Page 2 of 3 FirstFirst 123 LastLast
Results 11 to 20 of 23

Thread: Testing 60+ Intel/AMD/NVIDIA GPUs On Linux With Open-Source Drivers

  1. #11
    Join Date
    Nov 2012
    Location
    France
    Posts
    566

    Default

    Quote Originally Posted by dimko View Post
    With amount of work Michael has done.

    I am still gonna stick to my guns with Nvidia binary drivers and cards, since AMD clearly a bad choice when it comes for Bang for the Buck.
    interesting, non the less!

    I'd love to play games on open source GPU, but i don't think it's happening any time soon. May be when Weston will become reality... In a few years.
    You can't play a game on an open source GPU right now, obviously.

    Wayland won't change gaming performance in the slightest, except maybe in windowed mode with compositing. But on recent cards, compositing only has a small negative effect on gaming performance in windowed mode.

  2. #12
    Join Date
    Feb 2008
    Location
    Linuxland
    Posts
    5,070

    Default

    The box plot needs to have an explanation text (in the image) saying what percentages you used for the whiskers. The common ones (25% and 75%), or the 2%, 98% that you reported trying?

  3. #13

    Default

    Quote Originally Posted by curaga View Post
    The box plot needs to have an explanation text (in the image) saying what percentages you used for the whiskers. The common ones (25% and 75%), or the 2%, 98% that you reported trying?
    2% and 98% for whiskers, the box is around 25% and 75%, dots for outliers on the extremes past the whiskers. Min/Avg/Max reported to left of graph.

  4. #14
    Join Date
    Jan 2013
    Posts
    43

    Default

    Quote Originally Posted by dimko View Post
    I'd love to play games on open source GPU, but i don't think it's happening any time soon. May be when Weston will become reality... In a few years.
    I use the oss driver on my 7870 in linux. springrts runs fine, tesseract runs fine except for the sky because of one shader that causes crashes by register spilling, and that is still broken. But the performance is very good and way better then I expected.

  5. #15
    Join Date
    Feb 2008
    Posts
    902

    Default

    Michael, you should mention that radeon driver have shadows broken . So better avoid benchmarking Unigine Sanctuary, because numbers can be incorrect when you have broken render

    http://www.phoronix.com/forums/showt...131#post418131

  6. #16
    Join Date
    May 2014
    Posts
    23

    Default

    Quote Originally Posted by arabek View Post
    nouveau now supports reclocking btw: http://cgit.freedesktop.org/~darktam...7dab04e76b8742 Retest please
    Sadly, this isn't dynamic reclocking - the user needs to set the clock rates.

    I cannot express how awesome the box plot graphs are.

  7. #17
    Join Date
    Nov 2013
    Posts
    144

    Default

    Quote Originally Posted by tusharkant15 View Post
    Hi Michael,

    First of all, bravo on publishing such a comprehensive benchmarking article. I don't thing anyone, even on the windows side, does this.

    I currently have a MSI R7 260X which I bought in January. It was a bumpy ride initially but since the past few months I am really impressed with this GPU. My setup is Ubuntu 14.04 with stock Mesa and an updated kernel. I have seen a bunch of your articles in which your numbers for this card don't seem to match with mine. Here are the results of my furmark running at 1920x1080

    https://dl.dropboxusercontent.com/u/...%3A42%3A33.png

    As you can see, the results are quiet different. You data shows a total points of 114 while mine is over 500. A difference of a few points is understandable but >4x difference suggests there's something wrong with the test. I have also noticed under separate articles where you have mentioned that this card gets around 10 FPS in Xonotic under high settings while I always seem to get above 40FPS. Ultra crashes though.
    IIRC the 260x needs an updated firmware, Michael probably wasn't using it.

  8. #18
    Join Date
    Feb 2008
    Posts
    902

    Default

    Quote Originally Posted by Spittie View Post
    IIRC the 260x needs an updated firmware, Michael probably wasn't using it.
    Those firmwares landed yesterday in linux-firmware and here in Debian Sid as package too .

    http://git.kernel.org/cgit/linux/ker...x-firmware.git
    http://packages.qa.debian.org/f/firm...3T224907Z.html
    Last edited by dungeon; 06-04-2014 at 02:16 PM.

  9. #19
    Join Date
    Feb 2011
    Location
    Ukraine
    Posts
    132

    Default

    Quote Originally Posted by tusharkant15 View Post
    https://dl.dropboxusercontent.com/u/...%3A42%3A33.png

    As you can see, the results are quiet different. You data shows a total points of 114 while mine is over 500. A difference of a few points is understandable but >4x difference suggests there's something wrong with the test. I have also noticed under separate articles where you have mentioned that this card gets around 10 FPS in Xonotic under high settings while I always seem to get above 40FPS. Ultra crashes though.
    With kernel 3.15\3.16 and mesa-git your result must be much higer. My bonaire (hd 7790) - 1659 (27 fps)

  10. #20
    Join Date
    Jan 2014
    Posts
    18

    Default

    "Overall though, in the past ten years the open-source drivers have advanced a heck of a lot... "

    yeah, no joke, it's insane..

    I made made a naive approximation of % implemented of the OpenGL standard based on the amount of lines in GL3.txt are marked as DONE.



    Note that the people working on mesa drivers have gotten from sub 20% implementation with OpenGL4.0 as a target (16 of 85 extensions), to almost 75% towards OpenGL4.4 today (102 of 137 extension), in just 4 years! I guess we can thank intel to a large extent for that, but save for that pretty much without any help.. So that's quite an achievement!

    It seems some people think like: "The open source drivers are *not even working*", they really have to revise their expectations..

    If we are lucky we might see a feature complete mesa within 2 years, if OpenGL4.5 or something doesn't increase the target again, and after that optimization and working towards beating the closed drivers Or at least that is what I hope to see.

Posting Permissions

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