Announcement

Collapse
No announcement yet.

"Ask ATI" dev thread

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

  • @LostInSpacetime: your bug might be in fglrx, or it might be in Skype... Try using Skype without fglrx (use the FOSS ati/radeon/radeonhd driver) to see if you can reproduce the bug.
    You can also try disabling composite in fglrx (it doesn't deal well with it).

    @mirv: there is one major difference between your Nforce3 chipset and mine: Nforce3 uses AGP (which is now old enough to have all quirks known and worked around), while Nforce430 (mine) and more recent (baskin) use PCI-express. I suspect the problem is to be found in fglrx's internal, proprietary, graphics bus controller: PCI-E being still evolving, I betcha there's a little regression in it somewhere!

    One path worth investigating: how does fglrx deal with PCI-E 2.0 cards plugged in PCI-E 1.0 motherboards? Did it change in any way for Catalyst 9.9?

    Comment


    • Originally posted by mitch074 View Post
      @LostInSpacetime: your bug might be in fglrx, or it might be in Skype... Try using Skype without fglrx (use the FOSS ati/radeon/radeonhd driver) to see if you can reproduce the bug.
      You can also try disabling composite in fglrx (it doesn't deal well with it).

      @mirv: there is one major difference between your Nforce3 chipset and mine: Nforce3 uses AGP (which is now old enough to have all quirks known and worked around), while Nforce430 (mine) and more recent (baskin) use PCI-express. I suspect the problem is to be found in fglrx's internal, proprietary, graphics bus controller: PCI-E being still evolving, I betcha there's a little regression in it somewhere!

      One path worth investigating: how does fglrx deal with PCI-E 2.0 cards plugged in PCI-E 1.0 motherboards? Did it change in any way for Catalyst 9.9?
      My mistake in numbering of nforce - I do use a PCI-E card, a radeon3650 from sapphire to be exact.

      Comment


      • Ah. Well then, it's something else. What chipset are you using exactly?

        Comment


        • Originally posted by mitch074 View Post
          Ah. Well then, it's something else. What chipset are you using exactly?
          I'll post details later tonight if I remember after work.

          -- ok I use an asus m2n-e motherboard (nforce 570 or something like that). Apologies, I had confused my old computer previously, which used an nforce3 (this is what happens during crunch time at work, you mix up strange things). Kernel is 2.6.32-gentoo-r1, catalyst 9-11, AMD64 X2 3800+, radeon 3650. Obviously I use gentoo 64bit. xorg-server is 1.6.4. Let me know if you'd like more details.
          Last edited by mirv; 03 December 2009, 02:05 PM.

          Comment


          • Originally posted by mitch074 View Post
            @LostInSpacetime: your bug might be in fglrx, or it might be in Skype... Try using Skype without fglrx (use the FOSS ati/radeon/radeonhd driver) to see if you can reproduce the bug.
            You can also try disabling composite in fglrx (it doesn't deal well with it).
            I tried some things. The bug does only appears with fglrx and compiz(or metacity with compositing)+xv. If I force skype to use x11 instead, the video window works as expected. The problem with x11 is just that it's slow and doesn't scale (with skype).

            So it seems it's a fglrx bug affecting xv+compositing.. at least on my hardware (HD3200)

            Comment


            • Mine is a NVIDIA? nForce 750a SLI Chipset.

              Comment


              • Well, I use Mandriva 2009.1 (I also tried 2010.0) with kernels 2.6.29.6 and 2.6.31.5 (2009.1 backported 2010.0's kernel) with xorg-server 1.6.4/1.6.5 in 64-bit. It seems that kernel and Xorg versions have no real impact on the bug, however all our builds are 64-bit.
                We lack statistical evidence; from what we're getting here, 2 out of 4 Nvidia chipsets, of varying ages, got us a failure, the only AMD chipset succeeded with a kernel/server/card combination that failed on an Nvidia-based mobo. More reports would help (but we're not finding any).
                This could be a combination of things:
                - an Nvidia chipset
                - with a certain BIOS version
                - and certain cards
                may cause failures.

                I'd love finding an updated VBIOS for my card; the one it's using is shared by several early 4850 manufacturers; some updated their BIOS, but not all did; and I don't feel like experimenting with BIOSes I didn't modify myself

                Now though, I'd like to add some things:
                - my previous motherboard was made by Asus, a company that is tepid at best towards Linux (and at the time I bought the mobo, didn't acknowledge its existence), while my current one is from Gigabyte (first mobo maker to support LinuxBIOS officially on one of its products).
                - my previous mobo used a beta BIOS: I didn't feel like downgrading it to the latest 'official' BIOS, a previous experiment having me computerless for weeks when I had to have the BIOS chip re-flashed, and which was hardly supposed to help.
                - my previous mobo saw four years of heavy use; although no capacitors leaked and it passes all RAM and CPU tests, I did have overheating problems with this system at times, and it may not be considered rock stable anymore.

                Comment


                • "Ask ATI" dev thread

                  Think youre right that one ATI card will work but that the Nvidia chipset will not support Crossfire.

                  Comment


                  • The 9.12 driver brought notable improvements in card initialization routines; while I won't rebuild my previous rig to test, I wonder if the bug still exists.

                    Comment


                    • Where can I download the Linux version of atiflash?

                      Comment

                      Working...
                      X