Catalyst 8.6 + Cedega = graphics corruption

[Saist]

It seems that the latest update from AMD has some problems running with Cedega... Took pictures with my cellphone to kind of demonstrate the problem if it helps figure why the corruption is happening.


In most cases OpenGL support seems to be functional, with UT2k4, Warsow, Nexuiz, and Alien Arena all working perfectly...

However, attempting to start up a game through Cedega on the 8.6 driver results in well...


RadeonHD 3870
http://i20.photobucket.com/albums/b209/je-saist/hd3800-cedega-activecorruption-smal.jpg (http://i20.photobucket.com/albums/b209/je-saist/hd3800-cedega-activecorruption.jpg)


Using Control Alt Backspace, then killing WINE, gives me something like these

RadeonHD 3870
http://i20.photobucket.com/albums/b209/je-saist/hd3800-cedega-post-corruption-small.jpg (http://i20.photobucket.com/albums/b209/je-saist/hd3800-cedega-post-corruption.jpg)

Radeon 9600
http://i20.photobucket.com/albums/b209/je-saist/9600-cedega-post-corruption-small.jpg (http://i20.photobucket.com/albums/b209/je-saist/9600-cedega-post-corruption.jpg)


Anybody have any ideas what went screwy with the new driver?

[Dandel]

1 question, do you have compiz fusion disabled? if so try enabling it and looking again... this appears very similar to a bug i face with FGLRX 8.5 and 8.6 with it disabled.

[Saist]

1 question, do you have compiz fusion disabled? if so try enabling it and looking again... this appears very similar to a bug i face with FGLRX 8.5 and 8.6 with it disabled.

neither compiz-fusion nor beryl are installed.

[Dandel]

you could try installing that, although it'd be good to see the xorg config file to see if any changes are needed to that, because i don't have any of those issues... and as a quick note i don't see any of these problems... here's a quick spec of what i am using...

Ubuntu 8.04
FGLRX 8.6 ( 8.5 contained major bugs that made it unusable. )
wine 1.0
and of course my video card is a Radeon X1900GT.

[aniruddha]

I have exactly the same problem with the new drivers. It happens with Cedega, Wine and crossover games. I don't have compiz-fusion enabled.


*Adding x11-drivers/ati-drivers-8.501 to package.mask*

[hehe2]

Same problem,

unable to use wine or cedega with 8.6 ati drivers :x

Slackware 12.1 with wmaker, no compiz fusion ... tested with 2.6.24.7 and 2.6.25.7 recompiled kernels ...

8.5 ati drivers are working (using the workaround to get rid of the black borders) and so are 8.4 drivers.

That's "a bit" sad ... how comes such problems can happen on so many comps we got and ATI are still releasing their trash ??

I wonder if I'll be able to use my 3870X2 one day ... so sad :X

...

[Vighy]

Here on a amd64 machine doesn't happen!

fglrx 8.6, kernel tuxonice-sources-2.6.25-r4 (from gentoo tree)

could be just x86 related?

[dscharrer]

I get simmilar corruption with Wine on a X1600 XT on Gentoo x86.

Since I couldn't find a bug report, I created one: http://ati.cchtml.com/show_bug.cgi?id=1173

[bierbo24]

same results with gentoo amd64 and wine 1.0 here

[Dandel]

i stand corrected, did a few changes and ran a 2D app on wine and it happened to me... the fix is to restart your xorg server.

[aniruddha]

the fix is to restart your xorg server.
You mean each time you run a wine app???

[Dandel]

yes, for wine, however i was also able to get this to appear using smplayer ( a front end to mplayer ) with the video output set to...
gl:yuv=3

and this time, the fix for this was to close the program.

[Vighy]

yes, for wine, however i was also able to get this to appear using smplayer ( a front end to mplayer ) with the video output set to...
gl:yuv=3

and this time, the fix for this was to close the program.

Same here! i reproduced the bug playing a video with gmplayer on the external monitor of a dual head configuration!
Playing at full screen it came out. Closing it solved the problem!

edit: i think it's related to the new amdxmm... ysterday when i hacked in an ugly way the ebuild to install the new driver it didn't get loaded and the problem was not there! (if i'm not wrong! XD)

[sundown]

I could reprduce the 3rd picture when I start Nexuiz. To solve it I had to completely shutdown the internal monitor and keep only the external.

[hmmm]

i had something similar on a hd3870 when using crossover games to load steam with 8.6 - 8.5 n 8.4 were fine.

[ZedDB]

I have the same problem but with Blender. The screen becomes corrupted when i start Blender. But when i shut it down it goes away.

[dscharrer]

edit: i think it's related to the new amdxmm... ysterday when i hacked in an ugly way the ebuild to install the new driver it didn't get loaded and the problem was not there! (if i'm not wrong! XD)

I renamed my amdxmm.so and restarted X. The log files showed that the amdxmm module could not be loaded, however the problem remained (while native OpenGL apps still worked).

So far, everyone that posted in this screen seems to get the corruption, even if ther initially reported being fine:

- Saist: Cedega (fine with native OpenGL), no Compiz, Radeon HD 3870 / Radeon 9600
- aniruddha: Cedega/Wine/Crossover Games, no Compiz, Gentoo
- hehe2: Wine/Cedega, Slackware 12.1 (kernel 2.6.24.7/2.6.25.7), no Compiz, Radeon HD 3870X2
- bierbo24: Wine 1.0, Gentoo amd64
- Dandel: Wine 1.0 (2D app) / smplayer /w video outpu set to gl:yuv=3, Ubuntu 8.04, Radeon X1900GT
- Vighy: gmplayer on external monitor, Gentoo amd64 (kernel tuxonice-sources-2.6.25-r4) (corruption solved by closing)
- sundown: Nexuiz (solved by shutting down internal monitor)
- hmmm: Crossover Games loading Steam, Radeon HD 3870
- ZedDB: Blender (corruption solved by closing)

- Me: Wine 1.0, Gentoo x86 (kernel gentoo-sources 2.6.24-r8), no Compiz, xorg 7.3 / server 1.4.2, single monitor (configured and connected), Radeon X1600 XT
solved only by restarting X; closing the program and changing resolution with xrandr did not remove corruption

Hmm, I don't see any patterns except, that only people that get the screen corruption with native OpenGL apps reported that closing the programm solved the problem and closing the program (plus 'wineserver -k') did NOT solve the screen corruption for me.

About time for someone from AMD to take notice of this.

edit: I played around a bit with the 8.5 and 8.6 drivers: Interestingly, when loading the system and X with the Catalyst 8.6 driver and the downgrading to 8.5, but neither reloading the kernel module nor restarting X, I don't get any corruption anymore. It even works with the 8.6 configuration files ( etc/ati/amdpcsdb.default, /etc/ati/atiogl.xml, /etc/ati/control and /etc/ati/signature) in etc. Hope this helps.

[brot]

Hi everyone, i looked here if anyone has the same problem that i have, and it looks like i am not the only one. i am back to 8.5 now.

I bought that nice radeon 3850, as i really like what amd does with their linux support and opensource etc. so if anyone at ati/amd is reading this: you are doing the right thing, i was a hardcore nvidia user, but since you started your linux support thingy i wanted to switch.

back to topic ;)

the 8.6 driver has the same problem as the first post shows in the screenshots. But i discovered that this only happens if its an opengl window. so if i start etqw, everything works and is displayed right. As soon as i switch to windowed mode i get the corrupted screen.
Wine gets always started in a window here so i have the same problem here.

The issues are only with the new 8.6 drivers, so somewhere there is a regression in between those releases. Someone knows if amd/ati knows about this bug, or where to report it to them ?

If screenshots could help i could make some btw.

System is a Opteron 170, dfi nf4-sli-d, gentoo ~amd64, gcc 4.3.1, kernel 2.6.25.7.

Thanks for having such a great forum btw, and have a nice day everyone :D

[ZedDB]

the 8.6 driver has the same problem as the first post shows in the screenshots. But i discovered that this only happens if its an opengl window.

THANK YOU for pointing this out!
Blender runs without corruption in fullscreen mode. (blender -W)

[wfeltmate]

This is the first time I have ever had an issue with any other their driver releases. Sure, I've had stuff not work before and had to wait for them to fix it, but this is the first time that it is an issue causing me to downgrade to a previous release.

I am also getting the issue, primarily when trying to run wine and occasionally while using mplayer. However, I don't always get ti while running wine. I can open up an use Guitar Pro 5.x with no issues whatsoever, yet when I try to load GTA - San Andreas I get the corruption immediately.

Perhaps it has something to do with the recent changes they made to the Textured Video and related parts of the driver code?

[dscharrer]

THANK YOU for pointing this out!
Blender runs without corruption in fullscreen mode. (blender -W)

Warsow, Sauerbraten, PlanetPenguin Racer, Super Tux, Neverball all work fine for me both in windowed and in fullscreen mode. Howeverm, when in fullscreen mode and then switching to windowed mode, the game resizes to what the window would be, but stays in the top-left corner and doesn't get any window decoration (borders, title bar). The rest of the screen somehow mirrors the game, but messed up like the screens in this thread (and like games I run in wine). However, when starting the game directly into windowed mode everything works fine.

Somehow I only hit this bug when using wine.


Someone knows if amd/ati knows about this bug, or where to report it to them ?
I filed a bug at the unofficial bugzilla at http://ati.cchtml.com/show_bug.cgi?id=1173 and reported a ticket at amd.com: http://support.ati.com/ics/support/default.asp?deptID=894&task=ticket , but no response yet. :(
But after all, it has been weekend.

[Ower]

im using ceega to run wow. and i have the same corruption.. but i just switch to console and back to X and corruption disappears

[Wise Ferret]

I have the same problem with 8-6 driver (splitted, garbled screen) trying to run Lord of the Rings Online over Wine on Ubuntu Hardy 32-bit, mobility radeon x1600. Switching to console and back does not help - I have to restart X.

[kiwi_kid_aka_bod]

For those of you experiencing screen corruption/splitting:

Check what resolution the corrupt application is launching. I had this issue with mythtv (fullscreen or windowed) on my X1400 driven 1680x1050 laptop screen. By forcing mythtv to 1680x1024 or 1680x1080 the corruption cleared up! This is again some wacky thing about the 1050 lines resolution.

And I registered to bring you this... ain't I nice ;-)

[Wise Ferret]

For those of you experiencing screen corruption/splitting:

Check what resolution the corrupt application is launching. I had this issue with mythtv (fullscreen or windowed) on my X1400 driven 1680x1050 laptop screen. By forcing mythtv to 1680x1024 or 1680x1080 the corruption cleared up! This is again some wacky thing about the 1050 lines resolution.

And I registered to bring you this... ain't I nice ;-)

You are very nice indeed! Google earth works fine for me, as it runs in 1680x1050. However, most of my OpenGL apps cannot utilise such a high resolution, so I'm forced to live with the corruption.

Another nasty issue I experience with my x1600 on 32-bit hardy: on some sessions, using Xv video crashes my X or the hosting app. This happens in about 30%-40% of the sessions. Restarting X does not solve the issue, and I have to reboot the machine hoping for a "good session" this time. Any ideas? How can I help debugging this?

[kiwi_kid_aka_bod]

You are very nice indeed! Google earth works fine for me, as it runs in 1680x1050. However, most of my OpenGL apps cannot utilise such a high resolution, so I'm forced to live with the corruption.

Another nasty issue I experience with my x1600 on 32-bit hardy: on some sessions, using Xv video crashes my X or the hosting app. This happens in about 30%-40% of the sessions. Restarting X does not solve the issue, and I have to reboot the machine hoping for a "good session" this time. Any ideas? How can I help debugging this?

Yeah, googleearth (henceforth known as "The Great Time Waster") worked for me too (windowed and fullscreen). MythTV was the first time I experienced this. Some more odd data points:

1680x1049 and 1680x1051 work. Only 1680x1050 seems to break things. Unfortunately the top and bottom gnome panels are still visible if the res isn't 1680x1050.
The breakage is with the MythTV Qt renderer. If I change the renderer to OpenGL, I get corruption while it scales the theme (I presume this is using Qt) then when the OpenGL renderer starts the corruption dissappears. Unfortunately there's still some issue with OpenGL renderer active that makes MythTV unusable. I haven't yet sussed out if it's MythTV or fglrx at fault.

I'm on 32-bit hardy as well. No real help for you, but I am seeing much more instability than I was used to with fiesty.

[Thetargos]

I have also been hit by this bug on my Radeon X1200 in my laptop. Most prominently this bug happens with Wine applications, either accelerated windows, or 2D DirectDraw windows. However, I have seen this problem with other native programs as well, for example OpenGL applications running at non-native (for the laptop's display) resolutions, and on this lower-end card, that means pretty much any accelerated application (as the native resolution for the display is 1280x800).

From my own experience and what I've seen on this thread it would appear that:

Wine is the main trigger for this corruption.
Some native applications may be affected as well, and it seems to be "resolution dependent".


What I have not seen mentioned here, is that most native GLX applications seem to crash (segmentation fault) at exit (to test this, just see the output of glxinfo, in my case, the last message is a segmentation fault notice, also registered in dmesg).

[dscharrer]

Ok, here's some more data in case any of you dev's actually read this:

My monitor's native and my desktop resolution is 1280x1024.

- Wine
Corruption always appaears, even some game launchers (Track Mania Nations Forvever) / game setup utilities (SW: Knights of the Old Republic) that are seperate executables and don't touch the current screen resulotion. (At least there is no flickering or anything when using 8.5)

Closing the app and killing all wine processes or changing resultion with xrandr has no effect, only restarting X removes the corruption.


- Sauerbraten
In fullscreen mode, only resolutions between 1024x768 and 1280x1024 (both exclusive), namely 1152x768 1200x960 and 1280x800, cause corruption. Changing to resolutions above 1280x1024 has no effect.

When switching to windowed mode, the window gets resized and it's contens are drawn fine, but receives no window decorations and the remainder of the screen (the part not covered by the "window") mirrors the parts of the window content, but corrupted.

When in windowed mode, changing resolution (to any resolution, even 1280x1024 and 1024x768) shows the window decoration, but also corrupts the screen with the pattern described in this thread. Since Sauerbraten refuses to remember that it was in windowed mode, I didn't test how it behaves when started directly into windowed mode.

As reported for other native apps, closing Sauerbraten (cleanly or by killing the process) removes the corruption.


- Nexuiz
I haven't found any resolution causing permanent corruption, even changing to 1280x800 (the only one i can select that fails with Sauerbraten) succeeds.

However, when starting nexuiz or changing resolutions I sometimes get the split screen / checkerboard corruption, but only at my monitors native resolution or above. This corruption is preceded by the nexuiz window flashing a few times and the desktop becoming visible and only stays for a few seconds at most.

- Warsow
No corruption at all experienced in both fullscreen and windowed modes.

[GenTiradentes]

I have the same problem. An ATI Radeon HD 3870 with fglrx drivers 8.6, Wine 1.1.0, and direct rendering works great. Native linux games (mainly Urban Terror) work perfectly, better than any other linux games ever have, it's just as fast and responsive as in Windows. When I have Compiz enabled, that works great as well. When I run Steam in Wine, no corrution, Steam displays fine. When I start Counter-Strike: Source, the corrution appears halfway through loading, before the game display even comes up. The only way to fix it, as any users have already said, is to restart the X server.

I tried editing my xorg configuration file to see if it would help. I disabled composite rendering, then I replaced the glx module with dri, restarted the X server, tried again, and the corruption is still there.

As a side note, with any drivers before catalyst 8.6, (7.12 through 8.5) CSS and all other wine games crash. I was really hoping 8.6 would work with wine, as CSS crashes on startup in Vista. :D Funny how at times, wine works better than the operating system it's trying to emulate.

[Sekra]

Same problem with ATI Mobility Radeon X1400, catalyst 8.6, tried with wine 0.9.61 and 1.0.. Running a up-to-date Arch linux system with everything from the official repos.. Screen goes checkered and had to take a photo with my cam to get the look since screencap looks "fine".. Here are my photos:

http://www.saunalahti.fi/hotti/pics/pic1.png

and

http://www.saunalahti.fi/hotti/pics/pic2.jpg

wine worked just fine last week with catalyst 8.5....

edit: also using 1680x1050 screen resolution

edit2: tried with 1280x1024 resolution and same screen distortion happens..

[maligor]

HD 3870 here aswell, running with 1680x1050 resolution and seeing the same issue. 1280x1024 doesn't show the same behaviour but other types of artifacts instead, where GNOME Panels turn black.

It's interesting to note also that when running Blender in fullscreen, switching workspaces stops the screen updates. This same behaviour is in Savage2 if you don't have Grab Input enabled (so you can switch workspaces with hotkeys).

Savage 2 also has a minor issue that the textures or shaders in the login screen with 8.6 are totally bugged now. Ingame it works fine (Aside from the workspace switching).

[bridgman]

maligor, is the issue with Savege 2 login screen a regression from previous driver versions ? We're always looking for "easy to reproduce" problems where there's a good chance we might see them on our systems and not have to drive to your house :D

[storma]

maligor, is the issue with Savege 2 login screen a regression from previous driver versions ? We're always looking for "easy to reproduce" problems where there's a good chance we might see them on our systems and not have to drive to your house :D

No. Prior to the FCP2 savage update, we had no models at all. They seem to have fixed most of that problem, with only a few bits and pieces left.

[sid350]

I have same problems RadeonX200M(RC410) and Wine 1.0.0 x86. Ati still crappy :(

[ZedDB]

maligor, is the issue with Savege 2 login screen a regression from previous driver versions ? We're always looking for "easy to reproduce" problems where there's a good chance we might see them on our systems and not have to drive to your house :D
For me on my X1900XT the 8.5 did not have the shader problems with FCP2 (everything was displayed as it should). So the 8.6 partly broke what they fixed, but now we can at least see the models :D Besides that the ingame updater doesn't crash my computer anymore with Savage2 and 8.6... (Other than that "Penumbra:Overture/Black plague" now has working shadows with 8.6. With 8.5 the game crashed with them turned on)

[maligor]

maligor, is the issue with Savege 2 login screen a regression from previous driver versions ? We're always looking for "easy to reproduce" problems where there's a good chance we might see them on our systems and not have to drive to your house :D

Yes, it is.

Savage 2 64bit version 1.5.0 on linux and Catalyst 8.5 works to all appearances, flawlessly.

The issue only shows up on the login screen when using Catalyst 8.6. It looks like it looses all diffuse shading. In game I haven't seen any change.

This thread (http://forums.s2games.com/showthread.php?t=11653) seems to have other people mentioning the same issue. (The texture issue mentioned, not the crash)

My system is on debian unstable amd64 and intel core2.

[Vighy]

Yes, it is.

Savage 2 64bit version 1.5.0 on linux and Catalyst 8.5 works to all appearances, flawlessly.

The issue only shows up on the login screen when using Catalyst 8.6. It looks like it looses all diffuse shading. In game I haven't seen any change.

This thread (http://forums.s2games.com/showthread.php?t=11653) seems to have other people mentioning the same issue. (The texture issue mentioned, not the crash)

My system is on debian unstable amd64 and intel core2.

Same here: Savage2 64bit version 1.5.0 on Catalyst 8.6 has that strange issue in the login screen.

[duffster]

I'm seeing the same sort of corruption when I use MythTV.

It worked up to 8.5 (barring some video problems when changing channels, but that's a different issue), but now I get the weird corruption when I start MythTV - even if I start it in its own window, the whole display is corrupted. This happens on the menu screens as well as during TV playback.

Strangely, when I quit out of MythTV, the corruption remains for a second or 2 after MythTV has exited, and then the screen goes back to normal again.

This never happened before 8.6 and makes it unusable for me.

btw I'm using Debian/testing and have a 3450 with monitor plugged into DSub connector.

[LordRaiden]

http://www.phoronix.com/forums/showpost.php?p=36756&postcount=3


I haven't been able to go anywhere since.

Summary:
It occurs with when running Warcraft II using OpenGL or WoW using WINE's implementation of Direct3D (OpenGL doesn't work for me here). Catalyst 8.5 is fine, 8.6 gives me a dual checker-boxing, which oddly does not appear at all when I do a Print Screen.

If you need logs, as away (just tell me what command you need run, so that we're on the same page).

Thanks in advance

[Forge]

maligor, is the issue with Savege 2 login screen a regression from previous driver versions ? We're always looking for "easy to reproduce" problems where there's a good chance we might see them on our systems and not have to drive to your house :D

Well, the Catalyst 8.6 + any Wine-product certainly qualifies. I have not yet had my 4850 plus Wine manage anything but the checkerboard of doom.

[brot]

Well, the Catalyst 8.6 + any Wine-product certainly qualifies. I have not yet had my 4850 plus Wine manage anything but the checkerboard of doom.

Same here, wine seems to be the way to go if one wants reproduce the bug. By the way "checkerboard of doom" is a nice new name, i vote that its the official name for this from now on ;)

[Thetargos]

i vote that its the official name for this from now on ;)
You're on...

However, the problem would be to actually "correct" the problem rather than giving it a "cool" name ;)

[vjeko]

Once again. Same problem here. Could it be the case that this is an issue with hsync? Data is being rendered correctly, but the signal that is being fed to the monitor is corrupted, indicating synchronization problems.

So I suppose that the current solution is to wait for 8.7?

[brot]

maligor, is the issue with Savege 2 login screen a regression from previous driver versions ? We're always looking for "easy to reproduce" problems where there's a good chance we might see them on our systems and not have to drive to your house :D

Well, if you want to see the screen corruption i could take my pc and drive to you :D . According to googlemaps it would be less than an hour to Starnberg, which is what the fglrx driver says:

fglrx: module license 'Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY' taints kernel.

So if this would help in any way, just answer here and i will take my pc and take a ride. ;)

[aelschuring]

Just got bitten by the same issue, on AMD64 (Ubuntu 8.04). Just bought a new HD4850, but if I'd known the card was unusable with MythTV I'd not have bought it just yet...

Anyways, I agree with what one of the previous posters said, it appears to be an issue with horizontal sync / pixelclock. I have tried MythTV on every available resolution that my TV supports, and the checkerboard pattern varied in size accordingly:

1360x768 (max for my screen): about 10 chunks, blacks moving from upper right to lower left
1280x768: same 10 chunks, but moving from upper left to lower right
.
.
800x600: only about 6 chunks, moving right to left again

I have tried with some Wine games as well, and found that corruption caused by Wine games is non-recoverable. With MythTV, all I had to do was quit the app, and the screen would restore itself to normal. But with Wine (even when run in windowed mode, so no resolution changes), I had to do a ctrl-alt-backspace to get my normal screen back.

mplayer works fine for me, with both x11 and xv output. I also tried different settings for GLOverlay and VideoOverlay, but that changed nothing. Obivously, I can't test with 8.5 drivers because they don't support my card...

[storma]

Just got bitten by the same issue, on AMD64 (Ubuntu 8.04). Just bought a new HD4850, but if I'd known the card was unusable with MythTV I'd not have bought it just yet...

<--snip-->

As pointed out by someone in another thread, with Mythtv, if you use a non standard resolution, the problem is solved.

Start Mythtv with something like 'mythfrontend --geometry 1680x1049'

[aelschuring]

Cool. Must have missed that one. Indeed, telling mythtv to use 1359x768+1+0 fixes the corruption, but also makes it appear below the taskbar. But at least it's usable. Makes me curious about what the real reason for this bug is...

Curious thing: running mythfrontend in a window (using mythfrontend.real -w) causes the same corruption unless you provide the geometry workaround. So it's not (just) a full-screen issue.

Anyone else noticed screen flicker before the corruption sets in? It's almost like fglrx tries to switch to a new resolution, then decides that it can't or won't, and corrupts the screen in the process...

[storma]

Cool. Must have missed that one. Indeed, telling mythtv to use 1359x768+1+0 fixes the corruption, but also makes it appear below the taskbar.

<--snip-->

Check to see if your taskbar has an option 'keep below'. I use that setting with kiba-dock to get around that issue.

[hmmm]

New instance of checkerboard of death that might help spot where the regression is :S

(still on 8.04 i386 with cats 8.6 & hd3870)

Mirc 6.32 under wine 1.1.0 runs perfectly fine
Upgrade to wine 1.1.1 - checkerboard :S

I think the new Richedit or gdi+ fixes enable certain functionality that trigger the checkerboard.

BTW anyone notice super sluggish performance in Aisleriot Solitaire when maximised?

Hope it helps.

[Thetargos]

In general Catalyst 8.6 seems much slower on my mobile X1200 (RX690M) based ATi graphics chip. 8.5 was a tad faster (not by much, though), and feels especially sluggish in 2D (and even AIGLX, like some effects with Compiz Fusion [rotate-cube, close-window-effect [fire], etc])

[hmmm]

8.7 fixes the mirc issue - i dont know what you did ati but thankyou :)

Both mirc 6.33 and 6.32 are fine under wine.

checkerboard still popping up after opening system monitor :S

crossover games still checkerboarding with steam

aisleriot solitaire still sluggish...

[dscharrer]

checkerboard still popping up after opening system monitor :S

crossover games still checkerboarding with steam

What distribution / X version are you using?

For me (on Gentoo amd64) settting the ~amd64 keyword for all X packages (not just the server, also all the different libraries and protocol headers) and removing and then re-emerging all those packages removed the corruption for me with Catalyst 8.7 on a Radeon HD 4850.

[hmmm]

What distribution / X version are you using?

For me (on Gentoo amd64) settting the ~amd64 keyword for all X packages (not just the server, also all the different libraries and protocol headers) and removing and then re-emerging all those packages removed the corruption for me with Catalyst 8.7 on a Radeon HD 4850.

hardy (8.04.1) & 7.3 - with crossover games 7.1 steam still corupts :S

i wasn't expecting anything special from a rebase to wine 1.0 though - at least i can focus on uni work until cats 8.8 comes round.

Thankfully wine 1.1.2 hasn't broken mirc 6.33 either - i'm a bit worried as the gdi+ n richedit stubs get replaced it'll kill it again...

[maligor]

It seems atleast it is possible start blender without causing the corruption:

1) Switch to a smaller view (Ctrl-Alt-+, 1280x1024 here)
2) Start blender
3) Switch back to 1680x1050
* You'll notice now that the opengl view isn't fully scaled to the window
4) Resize the blender window to fit your needs.

[rob2687]

Switching to a lower resolution worked for me.
I can start Blender with corruption. Also Foobar on Wine works fine too.

[dscharrer]

It seems that at least for me (on Gentoo amd64 multilib, Radeon HD 4850) substituting the xorg-x11 libGL.so for the ATI one while using the fglrx driver is a quick&dirty fix for this.

Starting Kaffeine (opengl output) with
'LD_PRELOAD=/usr/lib64/opengl/ati/lib/libGL.so kaffeine'
and then switching to fullscreen and back corrupts the screen, while when starting with
'LD_PRELOAD=/usr/lib64/opengl/xorg-x11/lib/libGL.so kaffeine'
there is no corruption.

'eselect opengl set xorg-x11' has the same effect but system wide and permanent.

glxinfo reports the same info for both the ati and xorg-x11 libGL.so (direct rendering, OpenGL version 2.1.7769 Release) except for the GLX_MESA_copy_sub_buffer extension, which is only reported for the xorg-x11 version.

There is also no corruption in wine and native OpenGL games seem to run stable (at 1920x1200).

Can anyone else confirm this?

[Forge]

It seems that at least for me (on Gentoo amd64 multilib, Radeon HD 4850) substituting the xorg-x11 libGL.so for the ATI one while using the fglrx driver is a quick&dirty fix for this.

*snip*

glxinfo reports the same info for both the ati and xorg-x11 libGL.so (direct rendering, OpenGL version 2.1.7769 Release) except for the GLX_MESA_copy_sub_buffer extension, which is only reported for the xorg-x11 version.

There is also no corruption in wine and native OpenGL games seem to run stable (at 1920x1200).

Can anyone else confirm this?

WTF? Your post both excites and confuses me. I must not be understanding something somewhere, but I was certain that using the xorg-x11 GL libs put you solidly into software Mesa land. You're still getting functional hardware acceleration??

I'm on ~amd64 myself, I'll be trying this later tonight. If it works the way you describe, I'll name my nextborn child after you.

[dscharrer]

WTF? Your post both excites and confuses me. I must not be understanding something somewhere, but I was certain that using the xorg-x11 GL libs put you solidly into software Mesa land. You're still getting functional hardware acceleration??
Unless glxinfo is lying AND software acceleration can get sauerbraten with all settings to highest and 1920x1200 pixels over / close to the 200fps cap, yes, I'm getting full HW acceleration.

I'm on ~amd64 myself, I'll be trying this later tonight. If it works the way you describe, I'll name my nextborn child after you.
Wish you luck. ;)

[bridgman]

AFAIK libgl.so can connect to either a direct rendering driver (via Direct Rendering Interface aka DRI) or can encode the GL commands through the X protocol to run the commands indirectly on the X server. The X server, in turn, can load an accelerated driver, giving you Accelerated Indirect GL X aka AIGLX.

The LIBGL_ALWAYS_INDIRECT option forces libgl.so to chose the indirect path. Unfortunately almost all of the documentation refers to this as the "indirect, unaccelerated" path but that is of course no longer the case, since AIGLX is all about having the X server connect to the accelerated HW driver rather than good 'ol Mesa.

If the accelerated HW driver fails to open, either because it is not there or because the kernel module (drm) did not load properly, then presumably both direct (libGL => 3d driver) and indirect (libGL => X server => 3d driver) could fall back to Mesa but I don't know the details of that fallback (yet).

The libgl included with fglrx usually (but not always) behaves the same as the X libgl (remember that the xorg libgl supports direct and indirect accelerated 3d too) so it's certainly possible that this could work with acceleration and if it addresses the Wine issues is a useful clue.

[Forge]

Dscharrer - You are my hero.

Last night I put the 4850 in, unmerged all my NV crap, merged fglrx, and then reset the GL libs back to the stock xorg ones.

Poof, everything just magically works.

Left the system remerging world to re-establish coherency last night, should be ready for real testing tonight.

Non-sequiteur: It's really surprising just how many packages in Gentoo link in dependency against fglrx/nvidia. I removed the Nvidia drivers and hand-remerged half a dozen others, and STILL things were trying to pull in the nvidia drivers.


Bridgman : No offense intended to the fglrx devs, but using xorg's libs instead of ATI's appeals to me, plus it has solved a *lot* of little issues already. Might want to make a note of this usage of AIGLX+Xorg libs as a possible troubleshooting step in the future.

[maligor]

This seems to work yes.

Removing the fglrx-glx/fglrx-glx-ia32 packages in debian worked (had to reinstall libgl1-mesa-glx afterwards, didn't restore redirections)

LIBGL_DEBUG=verbose reports that it's using the fglrx_dri.so driver also so acceleration should be working correctly.

This also fixes the switching workspaces while fullscreen OpenGL apps running issue curiously.

(edit): Requires to make a manual copy of the fglrx_dri.so if you remove fglrx-glx-ia32 for 32bit apps.

[Inkaine]

No, doesn't do the trick on ArchLinux. I uninstalled fglrx and installed the mesa-libgl, copied that to a safe location and reinstalled catalyst. Then tried to use the mesa-libGL.so both with LD_PRELOAD as well as by copying over the catalyst libGL. Still corruption.

Switching display mode by Ctrl-Alt-+, running wine, then (still within wine) Ctrl+Alt+- works. No corruption then and after playing back to the original desktop. Weird.

[Guilo]

Same problem with opensuse 11.0 32 bits, catalyst 3.7 and radeon hd3850

[Extreme Coder]

dscharrer - THANK YOU!!!!!!!!
Your solution fixed everything for me here! :)
When updating to 8.6 from Mandriva repos (by accident), Cedega couldn't run Warcraft III (screen corruption), Urban Terror switching to window mode would cause corruption, and so on. Copying the Mesa libGL.so to fglrx folder fixed everything!

[Extreme Coder]

Ok, I tried this again, I downloaded ATI's installer, installed 8.7, before copying the file, everything is as usual with the corruption. After I copy /usr/lib/mesa/libGL.so.1 to /usr/lib , it worked!

[Thetargos]

My only question in this regards would be about the performance. I'm sure that ATi ships their own libGL.so due to some optimizations or similar. What about the performance with the Mesa libGL.so instead of the fglrx shipped one.

[Extreme Coder]

My only question in this regards would be about the performance. I'm sure that ATi ships their own libGL.so due to some optimizations or similar. What about the performance with the Mesa libGL.so instead of the fglrx shipped one.
Well, it looks the same to me(same fast performance) although I noticed the slight graphical glitches in Cedega with Warcraft III have gone.
And my glxgears have rosen from about 3300~ FPS to 4500~ FPS (I have an ATI x1650 Pro)
Thanks again to dscharrer, now I can feel easy about upgrading to the newest driver release, with a hotfix in hand in case something goes wrong :P

[Thetargos]

I will have to try this work around...

@Bridgman:
What may be the impacts of using Mesa libGL instead of the shipped fglrx libGL with fglrx? Should there be any features missing or performance issues?

[Thetargos]

Ok. I have tried this work around for the corruption with Wine, and since I run a 64-bit system, I did that for both /usr/lib and /usr/lib64. While older games like Starcraft run now much better and if used some of the tricks to speed it up (for instance, force OpenGL rendering for DirectDraw) it is orders of magnitude faster than previously using libGL.so.1.2 (through the libGL.so.1 symlink). However, I lost the ability to run Compiz, even though glxinfo does report that GLX_EXT_texture_from_pixmap is present.

[Extreme Coder]

Ok. I have tried this work around for the corruption with Wine, and since I run a 64-bit system, I did that for both /usr/lib and /usr/lib64. While older games like Starcraft run now much better and if used some of the tricks to speed it up (for instance, force OpenGL rendering for DirectDraw) it is orders of magnitude faster than previously using libGL.so.1.2 (through the libGL.so.1 symlink). However, I lost the ability to run Compiz, even though glxinfo does report that GLX_EXT_texture_from_pixmap is present.
Hmm.. I didn't try Compiz, I will give it a try today, if I'm free.
For me, gaming is much more important than desktop effects ;)

EDIT: I tried it out now, here is what it gives as an error:
[ahmad@localhost ~]$ compiz --replace
compiz (core) - Fatal: No GLXFBConfig for default depth, this isn't going to work.
compiz (core) - Error: Failed to manage screen: 0
compiz (core) - Fatal: No manageable screens found on display :0.0

[Thetargos]

Hmm... Is there an equivalent for ATi drivers to force a mode? Something like the environment variables for nVidia?

[hoho]

I hope this is fixed in the 8.8 release.
I get the same corruption as everybody on Archlinux, Radeon 9800 Pro (both 8.6 and 8.7, 8.5 worked fine).
And with fullscreen opengl games, if you switch workspaces the game doesnt go away, even keeps updating, but it doesnt respond to mouse or keyboard presses until you switch back to the same workspace where it was.

[Kano]

Do you really think they fix something? You must be lucky when they don't add a new error!

[LordRaiden]

http://bbs.archlinux.org/viewtopic.php?pid=395056#p395056

http://www.thinkwiki.org/wiki/Problems_with_fglrx#Corrupted_3D_display

From the second link

2/ As suggested by ATI support, edit the /etc/X11/xorg.conf and find the section "Display". Add the following line into the "Display" section:

Virtual <width> <height>

where <width> is the width of your screen in pixels rounded up to the next multiple of 64 and <height> is the height of your screen in pixels. For example, if your native resolution is 1400x1050, use

Virtual 1408 1050

After starting the X server you can run xrandr -s 0 to restore the X server to a native display resolution, and 3D rendering will still work.

I also use the start up script for playing WoW in it's own X server http://gentoo-wiki.com/HOWTO_Install_and_update_World_Of_Warcraft_with_wi ne#Startup_Script. Specifically for me, since I have a 1024x768 screen, I used Virtual 1048 768. Once I'm in an X session, I run xrandr -s 0 then execute the script.

It's working great so far on catalyst 8.7.

[Extreme Coder]

That sounds great, but using xrandr immediately messes up things again, just like the post from the Arch Forums..
ATI better not have this problem in 8.8..

[LordRaiden]

Just to summarize, my steps were:

1) install catalyst 8.7
2) change xorg.conf to have a "Virtual X Y" line in the Display section
3) Start up X
4) run "xrandr -s 0"
5) run WoW ( I do it in a separate script which spawns a separate X session, you might need to do this too. )

Notes
- Right now, I am not sure if I deleted /etc/ati/amdpcsdb after I installed catalyst or not.
- I use d3d instead of opengl as the engine

I'll also try invoking WoW directly, without creating a separate X session.

Good Luck

[Lukyn]

Hi all,

I am really sad about this screen corruption bug. It is all ok in the 8.5, but these drivers are much slower for my 9700pro.

However I installed archlinux today (first time) with ATI drivers and with glxgears I've got this:

libGL error: failed to open DRM: Operation not permitted
libGL error: reverting to (slow) indirect rendering

FPS was something about 3100/s so I tried wine game and it works! Well...frame rate was lower, but it works!

Then I fixed permission in xorg.conf and glxgears was fine with cca 5600/s so I run wine and again graphic corruption :-(

However, maybe there can be any solution..? I have no idea what this actually do before and after permissions, bud maybe someone..?

[hmmm]

8.8:

crossover games (tf2 still checkerboarding)
system monitor - still checkerboards but usable now (i.e. can now see system monitor window in front of checkerboards) - i noticed under the resources tab all 3 line graphs experience graphics corruption while updating)

[hoho]

Wow, just wow.

[Extreme Coder]

Just to summarize, my steps were:

1) install catalyst 8.7
2) change xorg.conf to have a "Virtual X Y" line in the Display section
3) Start up X
4) run "xrandr -s 0"
5) run WoW ( I do it in a separate script which spawns a separate X session, you might need to do this too. )

Notes
- Right now, I am not sure if I deleted /etc/ati/amdpcsdb after I installed catalyst or not.
- I use d3d instead of opengl as the engine

I'll also try invoking WoW directly, without creating a separate X session.

Good Luck
Maybe when you run it in a separate X server, the problem doesn't happen there, try running it in the same X server.
Also, why use D3D? I thought Wine worked best for WoW with OpenGL. But I use Cedega anyways, so I use D3D as default (since OpenGL can't run Specular Lighting and some other effects)
but with both Wine and Cedega, in both D3D and OpenGL, it's nowhere near Windows XP performance :/ I will try 8.8, since I stuck with 8.5 for so long now :P

[blabub]

Hi,

I got yesterday my 4870, then I installed thew 8.8 driver.

First I tried to run sauberbraten, works fine and good.


After that I tried to start WoW but had some problems.
With the normal libGL.so.1.2 my screen got strange effects. So that I have to restart the xserver.

Then I read that wine worked well with 8.4, so I replaced the libGL.so.1.2 with that one.

Now WoW starts and I can login, but there are no graphics, only a black screen.

After I selected my char the game load but the in-game graphic is not usable.

Here a few screens:

Login:
http://www12.file-upload.net/thumb/22.08.08/iasx9i.png (http://www.file-upload.net/view-1059267/login.png.html)

Loading:
http://www12.file-upload.net/thumb/22.08.08/qlbfqs.png (http://www.file-upload.net/view-1059268/load.png.html)

in-game:
http://www12.file-upload.net/thumb/22.08.08/eglhra.png (http://www.file-upload.net/view-1059269/game.png.html)

Anyone an idea how to solve it?

[Extreme Coder]

http://bugs.winehq.org/show_bug.cgi?id=12300
See the solution at the end of the post (the one with the M2UseShaders thing)

[blabub]

Hi,


Thanks for the link. It had set this already, will not work.
So the waiting will start until a better driver.

[Extreme Coder]

Hi,


Thanks for the link. It had set this already, will not work.
So the waiting will start until a better driver.
Maybe this is because of the replaced libGL file, I had some wierd effects too so I returned it to default again.

[mo0n_sniper]

See post
http://www.phoronix.com/forums/showthread.php?p=43617&posted=1#post43617

[LurkerMan]

Hi,



Then I read that wine worked well with 8.4, so I replaced the libGL.so.1.2 with that one.

Now WoW starts and I can login, but there are no graphics, only a black screen.





I did the same thing in Ubuntu 8.04. I simple ran the installer for the ATI fglrx driver and with the installer running but idle and retrieved the file from the temporary directory that the installer created. I dont play WOW but it sure made a difference with some of the native 3d apps that I am using. As well as 2d and 3d apps with Wine.
If Bridgeman or someone else from AMD/ATI is following this thread, it sure looks like a bug was introduced after the 8.4 release that is causing some of these screen corruptions and wine issues.

BTW. I use a HIS Radeon HD 3850.

[LordRaiden]

Maybe when you run it in a separate X server, the problem doesn't happen there, try running it in the same X server.
Also, why use D3D? I thought Wine worked best for WoW with OpenGL. But I use Cedega anyways, so I use D3D as default (since OpenGL can't run Specular Lighting and some other effects)
but with both Wine and Cedega, in both D3D and OpenGL, it's nowhere near Windows XP performance :/ I will try 8.8, since I stuck with 8.5 for so long now :P

With 8.8, I'm getting flickering/tearing even with the nopat kernel option. I use d3d because opengl has some issues when inside raids and instances ( map doesn't show up), and also some item icons seem to be wrong ( the wrong icon altogether).

I also get the benefits of specular lighting.

It's true that I use a separate X to play the game, but I still need the Virtual line in xorg.conf.
I'm OK with running the game in a separate X, since it has been a lot more stable ever since, and there has been no hits no performance.
I use an X1250 by the way.

[Extreme Coder]

With 8.8, I'm getting flickering/tearing even with the nopat kernel option. I use d3d because opengl has some issues when inside raids and instances ( map doesn't show up), and also some item icons seem to be wrong ( the wrong icon altogether).

I also get the benefits of specular lighting.

It's true that I use a separate X to play the game, but I still need the Virtual line in xorg.conf.
I'm OK with running the game in a separate X, since it has been a lot more stable ever since, and there has been no hits no performance.
I use an X1250 by the way.
I have an X1250 on my mobo, but I don't use it (I use the X1650 Pro)
BTW, how high have you set the settings ingame, and what FPS do you get?

[Guilo]

I have tried to replace libGL from ATI with libGL from Mesa. Although it solved the problem with wine, it is now impossible for me to change the settings in amdcccle.

Any idea ?

Thanks

[Extreme Coder]

I have tried to replace libGL from ATI with libGL from Mesa. Although it solved the problem with wine, it is now impossible for me to change the settings in amdcccle.

Any idea ?

Thanks

probably because the Mesa libGL is incompatible with amdcccle.

[blabub]

I get an libGl.so.1.2 from the mese package, now wine works but fall back zu software rendering.

I found some information about this:

# For direct rendering libGL has to be able to load one or more
# dri modules (files ending in _dri.so, like fglrx_dri.so).
# Gentoo's mesa looks for these files in the location specified by
# LIBGL_DRIVERS_PATH or LIBGL_DRIVERS_DIR, then in the hardcoded
# location /usr/$(get_libdir)/dri. Ati's libGL does the same
# thing, but the hardcoded location is /usr/X11R6/lib/modules/dri
# on x86 and amd64 32bit, /usr/X11R6/lib64/modules/dri on amd64
# 64bit. So we can either put the .so files in that (unusual,
# compared to "normal" mesa libGL) location or set
# LIBGL_DRIVERS_PATH. We currently do the latter. See also bug
# 101539.

# The problem with this approach is that LIBGL_DRIVERS_PATH
# *overrides* the default hardcoded location, it does not extend
# it. So if ati-drivers is merged but a non-ati libGL is selected
# and its hardcoded path does not match our LIBGL_DRIVERS_PATH
# (because it changed in a newer mesa or because it was compiled
# for a different set of multilib abis than we are) stuff breaks.

I think this is the source of the problem with wine.

LDD gave me that.

MESE-Lib:
linux-gate.so.1 => (0xf7ee6000)
libX11.so.6 => /usr/lib32/libX11.so.6 (0xf7d6b000)
libXext.so.6 => /usr/lib32/libXext.so.6 (0xf7d5d000)
libXxf86vm.so.1 => /usr/lib32/libXxf86vm.so.1 (0xf7d58000)
libXdamage.so.1 => /usr/lib32/libXdamage.so.1 (0xf7d55000)
libXfixes.so.3 => /usr/lib32/libXfixes.so.3 (0xf7d4f000)
libm.so.6 => /lib32/libm.so.6 (0xf7d2b000)
libpthread.so.0 => /lib32/libpthread.so.0 (0xf7d14000)
libdl.so.2 => /lib32/libdl.so.2 (0xf7d10000)
libdrm.so.2 => /usr/lib32/libdrm.so.2 (0xf7d07000)
libc.so.6 => /lib32/libc.so.6 (0xf7bb5000)
libxcb-xlib.so.0 => /usr/lib32/libxcb-xlib.so.0 (0xf7bb2000)
libxcb.so.1 => /usr/lib32/libxcb.so.1 (0xf7b9a000)
libXau.so.6 => /usr/lib32/libXau.so.6 (0xf7b97000)
/lib/ld-linux.so.2 (0xf7ee7000)
libXdmcp.so.6 => /usr/lib32/libXdmcp.so.6 (0xf7b92000)


ATI-Lib:
linux-gate.so.1 => (0xf7fdd000)
libpthread.so.0 => /lib32/libpthread.so.0 (0xf7f22000)
libXext.so.6 => /usr/lib32/libXext.so.6 (0xf7f14000)
libgcc_s.so.1 => /usr/lib32/libgcc_s.so.1 (0xf7f07000)
libc.so.6 => /lib32/libc.so.6 (0xf7db5000)
/lib/ld-linux.so.2 (0xf7fde000)
libX11.so.6 => /usr/lib32/libX11.so.6 (0xf7cc5000)
libXau.so.6 => /usr/lib32/libXau.so.6 (0xf7cc2000)
libxcb-xlib.so.0 => /usr/lib32/libxcb-xlib.so.0 (0xf7cc0000)
libxcb.so.1 => /usr/lib32/libxcb.so.1 (0xf7ca8000)
libdl.so.2 => /lib32/libdl.so.2 (0xf7ca4000)
libXdmcp.so.6 => /usr/lib32/libXdmcp.so.6 (0xf7c9e000)

[baqtor]

I've been following this thread for almost a month, but its hard for me to understand the fix completely (i am a Spanish speaker), can anyone post the fix for me? i would be great to have it in the first page.

thanks