XPRESS users, please test!

I’ve just committed code to hopefully fix some of the remaining issues with RS4xx (XPRESS chips). If you have an XPRESS chip and have problems with dualhead, or display corruption, or DVI, please test the latest code in ati git master!

17 Responses to “XPRESS users, please test!”

  1. Andrius Says:

    I have just installed git version (6d5066a45…). It works much better than 6.7.195. Switching between tty and X now always work as expected.

  2. Andrius Says:

    By the way, Xorg.0.log says that RENDER Accel is unsupported but it seems that it works. Maybe some unneeded warnings have been left in the source.

  3. Linas Žvirblis Says:

    Just tested it on RS485 (a.k.a. Xpress 200M, or Xpress 1100 IGP as it is detected on my system). I have DRI forced on (yes, it somewhat works), composite, and AIGLX. No significant changes here. Will try to test TV-out and dual head later.

    I did notice that some OpenGL applications that used to work no longer do so, but I cannot claim that this is what broke them.

    As a not-exactly-happy Xpress owner I am more than willing to test things. Just say what needs to be tested and what information you need.

    And thank you for your work. It is greatly appreciated.

  4. agd5f Says:

    Please try dualhead and report back. TV will not work yet as XPRESS boards use a different reference clock and I haven’t sorted out the pll timing for that clock yet.

  5. Brice Goglin Says:

    I just pingued a dozen people who complained about Xpress support on the Debian BTS, you might get lots of feedback soon.

  6. mar_rud Says:

    I have integrated Ati card in my nx6125 laptop (RS480?):
    “ATI Technologies Inc Radeon XPRESS 200M 5955 (PCIE)”

    Last commit 9bea60b3eb378de5e1d44cc02a2763f4feae7882 introduced image corruption:
    black horizontal lines ~100pix long and some 1pix shift between vertical fragments of screen (each fragment has regular black lines and is moved up/down from previous one).
    When I reverted this commit no corruption.

  7. beso Says:

    i have a much important issue for now with a x200m 5975 chip:

    googleearth continues to hang up the system and wine apps don’t work well; for example i don’t get iexplore toolbar buttons or url bar. also stellarium hangs up and doesn’t start and i still have exa corruption issues after rightclick.
    the right click menu remains until i refresh the screen after i rightclicked somewhere on the screen.
    i’ve been using the git driver for some months (since december) and i’ve been always using all the new features that got out, but since then these issues weren’t corrected.
    i’ve tried to run revenge to provide feedback but it won’t run on this chipset. if there’s something that i could provide to help out i’d be glad to do it.

  8. Alex Villacís Lasso Says:

    I have a RC410 XPRESS 200M (1002:5a61), in which DRI is still not officially supported. Any news on when this might get implemented?

  9. agd5f Says:

    Please try it out with latest xf86-video-ati and drm from git. It was disabled by default since it hung on some cards, but things may be working better now. If it works ok, let me know.

  10. Linas Žvirblis Says:

    I did not have a chance to test dual-head yet, but I got latest xf86-video-ati and drm from git, and played with that a bit. The results:

    o Simple OpenGL application like glxgears work fine. And I even get up to 1200 FPS in glxgears, which I consider quite good.
    o More complex ones like amoeba OpenGL demo render many things incorrectly. It looks something like weird colorful spikes.
    o Applications that use freeglut like torcs and gl-117 crash with r300EmitArrays: Assertion failed error.
    o Stellarium just segfaults at startup.
    o Blender and Celestia give warnings, but seem to work.
    o Xcompmgr works fine with any options. It is fast and seems stable.
    o Both EXA and XAA are stable and fast.
    o With EXA forcing application to do indirect rendering with LIBGL_ALWAYS_INDIRECT=1 results in X server crash. With XAA it resulted in total system freeze, and I did not feel like reproducing this. I use EXA on daily basis.
    o Compiz crashes X.

    None of these are regressions. Well, at least not obvious ones.

    Here are the relevant parts of my xorg.conf that I use on daily basis:

    Section “ServerFlags”
    Option “AIGLX” “on”
    EndSection

    Section “Extensions”
    Option “Composite” “enable”
    EndSection

    Section “Device”
    Option “Accel” “true”
    Option “AccelMethod” “EXA”
    Option “RenderAccel” “true”
    Option “DRI” “true”
    EndSection

    So, 3D is not really usable, but I keep DRI enabled because it makes desktop rendering way more pleasant. I do not use any composite manager.

    And now the good news. I can now have several instances video playing trough XV. It used to fail saying that XV is busy, and now it just works!

    RS485 [Radeon Xpress 1100 IGP] - says so in lspci
    XPRESS 200M 5975 (PCIE) - in Xorg.log

  11. Alex Villacís Lasso Says:

    RC410 XPRESS 200M (1002:5a61) in Fedora Core 8

    I have tried using latest GIT for both the radeon driver, and libdrm, but with stock mesa 7.0.2-3.fc8. When enabling DRI in xorg.conf, the X server (when started with startx) loads as far as the GNOME wallpaper, then screen remains without update. Mouse pointer can be moved around, but keyboard is dead. Network is still responsive, and I could successfully ssh in in order to reboot the machine. I have Xorg.0.log and dmesg log for reference.

    Of course, this might be due to the fact that I did not use the latest mesa source, so I will try that next.

  12. Alex Villacís Lasso Says:

    RC410 XPRESS 200M (1002:5a61) in Fedora Core 8

    I was wrong - the X server does not even reach a point in which it paints the wallpaper. It simply switches into graphics mode and whatever was in the screen memory is displayed. With EXA, the X server then hangs shortly and shuts the monitor down. With XAA, the mouse pointer can be moved, but the keyboard is dead, as described before.

  13. agd5f Says:

    Does your system have sideport memory? Does it help if you disable it in the bios? Also, if you have an option to explicitly specify the amount of video ram to use in the bios try that rather than “auto” or whatever.

  14. Alex Villacís Lasso Says:

    RC410 XPRESS 200M (1002:5a61) in Fedora Core 8

    This system is a desktop computer, not a laptop. The video chipset is onboard, and uses shared memory only - no dedicated video memory. BIOS already sets shared memory use at 32Mb - there is no “Auto” setting. Does this help?

    BTW, I feel I am hogging this particular blog entry for my own personal case. The relevant bug was already filed at http://bugs.freedesktop.org/show_bug.cgi?id=12048.

  15. Assaf Paz Says:

    RS480 Xpress 200M 5955, 64 MB shared memory.

    Running glxgears will show ~1000 fps @ XAA and ~200 fps @ EXA.
    Thats ~300 more fps in XAA from earlier releases!

    Using Textured Video gives me a diagonal line on fast image changes - other than that all is AOK.

    ppracer shows surfaces that block most part of the game view every couple of seconds.

    supertux shows squares everywhere (all across the screen), but game is playable.

  16. givemesugar Says:

    i cannot run anything on s-video output for some unknown reason. when i plug a tv (lcd 1440×900) i canno have the s-video see it as connected. maybe i’m doing something wrong with the configuration.
    is there a site where to find some info on how to configure the external s-video monitors?

    thanks

  17. agd5f Says:

    S-video is not yet supported on rs4xx chips at the moment since they use a 14 Mhz reference clock and I haven’t worked out the clock setup for tv-out in that case.