Page 1 of 2 12 LastLast
Results 1 to 10 of 17

Thread: Intel Windows vs. Linux GPU Performance Q4'2010

  1. #1
    Join Date
    Jan 2007
    Posts
    14,900

    Default Intel Windows vs. Linux GPU Performance Q4'2010

    Phoronix: Intel Windows vs. Linux GPU Performance Q4'2010

    Yesterday we shared benchmarks of the ATI R600 Gallium3D driver compared against the classic Mesa R600 driver and then the proprietary AMD Catalyst driver. The proprietary driver was much faster than the open-source drivers were, but the Gallium3D driver did possess higher performance in most of the tests than with the classic Mesa driver. This is similar to the R300 Gallium3D driver being faster than its now-deprecated R300 classic driver. Meanwhile though Intel continues to back only their classic Mesa DRI driver and there are no signs of them switching over to the Gallium3D architecture anytime soon. It is not as if Intel's current Mesa driver is feature-complete and performance-optimized as our tests from earlier this year show Intel's Linux graphics performance being far behind their Windows driver. In this article though we are seeing where the Intel Mesa performance is at when using the very latest DRM and Mesa code.

    http://www.phoronix.com/vr.php?view=15491

  2. #2
    Join Date
    Jan 2009
    Posts
    80

    Default

    This is much more comparable than it used to be.

  3. #3
    Join Date
    Mar 2009
    Location
    Hellas
    Posts
    1,063

    Default

    Yeah, when I read the title I said "Meh!" but when I read the post I was surprised with the boost of Linux driver.

  4. #4
    Join Date
    Jan 2009
    Posts
    80

    Default

    I looked at the previous benchmark, and wow, the improvement is huge.

    In this benchmark, there were four tests. In two Linux is the clear loser (OpenArena, Nexuiz), but by a factor of two at the maximum, which is bad but not devastating. In one, Urban Terror, it was kinda a tie -- Linux the winner at lower resolutions, Windows the winner at higher ones. And Warsow was a clear Linux win.

    The previous benchmark of half a year ago had only two tests (which in itself implies Intel is now able to execute more tests), in one of which -- Nexuiz -- Linux was lagging behind by a factor of ~10, and the second (Warsow) was slightly behind.

    Maybe Phoromatic should start tracking mesa-git? Seems like that is where a lot of work is going currently.

  5. #5

    Default

    Quote Originally Posted by loonyphoenix View Post
    Maybe Phoromatic should start tracking mesa-git? Seems like that is where a lot of work is going currently.
    The reason why I haven't done this is due to Mesa's dependencies on libdrm being bumped, etc. And the fact that some of the magic happens within the DRM code too, so technically the optimal way to track it would be to have one system tracking the DRM changes, another system tracking the Mesa changes, and then a third system tracking the DRM and Mesa changes simultaneously. But I don't have three identical systems to dedicate towards this task at the moment.

  6. #6
    Join Date
    Jan 2009
    Posts
    80

    Default

    Quote Originally Posted by Michael View Post
    The reason why I haven't done this is due to Mesa's dependencies on libdrm being bumped, etc. And the fact that some of the magic happens within the DRM code too, so technically the optimal way to track it would be to have one system tracking the DRM changes, another system tracking the Mesa changes, and then a third system tracking the DRM and Mesa changes simultaneously. But I don't have three identical systems to dedicate towards this task at the moment.
    Can't this be done on a single machine? Three different independent systems in three different partitions, with timed reboots between them. Seems like 8 hours should be enough to run the few graphics tests available right now.

  7. #7
    Join Date
    Apr 2009
    Posts
    557

    Default Beats AMD and NVIDIA!

    This beats AMD and NVIDIA by a long shot in relative terms. At least AMD supports Open Source, but for both AMD/ATI and NVIDIA, the open source drivers are several times (2 to 10) times slower than their best (binary) drivers. For Intel, the open source drivers are in the same ballpark, which is MUCH, much better.

    In my book, Intel > AMD/ATI > NVIDIA

    I wish Intel will start producing discrete graphics card, they would become the obvious choice (right now it is AMD for me).

  8. #8
    Join Date
    Jan 2010
    Posts
    159

    Default

    Quote Originally Posted by mendieta View Post
    This beats AMD and NVIDIA by a long shot in relative terms. At least AMD supports Open Source, but for both AMD/ATI and NVIDIA, the open source drivers are several times (2 to 10) times slower than their best (binary) drivers. For Intel, the open source drivers are in the same ballpark, which is MUCH, much better.

    In my book, Intel > AMD/ATI > NVIDIA

    I wish Intel will start producing discrete graphics card, they would become the obvious choice (right now it is AMD for me).
    ...No Intel wouldn't be a choice for GPUs because ... those Intel GPUs are dead slow. And a dedicated one would consume so much power it would be unbearable.
    Just see larabee thats your example. Also Intel developers don't seem to be able to loose on CPU building habits and because of that their GPUs are truly crippled CPUs.

  9. #9
    Join Date
    Apr 2009
    Posts
    557

    Default

    Quote Originally Posted by Ragas View Post
    ...No Intel wouldn't be a choice for GPUs because ... those Intel GPUs are dead slow. And a dedicated one would consume so much power it would be unbearable.
    Just see larabee thats your example. Also Intel developers don't seem to be able to loose on CPU building habits and because of that their GPUs are truly crippled CPUs.
    Oh, I had forgotten about Larrabee! Thanks for the pointer. I would like to see some performance and pricing analysis before jumping to conclusions. But according to the performance preview in Wikipedia, it seems like they will probably not compete in the high end, at least initially:

    http://en.wikipedia.org/wiki/Larrabe...rchitecture%29

    That would still be fine with me if they can compete with AMD/ATI's pricing at the low and mid range. We'll see. Cheers!

  10. #10
    Join Date
    Jun 2009
    Location
    NL
    Posts
    24

    Default

    Oh, I had forgotten about Larrabee! Thanks for the pointer. I would like to see some performance and pricing analysis before jumping to conclusions. But according to the performance preview in Wikipedia, it seems like they will probably not compete in the high end, at least initially:
    I'd doubt we'll ever see larrabee, at least in any form resembling a discrete, graphic-focused chip. I could imagine seeing something of a coprocessor, or something like this.

    Good to see that the Intel OSS drivers are catching up (at least compared to last year's OSS drivers), and being on par with the proprietary drivers in some areas.
    It could be that the intel drivers are limited by other things than the GPU (since they show flatline performance in some games), perhaps CPU limited (bad coding in the Win7 stack?), or some sync/energy save mode kicking in?

    Seemingly, investing in OSS drivers pays off in the end. This is actually pretty close to the ~80% closed-blob performance that are often named as a maximum for OSS drivers.
    I'd argue that especially the radeon drivers have not yet caught up to their closed blob couterparts this much because their cards have a *lot* more potential to be used by the fine-tuned blobs. Relative to previous versions, the (gallium) radeon drivers' catching-up is already failry impressive, i think.

Tags for this Thread

Posting Permissions

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