Page 1 of 2 12 LastLast
Results 1 to 10 of 12

Thread: Intel Performance With Ubuntu Linux vs. Windows 8.1 Is A Mixed Bag

  1. #1
    Join Date
    Jan 2007
    Posts
    13,420

    Default Intel Performance With Ubuntu Linux vs. Windows 8.1 Is A Mixed Bag

    Phoronix: Intel Performance With Ubuntu Linux vs. Windows 8.1 Is A Mixed Bag

    Earlier this week I published an extensive set of results from thirteen discrete AMD Radeon and NVIDIA GeForce graphics cards benchmarking various OpenGL games and comparing the Ubuntu Linux and Windows 8.1 performance when using the official AMD and NVIDIA drivers for each operating system. Those results were very interesting for both the AMD and NVIDIA GPUs, but now it's time to see how the Intel graphics are performing under Ubuntu and Windows 8.1 Pro x64. Making things even more interesting here is that Intel has only an open-source Linux driver and no closed-source solution.

    http://www.phoronix.com/vr.php?view=19280

  2. #2
    Join Date
    Jun 2012
    Posts
    22

    Default

    Michael, would it make sense to do the same tests on Wine as well? I wonder if some of the Linux results could come closer to Windows, due to running actually the same code. After all, Linux binaries and Windows binaries might be compiled completely differently.

  3. #3
    Join Date
    Jul 2013
    Posts
    146

    Default

    Hmm, I really wish Michael had waited a couple of weeks to do this comparison. There are a whole bunch of shader optimisations that have been posted to the mailing list this week. A number of i965 specific one's but also general glsl optimisations which should help all drivers. This one seem like it would have given much improved results on Unigine Tropics: http://lists.freedesktop.org/archive...er/047194.html
    "for tropics, that accounts for most of the effect, the FPS improvement is 11.67% +/- 0.72%"

    It would be good to see these tests redone with these patches applied.

  4. #4
    Join Date
    Nov 2011
    Posts
    257

    Default

    Graphics driver guys should push DE guys more. In case of a performance regression it's usually the DE bunch who screwed up
    https://bugzilla.gnome.org/show_bug.cgi?id=711028

  5. #5
    Join Date
    Jan 2007
    Location
    Germany
    Posts
    2,091

    Default

    Quote Originally Posted by tarceri View Post
    Hmm, I really wish Michael had waited a couple of weeks to do this comparison. There are a whole bunch of shader optimisations that have been posted to the mailing list this week. A number of i965 specific one's but also general glsl optimisations which should help all drivers. This one seem like it would have given much improved results on Unigine Tropics: http://lists.freedesktop.org/archive...er/047194.html
    "for tropics, that accounts for most of the effect, the FPS improvement is 11.67% +/- 0.72%"

    It would be good to see these tests redone with these patches applied.
    You can always wait for the next optimizations.... and the next... and the next...

    Or you do it like Michael and benchmark everything every month.

  6. #6
    Join Date
    Apr 2009
    Posts
    491

    Default

    Quote Originally Posted by tarceri View Post
    Hmm, I really wish Michael had waited a couple of weeks to do this comparison. There are a whole bunch of shader optimisations that have been posted to the mailing list this week. A number of i965 specific one's but also general glsl optimisations which should help all drivers. This one seem like it would have given much improved results on Unigine Tropics: http://lists.freedesktop.org/archive...er/047194.html
    "for tropics, that accounts for most of the effect, the FPS improvement is 11.67% +/- 0.72%"
    Nice to see singnificant improvements still coming down the line. Thanks for the info, tarceri! It's also nice to see, in the article, that the very latest stack is already showing some significant improvements in some areas. This means that end users will see a nice performance gain in early 2014 as they upgrade their distros.

  7. #7
    Join Date
    Dec 2010
    Location
    MA, USA
    Posts
    1,102

    Default

    Quote Originally Posted by RushPL View Post
    Michael, would it make sense to do the same tests on Wine as well? I wonder if some of the Linux results could come closer to Windows, due to running actually the same code. After all, Linux binaries and Windows binaries might be compiled completely differently.
    No, it wouldn't make sense. Due to how unstable it is (and the regressions between versions), wine is completely useless for a hardware benchmark, especially when you're comparing open source drivers against closed source. The only meaningful benchmarks with wine would be to compare between different releases of wine, using the proprietary nvidia drivers. I suppose comparing wine against Windows would be useful too, but only for platinum-rated programs, and again, with proprietary nvidia drivers.

  8. #8
    Join Date
    Jul 2013
    Posts
    146

    Default

    Quote Originally Posted by d2kx View Post
    You can always wait for the next optimizations.... and the next... and the next...
    I would normally agree with you but there has been many more Optimisation patches than normal posted to the list at one time . To add to the previous link there is also this 15 patch series: http://lists.freedesktop.org/archive...er/047185.html

    Quote Originally Posted by d2kx View Post
    Or you do it like Michael and benchmark everything every month.
    Yes but Windows vs Linux benchmarks are done a lot less often. As are benchmarks that include as many different benchmarks as this one did, for example lately the Unigine benchmarks haven't been getting run all that often. All I'm saying is it would be nice to see this all redone when the patches are merged, when Mesa 10 is released in a month or now with the patches applied to master.

  9. #9
    Join Date
    Feb 2013
    Posts
    77

    Default

    I have to disagree with the interpretation of the Xonotic "Ultra" and "Ultimate" results. Sure, Windows gets higher peak and average fps, but look at the minimum fps numbers: In "Ultimate", the Windows result at 14fps is closer to the out-of-the-box result of 12fps than it is to the Mesa10 result of 17! Of course none are really playable numbers, but I'd prefer to play a game that drops down to 17fps than one that falls to 12fps no matter how smooth it is when you stare at the floor! ;-)

  10. #10
    Join Date
    Feb 2013
    Posts
    77

    Default

    Quote Originally Posted by phoronix View Post
    Phoronix: Intel Performance With Ubuntu Linux vs. Windows 8.1 Is A Mixed Bag

    Earlier this week I published an extensive set of results from thirteen discrete AMD Radeon and NVIDIA GeForce graphics cards benchmarking various OpenGL games and comparing the Ubuntu Linux and Windows 8.1 performance when using the official AMD and NVIDIA drivers for each operating system. Those results were very interesting for both the AMD and NVIDIA GPUs, but now it's time to see how the Intel graphics are performing under Ubuntu and Windows 8.1 Pro x64. Making things even more interesting here is that Intel has only an open-source Linux driver and no closed-source solution.

    http://www.phoronix.com/vr.php?view=19280
    Michael, is there a openbenchmarking id for this? I'd like to compare with my Ivy Bridge system with the optimizations from mesa-dev.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •