, built a rig with:
    CPU Q9650 @~3.00 GHz
    6 GB DDR2 RAM
    NVidia 260GTX /w 1GB DDR3 VRAM
    --The problem is that 1] don't know jack about Linux and 2] Most high end games I own are Windows platform only - Could someone point me in the right direction?

  28. @Stephen

    Many Windows games do run under Linux without too many issues. Just know that most all Windows games do run at a lower frame rate under Wine/Linux.


  29. Chakra is based on Arch Linux, and of course, Arch Linux really is the BEST Linux Distro.
    The biggest and most obvious hurdle is the propretary Microsoft Directx implementations.
    Unfortunately still, Side-by-Side, Windows games just look and feel better than the same "Wine" game on Linux.
    Until Gaming companies actually start utilizing OpenCL,..., MS directx games,..., will always be slightly better. :(

  30. Just disable desktop effects and don't use Unity. Unity on tablets I understand, Unity on desktop I don't... but I don't understand the point of GNOME3 either.

  31. So you skew the appearance of the results by not starting the graphs off at zero, and you call yourself a 'maths guy'? Ubuntu doesn't do itself any favours (Unity, Gnome 3, etc) and is consistently behind, but it's not 50% behind like the graphs suggest, it's around 10%.

  32. Well the numbers alone dont speak for themselves. You should at least account for the variance of the scores. Otherwise it's hard to tell if you just happened to draw a bad sample...

    Also the scores of the individual benchmarks appear to be on different scales, so f.e. the sanctuary benchmark has a higher influence on the total score than the other two. Assuming all benchmarks are all equally important you should rather calculate the total score as the sum of the normalized scores.

  33. It would be really interesting to repeat this test because of all the talks about steam and opengl.