Announcement

Collapse
No announcement yet.

AMD Radeon HD 4250 880G On Linux

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

  • AMD Radeon HD 4250 880G On Linux

    Phoronix: AMD Radeon HD 4250 880G On Linux

    Back in May we looked at the AMD Radeon HD 4290 integrated graphics Linux performance of the 890GX motherboard chipset found on the MSI 890GXM-G65 and many other motherboards. This ATI Radeon 4290 (890GX) wasn't the most compelling integrated graphics processor we found with its OpenGL performance using the proprietary Catalyst driver being significantly lower than even the cheapest PCI Express graphics cards. X-Video Bitstream Acceleration, the method whereby AMD exposes their UVD2 engine capabilities on Linux, was also rather useless (and to this day still is the case) that makes this an ineffective Linux Home Theater PC too. Since the launch of the 890GX chipset, AMD rolled out the 880G chipset as a less-expensive solution for motherboard vendors and offers a stripped-down graphics processor of what is found in the 890GX and is branded as the ATI Radeon HD 4250. Today we have a few benchmarks of the Radeon HD 4250 (880G) under Linux for your viewing pleasure.

    http://www.phoronix.com/vr.php?view=15181

  • #2
    The plus side going to ATI hardware is the open-source support, for which the 880G/890GX chipsets there is kernel mode-setting, the classic Mesa DRI driver, and an emerging Gallium3D driver, but with this IGP hardware not being the fastest with the performance-optimized Catalyst driver, the open-source performance is sadly worse.
    It would be nice to see a comparison of Catalyst to the OSS stack anyway, to see where we are and how far we still have to go.

    Comment


    • #3
      Originally posted by pingufunkybeat View Post
      It would be nice to see a comparison of Catalyst to the OSS stack anyway, to see where we are and how far we still have to go.
      Yes, that could be usefull, too.

      I don't get it why HD4290 is even worse than HD4250. I thought 4290 is the most powerfull AMD IGP chipset.
      And this resolution comparision look strange that at 1024x768 and 1280x1024 resolutions you've get less FPS than on 1680x1050 and higher.

      Comment


      • #4
        IMO the strange performance in 1024x768 and 1280x1024 resolutions is a bug in the drivers. Was everything tested on the same stack (including Catalyst version)? Anyway, HD4290 should be faster than HD4250...

        Comment


        • #5
          Yeah - it's like some memory settings are being handled wrong.

          Since there are unified graphics shaders it might make sense to do vertex shading (up to a point) on the CPU, too...

          Comment


          • #6
            There's no mention in the article about sideport memory.
            A quick research show that both motherboards have 128 Mb of 1333 Mhz DDR3 sideport memory. Maybe Asrock is timing them more aggressively (usually they do this to appear in the first positions of the reviews).

            Comment


            • #7
              Conclusion - get an AMD 770/870 board and a GT220.

              Comment


              • #8
                Originally posted by DanL View Post
                Conclusion - get an AMD 770/870 board and a GT220.
                Get an nVIDIA video card.

                Comment


                • #9
                  It is unfortunate that AMD's Linux/X.org support remains half-heartedly inadequate (although basically functional).

                  Their IGP and (upcoming) "Fusion" combos would be excellent in terms of performance and cost-effectiveness everywhere where MS-Windows licensing fees and bloat favour Linux and OSS.

                  How would ULV-level performance at or near netbook cost sound like? Now how about the same but without proper power management and with significant features missing? That's what I thought.

                  Same issues affect AMD's desktop offerings. Potentially great (with Fusion potentially ideal), but kneecapped by AMD's own support.

                  Intel's GMA offerings starting from their low-end, while pricier than AMD's, enjoy solid OSS support so AMD's price advantage is lost before the race even began. That's in the high volume segment. At the other end the OSS-hostile Nvidia's proprietary drivers beat AMD/ATI's own preferred proprietary support as well...

                  Sadly I see no signs of Fusion gaining proper (parity) Linux/X.org support so what could be a game changer whimpers into a relative non-event. I can't even remember the last time I saw an AMD-based box or notebook preloaded with Linux by an OEM or a local shop. I hope I'm wrong as I'd really like welcome AMD back to my personal and recommended hardware list.

                  Comment


                  • #10
                    Other than video decode acceleration, what features do you think are missing from the proprietary drivers ? OEM preloads are done with fglrx, not the open drivers. The open source drivers are not supposed to replace fglrx, they are supposed to make it easy for distros to offer a great experience when installing on already-purchased hardware.

                    Re: power management and the open drivers, last time I checked there were no plans to stop work on power management - quite the opposite in fact. Did you hear something different ? I get the feeling you think that *we* are supposed to doing all of the open source driver development; that was not the plan originally and is not the plan today. We provide programming information, help with initial support, and provide direct developer support to the community but we are emphatically *not* writing the open source drivers ourselves.

                    Also note that Q is not actually the official keeper of the AMD product release calendar, despite appearances

                    Comment


                    • #11
                      If the AMD proprietary drivers are approaching feature parity with MS platforms, that sounds like good news for that set of Linux/Xorg users.

                      I admit not having followed the very latest goings-on since I'm not personally comfortable depending on manufacturers for closed-source drivers, although comparing feature sets and performance between Linux and MS platforms may be interesting.

                      OSS driver support however is what interests me and I'm glad that gains are being made and more older and some recent cards/chipsets have become quite usable under cutting egde settings. Throwing docs out there has certainly helped, but AFAIK Intel's OSS support remains well ahead (although not flawless by any means). The fast-moving graphics tech isn't ideal for purely volunteer-based development without significant input from the manufacturers (and their full-time team who have early access and advantage of grokking the whole ball of hair better).

                      I do find the following statement quite disheartening.

                      "The open source drivers are not supposed to replace fglrx, they are supposed to make it easy for distros to offer a great experience when installing..."

                      If that is AMD's official approach I'm afraid they're missing the actual point of OSS.

                      That said it'd be interesting to see reviews comparing OSS graphics support (performance, features, power mgmnt etc.) by both Intel (G3x/G4x/HD) and AMD (IGPs and 4xxx/5xxx/? cards) with the straitjacketed Catalyst driver also thrown in.

                      Comment


                      • #12
                        Originally posted by misGnomer View Post
                        Throwing docs out there has certainly helped, but AFAIK Intel's OSS support remains well ahead (although not flawless by any means).
                        That doesn't seem necessarily the case: r300-r700 have actively developed Gallium drivers with several community contributors (as well as the classic drivers), while the Intel Gallium drivers appear completely dormant, and the classic driver seems to have no contributors outside Intel.

                        Also, nVidia/ATI cards are generally 10-50 times faster than Intel integrated cards, so it's kind of pointless to support 3D graphics on the Intel cards.

                        Comment


                        • #13
                          Originally posted by misGnomer View Post
                          I do find the following statement quite disheartening.

                          "The open source drivers are not supposed to replace fglrx, they are supposed to make it easy for distros to offer a great experience when installing..."

                          If that is AMD's official approach I'm afraid they're missing the actual point of OSS.
                          Back when I could edit posts I used to write more nuanced responses. Now posting anything more than "your distro sucks, mine is great" is a big pain in the butt.

                          I expect that the open drivers will become the driver of choice for a significant segment of the Linux consumer market, but as long as there are users expecting complete feature & performance parity between Linux and other OSes there is also going to be a requirement for a proprietary driver that uses code sharing with other OSes to deliver a high degree of performance and functionality into a market which could not otherwise support the cost of writing and maintaining such a driver.

                          For the last few years the Xorg community has been putting a big chunk of their efforts into a major re-architecture of the open source driver stack (KMS, GEM/TTM, DRI2, Gallium3D etc..). This had two consequences -- one obvious (fewer devs available to work on other stuff) and one non-obvious (some things just didn't make sense to implement because the work would have to be mostly thrown away before completion as a result of the architecture changes). Power management was one of the non-obvious victims - implementing anything other than the most basic "don't melt my GPU" code before the move to KMS was completed would have been an almost total waste of resources.

                          Now that the re-architecture work is largely completed I think you will see relatively faster progress on features & performance of the open source stack, probably closer to what you were hoping for in the first place.

                          Comment


                          • #14
                            Fair enough. I'll go with the known (OSS) supportable hardware and steer away from the frustratingly mouthwatering Fusion news but will be open to give AMD a try when that day comes. And bridgman, I'm a number not a consumer dammit!

                            Comment


                            • #15
                              The 890GX numbers are really strange. Reminds me of the 785G that was actually faster than the 790GX.

                              Comment

                              Working...
                              X