Announcement

Collapse
No announcement yet.

AMD 8.41.7 Display Driver Released -- The Holy Crap Edition!

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

  • Ookay folks -- had a few extra minutes this afternoon when I got home from work -- and finally got a chance to chase my issues with X1650Pro AGP card.

    1) 8.34.8 fglrx ****WORKS****

    my X1650Pro is up and running in AGP mode 8 with DRI with opengl with only some weirdness

    I'm running gentoo-sources 2.6.22 - the driver fglrx will not compile with the default ebuild in gentoo tree - copy it to an overlay and add

    this line
    Code:
    epatch "${FILESDIR}"/8.37.6/fix-ioctl-for-2.6.22.patch
    to the ebuild in the stanza that looks like

    Code:
    # This patch is conditional because it contains an api change.
            if kernel_is ge 2 6 20; then
                    epatch packages/Fedora/ati-drivers-2.6.20.patch
            fi
    then ebuild {full path to the overlay ebuild file} digest
    emerge -v ={fully qualified ebuild}

    and reboot for good measure

    NOW -- caveat on this

    DO NOT SWITCH BACK TO CONSOLE after X starts. this locks my machine.

    I will now walk the driver tree and see if I can spot the regression.

    Please note that I've udpated the bug with the details.

    Comment


    • Okay -- walked through all the drivers in portage for my system with the X1650Pro --did some back to back testing (where the driver initialized things on the X1650Pro)

      Keep in mind I'm on Gentoo - so I'm sticking to ebuilds that are available - with a few fixes since the ebuilds typically fit the kernel that was current at the time, I've made minor fixes to get the driver to build with my 2.6.22 kernel.

      8.35.5 also enables the card in AGP mode 8

      However both 8.35.5 and 8.34 driver provide horrendous performance under
      load.
      glxgears and fgl_fglrxgears both run slower under these drivers with the
      X1650Pro than with the 9550

      I have not been able to test on 8.36 {I was unable to get a 64 bit build to
      complete}
      8.37 enables AGP mode, but the DRI setup never completes and the X setup
      does not complete nor do I get a display. (black screen lockup)
      8.38 never made portage - if it exists.
      8.39 fails to initialize AGP mode.
      8.40 fails to initialize AGP mode.
      8.41 fails to initialize AGP mode.

      Comment


      • You can use Kanotix Thorhammer to test the drivers, much more easy using

        install-fglrx-debian.sh -v VER

        or even in live mode with

        fglrx=v:VER

        The script is fully tested with lots of drivers and Xorg 7.1 is old enough to test even drivers for DX8 cards. Of course you can use the hard way too

        Comment


        • 8.41.7 works not so bad with my Sapphire Radeon 9600 256MB (not Pro) AGP on Kubuntu 7.04.
          I have ~2450 FPS on glxgears and ~500 FPS on fgl_glxgears with Dual Head (LCD 1280x1024x24 + CRT 1280x1024x24 + TV 1024x768x24).
          The only game that I've tried yet is Nexuiz 2.3.
          It works great (from ~40 to ~90 FPS with only disabled Vertex Mapping and World Dynamic Lights - these two options degrade FPS, everything else enabled).
          Game segfaults when I apply video changes in menu but no display corruption and other problems.
          Xv works well (only ~3% of CPU use when playing movie with MPlayer)

          I've managed to change display order with ForcedMonitors option in Device section but the driver doesn't react on this setting.
          The drivers uses CRT1 as Display1, TV as Display2 and TMDS1 as Display3 and CRT1+TMDS1 works as clone and TV is the second head. I wanted to use TMDS1 as Display1 (one head) and CRT1 as Display2 with TV as Display3 in clone second head but the driver doesn't look for ForcedMonitors option.

          Comment


          • Originally posted by xeros View Post
            8.41.7 works not so bad with my Sapphire Radeon 9600 256MB (not Pro) AGP on Kubuntu 7.04.
            Thats weird. I can't get my Radeon 9800xt to work. I can get X to start, an glxgears works, but anything that uses textures comes back with an x display error. Also, and this is way out there, but my gnome session wouldn't work until I moved .gconf out of the way and let it build a new one. I though this problem was isolated to gnome, but when I went back to 8.40.6 I put my old .gconf back, and it worked. I was getting Connection reset by peer errors in my .xsession with my old .gconf and 8.41.7.

            Comment


            • So it turns out my girlfriend bought a new laptop and the machine comes equipped with an Xpress 1100 IGP, anyone with a card like this tried the 8.41.7's? Or should we stick with the 8.40 for now?

              Comment


              • Originally posted by Alistair View Post
                Ookay folks -- had a few extra minutes this afternoon when I got home from work -- and finally got a chance to chase my issues with X1650Pro AGP card.

                1) 8.34.8 fglrx ****WORKS****

                my X1650Pro is up and running in AGP mode 8 with DRI with opengl with only some weirdness

                I'm running gentoo-sources 2.6.22 - the driver fglrx will not compile with the default ebuild in gentoo tree - copy it to an overlay and add

                this line
                Code:
                epatch "${FILESDIR}"/8.37.6/fix-ioctl-for-2.6.22.patch
                to the ebuild in the stanza that looks like

                Code:
                # This patch is conditional because it contains an api change.
                        if kernel_is ge 2 6 20; then
                                epatch packages/Fedora/ati-drivers-2.6.20.patch
                        fi
                then ebuild {full path to the overlay ebuild file} digest
                emerge -v ={fully qualified ebuild}

                and reboot for good measure

                NOW -- caveat on this

                DO NOT SWITCH BACK TO CONSOLE after X starts. this locks my machine.

                I will now walk the driver tree and see if I can spot the regression.

                Please note that I've udpated the bug with the details.
                The irony of this all is of course, that my X1950Pro AGP (with PCI-e bridge chip) only works with the 8.34 driver. Anything higher locks up my X session, i can ssh in, but as soon as I kill X my box hardlocks. So the only way i can reboot is to reset eitherway.

                Looks like they did 'something' starting with 8.35 if you ask me.

                Comment


                • Originally posted by Thetargos View Post
                  So it turns out my girlfriend bought a new laptop and the machine comes equipped with an Xpress 1100 IGP, anyone with a card like this tried the 8.41.7's? Or should we stick with the 8.40 for now?
                  The brave of heart are free to try the thing- if you don't have some serious Linux expertise under your belt you might want to hold off on anything past 8.40 for fglrx drivers...

                  Comment


                  • When you do it right, switching fglrx drivers is no big deal. Why don't you try the driver yourself?

                    Comment


                    • Originally posted by Kano View Post
                      When you do it right, switching fglrx drivers is no big deal. Why don't you try the driver yourself?
                      Heh... I plan on it- perhaps sometime this week/weekend if time permits to gain intel on the IGP series part behaviors with the codebase. Knowing something of things in this new driver, I'm sure there's some sharp edges on the IGP lineup like there's been with the AGP series of parts. Something that gives them reason to say it's NOT supported for most of the lineup.

                      Comment

                      Working...
                      X