Bug 29928

Summary: [RADEON-KMS, Evergreen, Eyefinity 5] Invalid output on 4th monitor
Product: DRI Reporter: ISHIKAWA Mutsumi <ishikawa>
Component: DRM/RadeonAssignee: Default DRI bug account <dri-devel>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: medium    
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg
none
BIOS on this graphic card (dump from /sys/bus/pci/devices/{pciid}/rom)
none
lspci -v output
none
kernel config
none
dmesg with drm_debug=1 none

Description ISHIKAWA Mutsumi 2010-09-01 01:22:04 UTC
Current Radeon KMS does not support Eyefinity 5 ?

I tested Eyefinity 5 with 4 monitors, 3 of these monitors outputs are fine,
but last monitor (4th monitor)'s output is very ugly.

Hardware setup:
 Graphic Card:
   - PowerColor Eyefinity 5 HD5770 1GB GDDR5
   - http://www.powercolor.com/Global/products_features.asp?id=240
   - This Graphic card has 5 mini DisplayPorts

 Monitors:
  2 x HYUNDAI W240D
   - Resolution: 1920x1200
   - connected to the card with mini-DP to Dual Link DVI passive convertor
     (this cable included the graphic card package)

  2 x BenQ G2400DW
   - Resolution: 1920x1200
   - connected to the card with mini-DP to DP cable (included the package)
     and DP to DVI active convertor (This one
     http://akiba-pc.watch.impress.co.jp/hotline/20100529/ni_cawesome.html
     this is a Japanese page. Sorry, I can not find English informations)

Software setup:
 Linux kernel version 2.6.36-rc3
 (from linus git tree, commit 2bfc96a127bc1cc94d26bfaa40159966064f9c8c)


 * Output images are attached.
   Picture 1 is monitor connected mini-DP port 1. It is a normal output of
   'ps aux on linux console'. Monitors connected on port 2 and port 3 are some as
   this one.
   Picture 2 is monitor conncted port 4 on the same time. I saw some kind of
   timing problem with noises.


 * On xf86-driver-ati (evergreen git branch) with individual output on all monitors
   xrandr setup, monitor 1,2,3's output are correct. But Monitor 4's output is
   same as Monitor 3 with some kind of timeing problem (timing problem and noises looks
   same as KMS console display).
   So, I think monitor 4 output on KMS is same as monitor 3 with some timing glitch.


 * A monitor connected largest number of port is ugly when 4 monitors are connected.
   When connected with only 3 monitors, these are fine.
   Ugly output is not depend on a kind of links and port.

   For example: (DP means connected with active convertor on this setting)

    Port 1 -> DVI1
    Port 2 -> DVI2
    Port 3 -> DP1
    Port 4 -> DP2
    Port 5 -> none

      Problem on DP2 monitor only



    Port 1 -> DVI1
    Port 2 -> DVI2
    Port 3 -> DP1
    Port 4 -> none
    Port 5 -> DP2

      Problem on DP2 monitor only



    Port 1 -> DP1
    Port 2 -> DP2
    Port 3 -> DVI1
    Port 4 -> DVI2
    Port 5 -> none

      Problem on DVI2 monitor only



    Port 1 -> DP1
    Port 2 -> DVI1
    Port 3 -> DP2
    Port 4 -> DVI2
    Port 5 -> none

      Problem on DVI2 monitor only



    Port 1 -> DP1
    Port 2 -> DVI1
    Port 3 -> DVI2
    Port 4 -> none
    Port 5 -> DP2

      Problem on DP2 monitor only



    Port 1 -> DP1
    Port 2 -> DVI1
    Port 3 -> DP2
    Port 4 -> DVI2
    Port 5 -> none

      Problem on DVI2 monitor only



    Port 1 -> DVI1
    Port 2 -> DVI2
    Port 3 -> DP1
    Port 4 -> none
    Port 5 -> none

      All 3 monitors are fine.



    Port 1 -> DVI1
    Port 2 -> DVI2
    Port 3 -> none
    Port 4 -> DP1
    Port 5 -> none

      All 3 monitors are fine.



    Port 1 -> DVI1
    Port 2 -> none
    Port 3 -> none
    Port 4 -> DVI2
    Port 5 -> DP1

      All 3 monitors are fine.


 * I know the Eyefinity restriction(related clock source). 2 of 4 monitor are linked
   with DisplayPort (with Active DVI convertor), other 2 are linked with DVI. These
   link settings are OK for Eyefinity 5. I checked this link config does works on
   Windows 7 (Individual Items are displyed on each monitor).

 * I think it is not a hardware problem. Monitors and link cables, active convetor
   exchange each other is not effect on the problems. And all hardwares works fine
   on other Graphic cards or Other OSs (e.g Windows).
Comment 1 ISHIKAWA Mutsumi 2010-09-01 01:26:29 UTC
Humm, pictures are larger than accept on this system.

Picture1 and 2 are put on the URL bellow:

Picture1: http://hanzubon.jp/tmp/radeon-kms/monitor1.jpg
Picture2: http://hanzubon.jp/tmp/radeon-kms/monitor4.jpg
Comment 2 ISHIKAWA Mutsumi 2010-09-01 01:27:44 UTC
Created attachment 38347 [details]
dmesg
Comment 3 ISHIKAWA Mutsumi 2010-09-01 01:34:27 UTC
Created attachment 38348 [details]
BIOS on this graphic card (dump from /sys/bus/pci/devices/{pciid}/rom)
Comment 4 ISHIKAWA Mutsumi 2010-09-01 01:36:53 UTC
Created attachment 38349 [details]
lspci -v output
Comment 5 ISHIKAWA Mutsumi 2010-09-01 01:38:59 UTC
Created attachment 38350 [details]
kernel config
Comment 6 ISHIKAWA Mutsumi 2010-09-01 02:15:22 UTC
Created attachment 38352 [details]
dmesg with drm_debug=1
Comment 7 ISHIKAWA Mutsumi 2010-09-19 20:09:38 UTC
I tested All 4 monitors connected with DP -> DVI acctive convertor.

It is same result (monitor connected on lates port show garbage).
Comment 8 Alex Deucher 2010-09-25 22:07:15 UTC
Does this set of kernel patches help (against 2.6.36)?
http://people.freedesktop.org/~agd5f/pll_fixes/
Comment 9 ISHIKAWA Mutsumi 2010-09-26 03:54:59 UTC
No. Same result with and without these patches (4th monitor display only show garbage)
Base kernel: linus git tree 32163f4b2cef28a5aab8b226ffecfc6379a53786
Comment 10 ISHIKAWA Mutsumi 2010-09-26 04:46:30 UTC
These pictures become any hint to solve the problem?

Picture1:
 http://hanzubon.jp/tmp/radeon-kms/monitor3_20100926.jpg

Picture2:
 http://hanzubon.jp/tmp/radeon-kms/monitor4_20100926.jpg

They are monitor3 and monitor4 pictures with X.

 * monitor3 looks fine.
 * monitor4 is NOT mirror mode (individual mode. rightof monitor3 on EXA setting in xorg.conf),
    but it displays same items as monitor3 with vertical/horizontal offset.
 * (Perhaps can not identify with these picture, but) monitor4's rightside/topside areas are drown with vertical line noise.
 * These offset and noises are same on the console without X.
Comment 11 ISHIKAWA Mutsumi 2010-11-03 08:51:45 UTC
I tested this problem on linux kernel 2.6.37-rc1 (from tar ball) again.

This problem is gone on the kernel. My 4th monitor looks fine.
(But I can not find which changeset(s) solve this problem from ChangeLog...)

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.