Announcement

Collapse
No announcement yet.

Intel Linux Driver Almost Neck-And-Neck With Windows 8.1

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • s_j_newbury
    replied
    Originally posted by Michael View Post
    All details can be found via Openbenchmarking.org.
    I must be completely blind, but I've never been able to find the link to the openbenchmarking.org results for your articles when you don't explicitly link to them in the text. Can't you just say whether you're running with powersaving, normal or performance perf_bias? Is this recorded in the openbenchmarking.org sysinfo? I can't see it anywhere.

    Leave a comment:


  • Michael
    replied
    Originally posted by s_j_newbury View Post
    Michael, you don't say what perf_bias setting you are using...
    ...on my Ivybridge system the BIOS sets it to performance mode, but the kernel switches to 'normal':

    [ 0.000203] ENERGY_PERF_BIAS: Set to 'normal', was 'performance'
    ENERGY_PERF_BIAS: View and update with x86_energy_perf_policy(8)

    Despite the kernel log above, the "cpupower" utility is, I believe, the standard way to set performance modes:

    cpupower set -b performance

    This really helps performance here. I've no idea what the policy is in Windows, but it most likely sets it to performance mode when running on AC.
    All details can be found via Openbenchmarking.org.

    Leave a comment:


  • [Knuckles]
    replied
    Thanks for all the answers! It seems we're already there for mid and top-range SKUs, but not yet for older/low-end ones.

    Leave a comment:


  • a user
    replied
    Originally posted by Kayden View Post
    There's apparently a bad interaction between Heaven 4.0 and our implementation of ARB_sample_shading, which results in the solid white screen. If you disable that by setting the MESA_EXTENSION_OVERRIDE=-GL_ARB_sample_shading environment variable, then Heaven 4.0 works fine on Mesa 10. I've run it a lot, actually.

    Valley 1.0 has a bug where it requests dual source color blending without indicating which outputs are color 0 and color 1. Without that information (which is required by the GL spec), we have to guess, and have a 50/50 chance of getting it right. Apparently, nVidia gets lucky. I have some hacks to work around that, and then Valley works as well.
    ohhh, thank you. finally some words that clear up things a bit for me. searched like mad all over the inet and found nothing usefull regarding this issue.

    many thanks!!! tried out the heaven demo and it works... slow and choppy on my i3 4330 but it works

    Leave a comment:


  • Kayden
    replied
    Originally posted by a user View Post
    i see that only the older unigine demos are used here.

    any news when the newer ones, like heaven 4.0 will work with intel systems (haswell)? in particular the shaders are not compiling and you see lot of errors on the console when starting the heaven demo and the demo on its own renders just plain white screen.
    There's apparently a bad interaction between Heaven 4.0 and our implementation of ARB_sample_shading, which results in the solid white screen. If you disable that by setting the MESA_EXTENSION_OVERRIDE=-GL_ARB_sample_shading environment variable, then Heaven 4.0 works fine on Mesa 10. I've run it a lot, actually.

    Valley 1.0 has a bug where it requests dual source color blending without indicating which outputs are color 0 and color 1. Without that information (which is required by the GL spec), we have to guess, and have a 50/50 chance of getting it right. Apparently, nVidia gets lucky. I have some hacks to work around that, and then Valley works as well.

    Leave a comment:


  • a user
    replied
    i see that only the older unigine demos are used here.

    any news when the newer ones, like heaven 4.0 will work with intel systems (haswell)? in particular the shaders are not compiling and you see lot of errors on the console when starting the heaven demo and the demo on its own renders just plain white screen.

    Leave a comment:


  • Pseus
    replied
    Originally posted by [Knuckles] View Post
    Wow, was not expecting them to be that good.

    Has there been any test using Source games? Aka "can I play dota2/tf2" with Intel integrated graphics yet?
    Core i5 3210M + HD4000 user here. Dota 2 plays fine with medium to max settings (720p, everything maxed except shadows and no AA). TF2 is playable with medium settings, but you really need low settings if you are expecting better framerates (important for online FPS gameplay).

    Edit: using Kernel 3.12.1 and mesa 9.2.3.

    Leave a comment:


  • s_j_newbury
    replied
    Michael, you don't say what perf_bias setting you are using...
    ...on my Ivybridge system the BIOS sets it to performance mode, but the kernel switches to 'normal':

    [ 0.000203] ENERGY_PERF_BIAS: Set to 'normal', was 'performance'
    ENERGY_PERF_BIAS: View and update with x86_energy_perf_policy(8)

    Despite the kernel log above, the "cpupower" utility is, I believe, the standard way to set performance modes:

    cpupower set -b performance

    This really helps performance here. I've no idea what the policy is in Windows, but it most likely sets it to performance mode when running on AC.

    Leave a comment:


  • chrisf
    replied
    Originally posted by b3nn0 View Post
    I tried to play dota 2 on a Laptop with an Intel HD4000 about 2 months ago. It was unplayable on Linux, while on Windows it works fine on low settings.
    However, I don't know if the HD4000 for the desktop is better than the laptop version.
    I have successfully tested Dota2 now on everything from GM45 to Haswell:

    - GM45 and Ironlake require mesa 10 (or git master) for correct rendering, but are unplayably slow even with everything turned right down.
    - Sandybridge and later both work correctly from 9.1.6 onward. Sandybridge is vaguely playable with everything turned right down; Ivybridge works reasonably well with mediumish settings (I'm limited by my own poor skill rather than the machine being too slow).
    - Haswell Iris Pro runs fine with almost max settings; requires 9.1.6+ for correct rendering, but mostly tested with 10 and master. I can't comment on performance of slower Haswell variants.

    Leave a comment:


  • Michael
    replied
    Originally posted by [Knuckles] View Post
    Wow, was not expecting them to be that good.

    Has there been any test using Source games? Aka "can I play dota2/tf2" with Intel integrated graphics yet?
    See the 21 way comparison results earlier as there's Haswell graphics and a CS: Source run.

    Leave a comment:

Working...
X