Announcement

Collapse
No announcement yet.

Catalyst 10.2 is out

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

  • #81
    Originally posted by cutterjohn View Post
    There are instructions buried in this forum somewhere about how to build a driver that should not show that unsupported hardware annoyance. I kind of liked it myself as it gave some feedback as to when X & fglrx started up... but it was otherwise, annoying and unhelpful.

    Congratulations ATI. You finally figured out that you can support mobile chipsets, then you went and blew a gaping hole in your head by allowing your tardy OEMs to skip out on the support. Thanks Sony, et. al. and ATI.

    Double post, since phoronix appears to be editing brain dead now.
    I googled for the issue, and the only trick I found documented was to replace a /etc/ati/control file with an older one. I replaced it from 8.10, but I still get the watermark

    Comment


    • #82
      Originally posted by cutterjohn View Post
      Congratulations ATI. You finally figured out that you can support mobile chipsets, then you went and blew a gaping hole in your head by allowing your tardy OEMs to skip out on the support. Thanks Sony, et. al. and ATI.
      Apologies, I'm a little confused about that statement. The windows driver just picked up mobility support directly, so what do the OEMs have to do with anything?

      Comment


      • #83
        for me, the DKMS build part of installing the driver fails, I wonder what the consequences are:

        Code:
        Errors during DKMS module removal
        
        Creating symlink /var/lib/dkms/fglrx/8.702/source ->
                         /usr/src/fglrx-8.702
        
        DKMS: add Completed.
        
        Kernel preparation unnecessary for this kernel.  Skipping...
        
        Building module:
        cleaning build area....
        su nobody -c "cd /var/lib/dkms/fglrx/8.702/build; sh make.sh --nohints --uname_r=2.6.32.8-custom --norootcheck"....(bad exit status: 1)
        0
        0
        [Error] Kernel Module : Failed to build fglrx-8.702 with DKMS
        [Error] Kernel Module : Removing fglrx-8.702 from DKMS
        this is on 2.6.32.8, Radeon Mobility HD 5830

        Comment


        • #84
          Originally posted by mirv View Post
          Apologies, I'm a little confused about that statement. The windows driver just picked up mobility support directly, so what do the OEMs have to do with anything?
          Well, it goes like ATI IS going to do what nVidia has done for what seems like forever: attempt to make their drivers generic, HOWEVER they are ALLOWING their OEMs to OPT OUT and already, apparently, Sony, Toshiba and a bunch of others have done so effectively just about negating the entire point.

          There IS mobility modder.net, which I currently use, but it's broken as of 10.x releases until it's updated. True you CAN modify the base driver by hand, but this is just silly.

          Anyways they also added a notebook entry IIRC for card type, even under linux but it just takes you to a page of OEMs with links to their sites which I find to be hilarious considering that most of them have no linux/X drivers. There's a radeon mobility option which takes you to the good old catalyst install package though.

          ATI's driver distribution page is a mess IMNHO and should be cleaned up/simplified. IIRC with nVidia they just have a single package to download for every supported card, and then a legacy driver package for older GPUs.

          So, this is kind of linux on topic, but the main point where i'tll be annoying will be under Windows.

          I also found it to be interesting from the Tom's Hardware article how they apparently implement crossfire by switching off entire screen rendering between cards when I though for sure that they'd have a bit more balanced approach which it sounds like they're going to try doing now. (I wonder how nVidia does SLI... never thought about multi-GPUs until now since I'm still holding off on my new desktop until either Sandy Bridge or hexacore i7's(kind of want to wait and see what nVidia's new GPU is like too) so it might be budgetable by the time I start buying parts.)

          Looks like the next catalysts are going to be ALL about crossfire stuff and fixes/optimizations.

          @ako.maroy
          you should list your distro too...

          Comment


          • #85
            d'oh missed the EDIT deadline again! Why don't they just remove editing altogether?!

            Anyways, just to clarify their drivers are already really generic ones as I don't think that I've seen an OEM make any truly significant changes to a gfx card that would require driver changes. After all they all get the same base GPU parts so about the only other different thing would be for overclocking, but even then there's the builtin overdrive option plus other tools.

            Comment


            • #86
              @cutterjohn

              Ah, cheers. Still, can't blame ATI for allowing the OEMs to do what they want, even if it sucks.

              Comment


              • #87
                I did some testing by enabling compiz from the command line.
                Certain features of compiz are working with the new driver.
                Enhanced Zoom is working.
                Desktop Cube just crashes. I don't know much about the inner workings of compiz. I'm not that great with the retrieval of the build dependencies for compiz in debian type distributions. I'm used to grabbing the source off a file sever...

                All my games work just fine. Nexiuz, quake3 engines based games work fine.
                I still need to test unreal tournament 2004 and neverwinter nights.

                I'd say it's a problem with compiz.

                On a plus, it's nice that the drivers build with 2.6.32. My testing of debian amd64 has gone much nicer.

                Comment


                • #88
                  Originally posted by squirrl View Post
                  I did some testing by enabling compiz from the command line.
                  Certain features of compiz are working with the new driver.
                  Enhanced Zoom is working.
                  Desktop Cube just crashes. I don't know much about the inner workings of compiz. I'm not that great with the retrieval of the build dependencies for compiz in debian type distributions. I'm used to grabbing the source off a file sever...

                  All my games work just fine. Nexiuz, quake3 engines based games work fine.
                  I still need to test unreal tournament 2004 and neverwinter nights.

                  I'd say it's a problem with compiz.

                  On a plus, it's nice that the drivers build with 2.6.32. My testing of debian amd64 has gone much nicer.
                  What video card do you have?
                  Interestingly it doesn carsh Desktop Cube on my mobility hd5650. Also I found that there's one crash situation here, when maximizing windows at first, compiz crashes. But if i switch to Emerald window decorator, there's no crash - with that workaround i have a fully working compiz + gnome.
                  Quake4 and Neverwinter Nights runs here okay.

                  Comment


                  • #89
                    I bought a 4770 about 2 months ago now. This card is awesome, but the driver!. fglrx is generally a lot better/more stable than it was when I last tried to use ATI hardware, but it is still an extremely disappointing experience.

                    As before, fglrx doesn't seem to get along with kdm at all. This pair still loves to get together for computer-crashing parties. I eventually just uninstalled the login manager entirely and started using startx instead to avoid having to hold my breath every time I logged in or out.

                    VT-switching while X is running is another hold-your-breath-and-hope-for-the-best experience. Sometimes when logging out of KDE and being returned to VT1 where I started X, I find that all the text inputted into the console isn't displayed. It's there and I can run commands but I can't see what I'm typing. Fortunately the other VTs aren't affected and I don't generally need to use them anyway, not when fglrx will let me start X anyway.

                    I've been holding my old xorg packages preventing the upgrade to 7.4 for a few weeks now. Of course I made the mistake of letting them install the first time, only to be greeted by X not starting. If it weren't for the fact that I hadn't cleaned my apt-cache in months, I would have had to do a fresh install to keep using fglrx. And I happen to know that at least one DD had that same problem. He probably never imaginaed that ATI could still be lacking support after all this time. Today I found that Catalyst 10-2 had made into testing. I had a brief moment where I worried about installing it because I thought it wouldn't support my Xorg version before getting a clue and realizing I'd been holding the older xorg packages back for weeks. So I went ahead and installed it, and was subsequently impressed by Debian's new support for dkms. Unfortunately starting X just gives a black screen! So now I'm holding the older Xorg packages AND the 10-1 Catalyst packages! What fun it is to be an ATI customer!

                    amdpcsdb - I hate this thing. For starters, it lives in /etc so every upgrade overwrites it with defaults. Second, IT'S NOT XORG.CONF. Third, it somehow relies on the contents of xorg.conf anyway, why not just use xorg.conf, and a . file in ~ if you have to have more settings outside of xorg.conf? I guess it wouldn't be enough like the Windows driver or something...

                    Still can't run "X :1" without a huge crash. But I forgot, fglrx isn't a driver for X so it doesn't have to worry about actually supporting X's features.

                    Here's what isn't totally, confoundingly annoying for me wrt fglrx: If you don't go outside the Windows-usage paradigm (meaning one Xserver, no VT switching) 9-12 and 10-1 at least seem to be pretty stable, performance is reasonable, Doom3 ran nicely enough. TVout support isn't missing essential features like overscan and position controls (though the fact that there doesn't seem to be an xorg.conf setting for them is not a plus). The fan isn't doing the repeated start/stop thing it does on the console and when booting Windows, so I'm guessing the power management is working well, even though I don't have atieventsd installed.

                    If I stay in that Windows-usage paradigm where there's only the one desktop(read:Xserver) with my TV attached to the side of it, and I don't VT switch while X is running, Catalyst 10-1 works well enough for me to actually keep using my 4770! That said, I am desperate for a newer version of the OSS driver to show up in Squeeze and I want to stop using fglrx as soon as I'm able to. It's still so chock full of stupid crap to deal with that I'm just flabbergasted. Still no Xorg 7.4 support after all this time? Still can't actually use features X supports? Can't reliably VT-switch? Can't upgrade the driver without worrying if it will work at all?? I'm sure I managed to have Nvidia's blob to crash my box a couple times over the years, but fglrx is just ridiculous with the number of ways it can and does crash your computer. And that's saying something because basically NOTHING else crashes my Debian system. Sure stuff crashes now and then, but it NEVER crashes the entire system to the point of being Utterly Bored by Raising Elephants - fglrx does excels at that.

                    All that said, I'm still happier owning an ATI card which I'll eventually have open drivers for than I was with my Nvidia card and their driver. ATI, you do not make it easy though!!!

                    Comment


                    • #90
                      well, seems I was wrong compiz failed to start after a reboot.

                      Ubuntu 9.10 x86-64 + Catalyst 10.2
                      mobility radeon hd 4850 512MB GDDR3

                      (Seems to run OK if enabled after getting to the desktop but fails presumably and an X.org re-start.)

                      Comment

                      Working...
                      X