Bug 8541

Summary: Strange effect?
Product: xorg Reporter: Asbjørn Sannes <ace>
Component: Driver/RadeonAssignee: xf86-video-ati maintainers <xorg-driver-ati>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: high CC: glisse, mat
Version: 7.1 (2006.05)   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
my xorg.conf file
none
Xorg.0.log
none
fglrx log
none
newer log with EDID data
none
newer log with EDID data, plain text
none
turns on ddx always patch none

Description Asbjørn Sannes 2006-10-07 03:52:12 UTC
I'm not sure how to word it correctly, so it seems to be that the rendered
screen seems to be offset incorrectly, and some shadow effect or artifact. I
have tried looking for a similiar bugs, but havn't found any.

I have some photos at http://www.sannes.org/misc/mesa-problem/ (especially the
527 one you can see the shadowing effect quite clearly), I see more or less the
same thing with quake2-icculus, qudos and enemy territory.

I have gentoo installed with 64bit and all the applications runs in 64bit mode.
mesa-6.5.1-r1
xf86-video-ati-6.6.3
xorg-x11-7.1

and kernel modules from git

I get the same when running with export LIBGL_ALWAYS_INDIRECT=1 before starting
the games.

Graphics card:
05:00.0 VGA compatible controller: ATI Technologies Inc R423 UK [Radeon X800SE
(PCIE)]
05:00.1 Display controller: ATI Technologies Inc Radeon R423 UK (PCIE) [X800 SE]
(Secondary)

I hope this isn't a duplicate bug, and if it is it is probably because I don't
know what to search for in this case. I can probably provide alot more
information if you just tell me what you want :)
Comment 1 Asbjørn Sannes 2006-10-07 03:57:12 UTC
Also, it is only in fullscreen mode that this happens. (works fine in windowed
mode).
Comment 2 Jerome Glisse 2006-10-07 05:53:30 UTC
Do you have enabled page flip in your xorg conf ? If so please
disable it and retry.
Comment 3 Asbjørn Sannes 2006-10-07 08:57:31 UTC
Created attachment 7278 [details]
my xorg.conf file

Tried to disable it (with "off") then disabled all options in my xorg.conf  
(which I had enabled) for the driver by uncommenting it, still same strange 
offset.. 

I'm attaching my xorg.conf, maybe it could shed some light on things? :)
Comment 4 Asbjørn Sannes 2007-01-13 03:48:56 UTC
Seems this has nothing to do with 3d acceleration at all, I can get the same
problem if I just change the resolution to something else with xrandr.

$ xrandr     
 SZ:    Pixels          Physical       Refresh
*0   1680 x 1050   ( 431mm x 272mm )  *60  
 1   1280 x 1024   ( 431mm x 272mm )   60  
 2   1280 x 960    ( 431mm x 272mm )   60  
 3   1152 x 864    ( 431mm x 272mm )   60  
 4   1024 x 768    ( 431mm x 272mm )   60  
 5    832 x 624    ( 431mm x 272mm )   60  
 6    800 x 600    ( 431mm x 272mm )   60  
 7    640 x 480    ( 431mm x 272mm )   60  
 8    640 x 350    ( 431mm x 272mm )   60  
Current rotation - normal
Current reflection - none
Rotations possible - normal 
Reflections possible - none

Some resolutions are good, others bad..
good: 1680 x 1050, 1152 x 864,832 x 624, 640 x 480, 640 x 350
bad: 1280 x 1024, 1280 x 960, 1024 x 768, 800 x 600

.. Any ideas?
Comment 5 Michel Dänzer 2007-01-15 07:59:09 UTC
Please attach the full log file.
Comment 6 Asbjørn Sannes 2007-01-16 02:24:36 UTC
Created attachment 8411 [details]
Xorg.0.log

I switched to 800x600 and back with xrandr and then copied the log (if it
matters).

Added two pictures at http://www.sannes.org/misc/mesa-problem/ that are not in
a game, but same effect.
Comment 7 Asbjørn Sannes 2007-03-29 04:54:40 UTC
I did additional tests with newest ubuntu amd64 live/install cd and it is the same problem.

Next I tried a different lcd panel and it worked! Of course, this led me to believe that it might be the panel all by itself, but then I took the panel to another computer, different OS (windows) and it worked fine .. then I noticed something, namely that when going through the menus of the panel and looking at display information it said 800x600 @75 hz ..

So I went back to my computer and looked at what my panel said when it was supposed to be in 800x600 @60 .. it said 1680x1048 @ 60 hz (1680x1048??) .. anyways, I tested with the fglrx drivers and when I put that into 800x600 it is at 75 hz, and it works fine.. then I did xrandr with -r 60 and that worked remarkedly well aswell, both times giving the correct information to the screen ..

On further note, I have tried this with both dri and glx disabled .. and the native resolution for my screen is 1680x1050 (at which it works perfectly)
Comment 8 Asbjørn Sannes 2007-03-29 05:18:59 UTC
Created attachment 9350 [details]
fglrx log

fglrx log ..xrandr gives this:
$ xrandr
 SZ:    Pixels          Physical       Refresh
 0   1920 x 1080   ( 432mm x 272mm )   30  
 1   1776 x 1000   ( 432mm x 272mm )   30  
*2   1680 x 1050   ( 432mm x 272mm )  *60  
 3   1280 x 1024   ( 432mm x 272mm )   75   70   60  
 4   1280 x 720    ( 432mm x 272mm )   60  
 5   1152 x 864    ( 432mm x 272mm )   75   70   60  
 6   1024 x 768    ( 432mm x 272mm )   75   72   70   60  
 7   1024 x 480    ( 432mm x 272mm )   60  
 8    960 x 720    ( 432mm x 272mm )   60  
 9    864 x 648    ( 432mm x 272mm )   60  
 10   856 x 480    ( 432mm x 272mm )   60  
 11   848 x 480    ( 432mm x 272mm )   60  
 12   800 x 600    ( 432mm x 272mm )   75   72   70   60   56  
 13   720 x 576    ( 432mm x 272mm )   60  
 14   720 x 480    ( 432mm x 272mm )   60  
 15   704 x 480    ( 432mm x 272mm )   60  
 16   640 x 480    ( 432mm x 272mm )   75   72   60  
 17   640 x 432    ( 432mm x 272mm )   60  
 18   640 x 400    ( 432mm x 272mm )   75   60  
 19   640 x 350    ( 432mm x 272mm )   70  
 20   512 x 384    ( 432mm x 272mm )   60  
 21   400 x 300    ( 432mm x 272mm )   75   60  
 22   320 x 240    ( 432mm x 272mm )   75   60  
 23   320 x 200    ( 432mm x 272mm )   75   60  
Current rotation - normal
Current reflection - none
Rotations possible - normal 
Reflections possible - none

two first resolutions give "out of range" on the panel
Comment 9 Jerome Glisse 2007-03-29 05:50:52 UTC
OK, this bug is due to wrong modeline, i believe this is
related to other bug about modeline validation.

For instance:
Radeon:
(II) RADEON(0): Modeline "800x600"  119.00  800 1728 1760 1840  600 1053 1059 1080 +hsync +vsync
Fglrx (more saner looking modeline):
(II) fglrx(0): Modeline "800x600"   45.50  800 840 920 1040  600 601 604 625 +hsync
Comment 10 Asbjørn Sannes 2007-03-29 06:40:59 UTC
Created attachment 9352 [details]
newer log with EDID data

It is the log from the radeon driver in git and seem to have more info..
Comment 11 Asbjørn Sannes 2007-03-29 06:42:22 UTC
Created attachment 9353 [details]
newer log with EDID data, plain text
Comment 12 Asbjørn Sannes 2007-03-29 07:13:30 UTC
Created attachment 9354 [details] [review]
turns on ddx always patch

Jerome Glisse gave me this to turn on ddx instead of rmx for my monitor, applied against git, it is a hack and not a solution, but it certainly solves my problem :)
Comment 13 Jerome Glisse 2007-03-29 07:15:24 UTC
It seems that the display doesn't like us sending RMXed mode
and wants to use it's own (ugly hack attached fix the problem).
I am not familiar with this RMX stuff so i let this to some
with more knowledge on it :)

(a small confusion in Asbjørn post, the patch disable use of
card RMX (at least this is my understanding, and let the display
use its own RMX unit).
Comment 14 Matthias Hopf 2007-04-18 09:20:23 UTC
Instead of using this patch, just setting

  Option "DDCMode" "true"

should have the same effect. This should probably be documented better. Can you test this, please?

This is related to bug #10620.
Comment 15 Asbjørn Sannes 2007-04-19 00:34:15 UTC
Works for me (only tested on git, could test on other stuff if you like).

I still believe this is something that should work out of the box :)
Comment 16 Matthias Hopf 2007-04-19 02:58:28 UTC
If you tell us, who is at fault (gfx card or monitor), and how to detect this situation, you might have a chance ;)
Comment 17 Asbjørn Sannes 2007-04-19 03:14:45 UTC
Well, it worked on a windows machine and with the proprietary driver, so I guess their default is to have "DDCMode" on as default perhaps .. question is, are there more monitors that work with this setting on or off? And maybe blacklist those monitors that fail .. I'm guessing it is a monitor issue since it is only this one monitor that is fails with .. however, having only tested two lcd panels ... it is  hard to say (well, havn't seen others with the same issue) ..
Comment 18 Matthias Hopf 2007-04-19 04:58:41 UTC
We would need to know how to identify monitors that need the chip's RMX scaler active. There are at least Apple flat panels that need it.
Comment 19 Michel Dänzer 2007-04-19 05:48:01 UTC
(In reply to comment #18)
> We would need to know how to identify monitors that need the chip's RMX scaler
> active. There are at least Apple flat panels that need it.

Apple panels only have their native mode in EDID. Are there any displays where this is true but which work better without RMX (currently - ss the idea of RMX is to always give the panel its native timings, there's probably a bug there somewhere)?
Comment 20 Asbjørn Sannes 2007-10-18 02:56:50 UTC
Seems like in the latest git versions I do not have to set DDCMode "true", so this bug could probably be closed?
Comment 21 Alex Deucher 2007-10-18 06:43:59 UTC
radeon in ati 6.7.19x and master will use DDC modes by default if available.
Comment 22 Alex Deucher 2007-11-12 15:03:31 UTC
closing

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.