Announcement

Collapse
No announcement yet.

Open-Source ATI 6.8.0 Driver Released

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

  • #16
    I can get some hw accel with fglrx on an AGP X1600 pro by using

    mplayer -vo gl:yuv=4:lscale=1

    Comment


    • #17
      Originally posted by Kano View Post
      You dream too much, there was never divx acc available, not even with win!
      not with windows or fglrx driver but with radeon is there. i can assure you.
      when playing divx-5 movies with xine and radeon i get the processor at 10% of the lower speed (800mhz) and a top thermal temp of 58° C.
      when running the same movie on windows or fglrx i get the processor at 35% of middle speed (1600Mhz) and a thermal temp of 69°C.

      when playing other sort of movies i don't get this behaviour except some xvid ones (probably coded for divx-5 compatibility). this has only one explanation:
      hw decoding capabilities.
      i got the confirmation on the xorg mailing list or on one of the radeon devs blog of the presence of hw decoding in the radeon driver. i have to find out the confirmation, but right now i'm in a hurry cause i got my favorite soccer team playing a game this evening in champions league so i'll post the link to this confirmation when i find it in the next days.

      Huh? What card are you using? Xv supported?

      This seems to contradict everything said so far.
      xpress 200m 5975 model.
      i get xv overlay, xv movie accel, font antialiasing, buggy opengl, dri forced to on, but unsupported by default and no render. i use the git versions of:
      libdrm, x11-drm-rework, xf86-video-ati, xorg-server, mesa, evdev, kdb, mouse and xcb. i have them compiled with xcb enabled and i have xinerama disabled. as movie player i use xine+kaffeine+external ffmpeg from svn. i have kaffeine and xine compiled with xcb and without xinerama. i use kde without any gnome package, with exception of gtk+ and some other needed packages for pidgin and pan. i'm on gentoo 2.6.23-r8 (the 24 branch is buggy with networkmanager and madwifi) with system on the amd64 branch while the other packages are from the unstable branch.
      there's no compiz support, but damage, compositing and aiglx work well. glxgears shows about 900fps and i've allowed the mapping ot mtrr registers. without it the xorg seems a little buggy.

      the vt switch is ctrl+atl+fx where x is the number of the virtual termina. normally the vt from 1 to 6 are console logins while the 7 to 11 are xorg ones. unfortunately the vt switch problem in radeon with some boards is well known and the devs are working on it.
      dinamic clocks option works very well to keep the system cool and the battery high (i got about 10% more battery time on radeon when compared to fgrlx + powerplay).

      the last good fglrx driver for this board is the december one; the january one doesn't start for the glx problem while the feb one is very angry with xorg-1.4 from git and is very slow and has a lot of artifacts even if it has rendering (100-150 fps on rendering). fglrx has opengl overlay and opengl acceleration work, while the radeon one is still faulty: for example i get a lot of problem running stellarium or google-earth. for that reason i have to run these programs with fglrx.

      Comment


      • #18
        All right! That's just too much for a newbie like me. I have ATI Mobility Radeon X1600. Will this driver work with it in the manner of running compiz, playing videos and games at some normal speed?

        Comment


        • #19
          Originally posted by givemesugarr View Post
          xpress 200m 5975 model.
          Ah, so an R300 core.

          Comment


          • #20
            Originally posted by DoDoENT View Post
            All right! That's just too much for a newbie like me. I have ATI Mobility Radeon X1600. Will this driver work with it in the manner of running compiz, playing videos and games at some normal speed?
            guess you'll have to try it out. without trying you cannot know. the board should be supported but maybe advanced features like compiz or high-level opengl acceleration may still not be at a high level. for videos and normal use (web browsing, listen to music, reading documents and so on) it should be satisfying. for games i don't know if there really are games that run pretty well with either fglrx or radeon.

            Ah, so an R300 core.
            r300 chipset architecture integrated into motherboard without vertex shaders and without dedicated memory (it's quite a bad board itself). it is similar to the rs690 as core arch so one should benefit from the other's specs and the same goes for driver implementation.

            Comment


            • #21
              There is no open source 3d or video acceleration for X1600 yet. We're working hard to change that.

              Comment


              • #22
                Originally posted by givemesugarr View Post
                r300 chipset architecture integrated into motherboard without vertex shaders and without dedicated memory (it's quite a bad board itself).
                Well... Not ALL iterations of the Xpress 200M GPU were UMA only. I know, my previous laptop from HP was one with "Sideport" memory, which was 128Mb of dedicated VRAM of the traditional type combined with either UMA or using UMA only. I could never get any of the modern fglrx drivers to work with anything other than UMA only mode with it, so, with all the other drawbacks, it got relegated to be my mother's computer, since it was more than serviceable under XP.

                What I want to know is what possessed ATI to make such stinkers in the first place- when they had the lead with the R200 mobility parts and they DID have T&L hardware? (No, I don't expect an answer...I just expect better from you guys going forward... )

                Comment


                • #23
                  Originally posted by Svartalf View Post
                  Well... Not ALL iterations of the Xpress 200M GPU were UMA only. I know, my previous laptop from HP was one with "Sideport" memory, which was 128Mb of dedicated VRAM of the traditional type combined with either UMA or using UMA only. I could never get any of the modern fglrx drivers to work with anything other than UMA only mode with it, so, with all the other drawbacks, it got relegated to be my mother's computer, since it was more than serviceable under XP.

                  What I want to know is what possessed ATI to make such stinkers in the first place- when they had the lead with the R200 mobility parts and they DID have T&L hardware? (No, I don't expect an answer...I just expect better from you guys going forward... )
                  As much as we dislike them, something like 70% of computers sold have IGP graphics on them. When graphics are more or less free when you buy the chipset, it make a compelling argument if you are an OEM

                  Comment


                  • #24
                    Originally posted by Svartalf View Post

                    What I want to know is what possessed ATI to make such stinkers in the first place- when they had the lead with the R200 mobility parts and they DID have T&L hardware? (No, I don't expect an answer...I just expect better from you guys going forward... )
                    It's call transistor count... fitting a TCL or Vertex processor onto an integrated chipset device requires lots of transitors. Also for that market segment they mostly don't expect tcl/vertex shaders because Intel didn't do it either.

                    Now that shaders are unified they will probably just put less of them on future integrated chipsets and you can do TCL with some of the threads. again what Intel have done with 965.

                    Comment


                    • #25
                      Originally posted by agd5f View Post
                      As much as we dislike them, something like 70% of computers sold have IGP graphics on them. When graphics are more or less free when you buy the chipset, it make a compelling argument if you are an OEM
                      Won't argue the compelling argument there- I've had to deal with some of the board vendors (ECS...) directly in the past when one of my other business ventures was doing embedded Linux for Internet Appliances. The only REAL complaint would be that the cuts to get the transistor counts down were...iffy...at best in the case of ATI's chips. The reality is that they managed to HAVE TCL support in the previous lineup of chips, even if it wasn't Vertex shader based, in spite of that argument.

                      Right now, it's water under the bridge, really.

                      Comment


                      • #26
                        Originally posted by airlied View Post
                        It's call transistor count... fitting a TCL or Vertex processor onto an integrated chipset device requires lots of transitors. Also for that market segment they mostly don't expect tcl/vertex shaders because Intel didn't do it either.
                        Heh... The problem with that thinking is that they managed to HAVE TCL in the previous series without this issue coming up. That's just...frustrating.

                        As for the market segment, they're the ones that actually MADE it (In laptops, that is...) in the first place with the Radeon Mobility lineup. Why slack off to what the competitor can manage when you're already vastly better than they are in the first place?

                        Now that shaders are unified they will probably just put less of them on future integrated chipsets and you can do TCL with some of the threads. again what Intel have done with 965.
                        Heh... Like I said... Just DO better...
                        Last edited by Svartalf; 02-20-2008, 11:05 PM.

                        Comment


                        • #27
                          Originally posted by Svartalf View Post
                          Heh... The problem with that thinking is that they managed to HAVE TCL in the previous series without this issue coming up. That's just...frustrating.

                          As for the market segment, they're the ones that actually MADE it (In laptops, that is...) in the first place with the Radeon Mobility lineup. Why slack off to what the competitor can manage when you're already vastly better than they are in the first place?



                          Heh... Like I said... Just DO better...
                          None of ATI integrated chipsets have ever had TCL support for my knowledge, Mesa always disabled TCL on all the ATI integrated chipsets. So transistor count is still the problem, you have more space on a dedicated GPU die than on an integrated one.

                          Dave.

                          Comment


                          • #28
                            Originally posted by airlied View Post
                            None of ATI integrated chipsets have ever had TCL support for my knowledge, Mesa always disabled TCL on all the ATI integrated chipsets. So transistor count is still the problem, you have more space on a dedicated GPU die than on an integrated one.

                            Dave.
                            Heh... I offer this as evidence and it comes from the horse's mouth as it were:

                            http://ati.amd.com/products/comparis...omparison.html

                            The 7500 and 9000 mobility parts offered the Charisma Engine (TCL) and the 9000 provides Shader Model 1.4 support.

                            What happens if you use driconf to override the settings on the configuration? It's not signalling any warnings on start up nor is it crashing when I force it to use hardware TCL support on the laptop I've done this to.

                            Comment


                            • #29
                              Originally posted by Svartalf View Post
                              Heh... I offer this as evidence and it comes from the horse's mouth as it were:

                              http://ati.amd.com/products/comparis...omparison.html

                              The 7500 and 9000 mobility parts offered the Charisma Engine (TCL) and the 9000 provides Shader Model 1.4 support.

                              What happens if you use driconf to override the settings on the configuration? It's not signalling any warnings on start up nor is it crashing when I force it to use hardware TCL support on the laptop I've done this to.
                              Those are discrete parts. They have fully functionality. The IGP parts are different.

                              IGP chips:
                              ATI Radeon™ X1200 series
                              ATI Radeon™ XPRESS series
                              ATI Radeon™ 9100 IGP
                              ATI Radeon™ 9000 IGP
                              ATI Radeon™ 7000 IGP
                              ATI Radeon™ 340M IGP
                              ATI Radeon™ 320M IGP

                              Comment


                              • #30
                                Ah... That sort of explains things a bit, then.

                                Looking at a more in-depth comparison of the varying parts, the IGP parts not only were crippled in TCL, but also crippled in the fragment shader department as well. <*sigh*> That just makes for "fun" in trying to support stuff for titles and applications. Ugh...

                                I'll drop the gripes about that aspect of things- but it still goes back to the "why ever have done it" that I really
                                asked at the beginning of this aside. It's frustrating to have something claim to be 3D but end up being pretty much
                                like the Savage of this era of chips in hand- more of a decellerator than an accelerator... >:-)
                                Last edited by Svartalf; 02-21-2008, 10:37 AM.

                                Comment

                                Working...
                                X