Announcement

Collapse
No announcement yet.

Intel SNB Linux Driver Can Out Run Windows Driver

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

  • #11
    Originally posted by d2kx View Post
    Am I the only that, just seeing the first two screenshots of Windows 7 and Ubuntu on the first page, thinks that Windows visually looks like total crap in comparison to Ubuntu?
    Yeah, I thought so too. Not to mention in terms of functionality and space taken on disk. I never actually seen the two side by side so windows 7 never struck me as terribly ugly. However, I'm sure deanjo is just around the corner to tell you why windows 7 looks so much better than ubuntu.

    Back on topic, nice to see at least one linux graphics driver outperforming the windows equivalent

    Comment


    • #12
      Originally posted by Znurre View Post
      OnT: Interesting numbers, indeed.
      For my next laptop I might go for a Sandy bridge/Ivy bridge solution instead of NVidia.
      I'd wait for the mobile version of the A8-3550.

      Comment


      • #13
        on Lightsmark's side Linux's performance actually better than Windows.
        it is the regression that causing performance lag.

        just revert this commit, double performance




        actually Linux driver now supports more extensions than Windows. Only thing is left performance optimizations on specific extensions.

        Comment


        • #14
          Originally posted by Kivada View Post
          I'd wait for the mobile version of the A8-3550.
          Question is which one would actually give me best 3D performance with open source drivers.
          I am not all that sure it would be the integrated HD 6550 of A8-3550.

          Comment


          • #15
            Originally posted by smitty3268 View Post
            Quote Originally Posted by phoronix View Post
            such as S3TC texture compression support and other patented features that cannot be implemented in the open-source stack.
            Correction - this is already implemented in the open-source stack, you just have to turn it on.

            I'm not sure if this says more about the Intel Linux driver or the Intel Windows driver. But it's good news to have the hardware be competitive in Linux, I suppose, like the proprietary drivers do for ATI and NVidia hardware.
            One thing that isn't clear to me... wouldn't the rights to use the patented technologies have been paid for by purchasing the hardware? [it's not like vendors sell their drivers]

            I understand that there might be an interest to protect proprietary information behind a technology, or hide implementation details of DRM protections... but if there is an independent S3TC implementation wouldn't I have the right to use it simply because I have the hardware

            [yes I realize legal agreements are hardly written that way...]

            Comment


            • #16
              Originally posted by apholux View Post
              Intel recently announced 15.22 driver (15.​22.​1.​2361/15.​22.​1.​64.​2361) with OpenGL3.1 support - was this version used to capture performance on Windows?
              You guys should really address this. The 2361 build driver increases performance not just in OpenGL, but DX10-based games as well. Some testing done here: http://communities.intel.com/message/125916

              OpenGL performance increase is massive because 2361 not only enables OpenGL3.1, but hardware T&L/VS support(sounds similar to the Linux VS thread patch). I don't know why DX performance increased but it sounds like similar things might be going on too. According to 3DMark06 Batch Size tests, it has good gains in the smaller sizes, from which I've read shows how it should perform in unoptimized code.

              So Windows is getting outperformed not because Linux stack is better, but because Windows didn't receive the same updates Linux got yet.

              Another benchmark here: http://translate.googleusercontent.c...Shkxj-ntkd1EvQ

              DX10 games show significant gains.

              Comment


              • #17
                Originally posted by Craig73 View Post
                One thing that isn't clear to me... wouldn't the rights to use the patented technologies have been paid for by purchasing the hardware? [it's not like vendors sell their drivers]
                Licenses are very different written from time to time. Some Licenses are bound to the hardware, some licenses are bound to the software, and some licenses are bound to the combo thereof.
                It seems like nobody has been able to really figure out what the S3TC is bound to, and if it differs from manufacturer to manufacturer for the cards(i.e not Intel/nVidia/AMD, but ASUS, Gigabyte...), for the chips (Intel, AMD, nVidia), or some other rather random occurance. This just seems to be a mess no-one really wants to pick unless they really have to AFAIK.

                Comment


                • #18
                  Originally posted by DavidC View Post
                  You guys should really address this. The 2361 build driver increases performance not just in OpenGL, but DX10-based games as well.

                  OpenGL performance increase is massive because 2361 not only enables OpenGL3.1, but hardware T&L/VS support(sounds similar to the Linux VS thread patch). I don't know why DX performance increased but it sounds like similar things might be going on too. According to 3DMark06 Batch Size tests, it has good gains in the smaller sizes, from which I've read shows how it should perform in unoptimized code.

                  So Windows is getting outperformed not because Linux stack is better, but because Windows didn't receive the same updates Linux got yet.
                  DX10 games show significant gains.
                  I didn't read your message, but I suspected it and I did some tests:


                  Windows driver is still faster than linux one.
                  ## VGA ##
                  AMD: X1950XTX, HD3870, HD5870
                  Intel: GMA45, HD3000 (Core i5 2500K)

                  Comment

                  Working...
                  X