Announcement

Collapse
No announcement yet.

11-Way AMD Radeon GPU Comparison On Linux 3.12, Mesa 9.3

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

  • 11-Way AMD Radeon GPU Comparison On Linux 3.12, Mesa 9.3

    Phoronix: 11-Way AMD Radeon GPU Comparison On Linux 3.12, Mesa 9.3

    After last week delivering new AMD Radeon HD 7000 Gallium3D benchmarks from the Mesa 9.3 development driver and the Linux 3.12 Git kernel, up today is a much larger open-source AMD Radeon graphics comparison using the latest code that's yet to be officially released. From the Linux 3.12 kernel and Mesa 9.3-devel, eleven different AMD Radeon graphics cards spanning multiple generations were compared with the latest open-source Linux graphics driver code.

    http://www.phoronix.com/vr.php?view=19141

  • #2
    Good thing I own the HD 6950 then!

    Comment


    • #3
      So what we learn here is that in order to get good gaming performance and 3D application support you need Catalyst ( no suprise). Which means most desktop environments under linux for people wanting to play games are going to be a bit buggy and unstable from time to time meaning they will have a negative experience of linux. I keep reading how Devs don't use the catalyst driver before testing released and i die a bit inside.. the linux desktop should be stable with the offical driver too.

      Devs should make things better for all people and even though I applaud the opensource driver and its 2D desktop stability it may never reach the heights of a closed source driver. I wonder what Valves take on drivers and vendors will be in just a few hours from now with the potencial steambox revealing? Nvidia has total closed source and as much as people hate that it is almost on par with windows 7/8 in performance.. yet 2d and desktop integration can suffer for the same reasons as AMD.

      I expect the only way to get linux to work for all the people all the time is to Fork and close off certain things

      Comment


      • #4
        Originally posted by phill1978 View Post
        So what we learn here is that in order to get good gaming performance and 3D application support you need Catalyst ( no suprise). Which means most desktop environments under linux for people wanting to play games are going to be a bit buggy and unstable from time to time meaning they will have a negative experience of linux. I keep reading how Devs don't use the catalyst driver before testing released and i die a bit inside.. the linux desktop should be stable with the offical driver too.

        Devs should make things better for all people and even though I applaud the opensource driver and its 2D desktop stability it may never reach the heights of a closed source driver. I wonder what Valves take on drivers and vendors will be in just a few hours from now with the potencial steambox revealing? Nvidia has total closed source and as much as people hate that it is almost on par with windows 7/8 in performance.. yet 2d and desktop integration can suffer for the same reasons as AMD.

        I expect the only way to get linux to work for all the people all the time is to Fork and close off certain things
        This is such fud. If AMD cared, they could increase the development team on Gallium3d ten fold for 3 months and have complete GL 4.4 support, every bug fixed, diverse graphical tools, and every gpu they ever made supported with better performance than on Windows or by Nvidia.

        Hell, if every gpu developer came together and each contributed 5 developers a piece (Texas Instruments, ARM, Intel, Qualcomm, Nvidia, AMD, etc) they would all have extremely per formant fluid highly compatible drivers on every device under one implementation in the same way Linux supports dozens of cpu architectures and kicks the butt of every other kernel in terms of performance and feature parity.

        At the end of the day the proprietary drivers win because these companies are putting all their money and developers on them. In practice, a shared opengl implementation with deep kernel integration and all the magnificent multifunctional behavior that arises from that (being able to stick d3d on top of kms rather than ogl, for example) would kick the crap out of these dozens of megabyte binary blob monstrosities.

        It is just that businesses are very anti-freedom in general, and they trend towards closed source even when it makes zero sense and demonstrably reduces their products value by making it a black box with no documentation or information for the user. It is like giving someone a car with a soldered shut hood. I'm not getting a new laptop, for example, until next years Broadwell, when Crystalwell 2.0 should be very competitive in the graphics space, since Intel are the only ones doing GPUs right (full stack FOSS that still has the performance numbers, even if they should transition to gallium3d). But I still have a bad taste in my mouth buying Intel stuff because they are still hugely anti-freedom with their proprietary chipsets, trade secret designs, and scrambled microcode behavior on their chips.

        That is actually humorously in parallel to AMD, who support coreboot on their chipsets and usually are more open with their cpu hardware, even if they still use proprietary designs and implementations.

        Wish I could get an AMD mobo with an Intel GPU =P
        Last edited by zanny; 09-23-2013, 11:45 AM.

        Comment


        • #5
          Originally posted by phill1978 View Post
          So what we learn here is that in order to get good gaming performance and 3D application support you need Catalyst ( no suprise).(
          No. Catalyst was not even benchmarked in this comparison. Besides, all the "average user" requires is 60fps at 1920x1080, which is achievable with the open source drivers according to benchmarks published previously on Phoronix. (According to Steam's hardware survey, 96% of gamers are gaming at 1920x1080 or lower)

          Comment


          • #6
            Were the kernel params modified? i.e.:

            radeon.pcie_gen2=1 radeon.dpm=1

            Comment


            • #7
              Originally posted by FourDMusic View Post
              Were the kernel params modified? i.e.:

              radeon.pcie_gen2=1
              not needed as its enabled by default since 3.6
              radeon.dpm=1
              yes, as stated and shown on the very first page

              Comment


              • #8
                Originally posted by zanny View Post
                This is such fud. If AMD cared, they could increase the development team on Gallium3d ten fold for 3 months and have complete GL 4.4 support, every bug fixed, diverse graphical tools, and every gpu they ever made supported with better performance than on Windows or by Nvidia.
                That is such a big bullshit, it's not even funny. You don't hire a bunch of people and they are immediately productive. You don't hire software engineers and they are immediately capable of writing solid, high performance drivers. You don't hire software engineers and they know every nook and cranny about a until-then-unknown hardware.
                If you have a recipe for training people as fast as you claim, please share it or open a development company and earn big money. Then you can fund all those open source projects.

                Hell, if every gpu developer came together and each contributed 5 developers a piece (Texas Instruments, ARM, Intel, Qualcomm, Nvidia, AMD, etc) they would all have extremely per formant fluid highly compatible drivers on every device under one implementation in the same way Linux supports dozens of cpu architectures and kicks the butt of every other kernel in terms of performance and feature parity.
                Why should any company give up a competitive advantage?

                It is just that businesses are very anti-freedom in general, and they trend towards closed source even when it makes zero sense and demonstrably reduces their products value by making it a black box with no documentation or information for the user.
                Business are to earn money for the shareholders. Nothing else.

                It is like giving someone a car with a soldered shut hood.
                So for your or any modern car you are getting a complete repair manual with cheap tools?
                Don't think so or let me know model and brand. Though this will make a lot of mechanics very unhappy.

                I'm not getting a new laptop, for example, until next years Broadwell, when Crystalwell 2.0 should be very competitive in the graphics space, since Intel are the only ones doing GPUs right (full stack FOSS that still has the performance numbers, even if they should transition to gallium3d). But I still have a bad taste in my mouth buying Intel stuff because they are still hugely anti-freedom with their proprietary chipsets, trade secret designs, and scrambled microcode behavior on their chips.

                That is actually humorously in parallel to AMD, who support coreboot on their chipsets and usually are more open with their cpu hardware, even if they still use proprietary designs and implementations.

                Wish I could get an AMD mobo with an Intel GPU =P
                AMD is putting in a lot of money (seen from their revenue perspective) and time and you're buying Intel. Yeah, that is a hell of a support for AMD ...

                Comment


                • #9
                  A question

                  I have a ATI HD 6670... Why is never considered in benchmarks? Is it similar to other video card? perhaps 6570 or 6770?

                  Thanks in advance

                  Comment


                  • #10
                    Originally posted by tulioserpio View Post
                    I have a ATI HD 6670... Why is never considered in benchmarks? Is it similar to other video card? perhaps 6570 or 6770?
                    IIRC the 6570 and 6670 use the same GPU but the 6570 usually ships with slower memory than the 6670 (I'm not sure what kind of RAM is on Michael's HD 6570). Assuming it doesn't get hit by whatever is slowing down the 6570, I would expect your HD 6670 to fall ~midway between 6450 and 6770.

                    Comment


                    • #11
                      Originally posted by chrisb View Post
                      No. Catalyst was not even benchmarked in this comparison. Besides, all the "average user" requires is 60fps at 1920x1080, which is achievable with the open source drivers according to benchmarks published previously on Phoronix. (According to Steam's hardware survey, 96% of gamers are gaming at 1920x1080 or lower)
                      I am not an "average user". I want the same or at least comparable performance with the same settings as the Catalyst driver uses, I don't buy expensive hardware to get sub-par performance. If in a game in 1920x1080 with all settings to maximum the Catalyst driver has room to enable visual enhancements like MSAA and AF then the radeon driver should give a similar performance with the same visual quality. From what I get from comments on this forum at this point it is not even possible to enable/disable MSAA, it is kind of an automatic function. If that is wrong please correct me.

                      Having said that, all in all I am confident that we will get there in the not so far future, the radeon developers are doing a tremendous job.

                      Comment


                      • #12
                        Originally posted by bridgman View Post
                        IIRC the 6570 and 6670 use the same GPU but the 6570 usually ships with slower memory than the 6670 (I'm not sure what kind of RAM is on Michael's HD 6570). Assuming it doesn't get hit by whatever is slowing down the 6570, I would expect your HD 6670 to fall ~midway between 6450 and 6770.
                        So the real question is:

                        Has the HD 6670 the same problems the HD 6570 has or not?

                        Thanks for the answer.

                        Comment


                        • #13
                          Personally,

                          I never expect to run anything but Catalyst for 3D until AMD itself open sources the entire stack and support for FreeBSD, Linux, etc, whilst focusing on n-tier selling of service development for the likes of Adobe, Apple, Oracle, NSA, CIA, DoE, etc within their HSA initiative.

                          As one astute poster mentioned about large amounts of resources [relative to the present value of the company: Only Apple could dwarf Intel on R&D and beat them to the punch] already being leveraged, more will ramp up with the expansion of sales soon to be realized over the next several years.

                          Anyone complaining about their lack of effort is most likely the same person who wishes to pay zero taxes for services.

                          You want to help, one has lots of tools to debug and help narrow suspect areas of development that can expedite quicker release cycles.

                          Comment


                          • #14
                            Originally posted by tulioserpio View Post
                            So the real question is:

                            Has the HD 6670 the same problems the HD 6570 has or not?
                            I am not sure what is wrong with Michael's Radeon HD 6570, I don't believe this is a problem that applies to all of these GPUs. Must be something specific to the card manufacturer. A XFX Radeon HD 6670 works just fine here in terms of performance, and should be noticeably (20%) faster than the 6570.

                            Comment


                            • #15
                              This review reminds me why I still use nVidia with blob drivers.

                              Good work on AMD's behalf though, they're really pushing forwards with their drivers. I think AMD could fix this driver cycle issue by making the hardware and firmware more complex and making drivers more simple. I realise in time, this might come true, but for now this is just a pipe dream.

                              Comment

                              Working...
                              X