Problem with 23 inch (1920x1200) Philips monitor in Linux and during bootup

[erlend]

And when you use a dvi->vga adapter till you get a working dvi cable?

Interestingly, that doesn't work. I can use the VGA cable with another computer and everything seems to be fine (I can increase the screen resolution to 1920x1200), however when I put an adapter (VGA->DVI) on the cable and plug that into my graphics card's DVI port Xorg complains that the EDID information is corrupt.

To summarise briefly, if I use DVI to desktop I get no image on the screen during boot, only once the nvidia drivers kick in do I see something on the scren (and this is at a low resolution 640x480). Xorg.0.log complains that the EDID information has a corrupt header. If I use the VGA cable with other computers (laptops and windows desktops) then everything is ok - I can right click the desktop and increase the resolution to 1920x1200). If I put a VGA->DVI converter on the VGA cable and connect this to my desktop I get problems again (I do see output during boot, but I'm still getting a corrupted EDID).

I think it may be the graphics card.

[Kano]

Yes, best replace it and get a new series card with vdpau support. That would be better for videos anyway.

[rohcQaH]

so have you tried manually providing an EDID or specifying the modeline yet?
If different computers can read the EDID, why don't you try dumping it on one of them?

What happens when you connect the monitor to the other output of your gfx card? If a pin is loose, that should solve it.

Your 7900gt seemed to work fine with your previous monitor, so I doubt it suddenly broke. Besides, acquiring the EDID is mostly a software task, not something with lots of hardware involved. Getting a new card may very well be a waste of a hundred bucks.

[erlend]

so have you tried manually providing an EDID or specifying the modeline yet?
If different computers can read the EDID, why don't you try dumping it on one of them?

I have not had success in getting another computer to read the EDID because I have only connected it to laptops (and they always return the EDID of the LCD screen attached to the laptop) or my flatmates desktop, which only has a VGA connector.

I tried setting the modeline as Kano suggested but the Xorg.0.log file says, "No valid modes found for 1920x1200". I also specified the UseEDID false and UseExactDVITimings True options.

What happens when you connect the monitor to the other output of your gfx card? If a pin is loose, that should solve it.

That doesn't help. Last night I also tried a different DVI-D cable. I also borrowed a DVI->VGA adapter to try the monitor on other machines - this worked fine (I could increase the resolution to 1920x1200).

Your 7900gt seemed to work fine with your previous monitor, so I doubt it suddenly broke. Besides, acquiring the EDID is mostly a software task, not something with lots of hardware involved. Getting a new card may very well be a waste of a hundred bucks.

I have read some stuff about minor incompatibilities between different DVI implementations. As for not seeing an image at bootup I wonder if the DVI EDID is stuck on sleep mode.

[rohcQaH]

I have not had success in getting another computer to read the EDID because I have only connected it to laptops (and they always return the EDID of the LCD screen attached to the laptop) or my flatmates desktop, which only has a VGA connector.

EDID works over VGA connectors, too. Also, nvidia-settings can dump the EDID of all connected monitors. If you only get the EDID of the internal LCD, that's a limitation of your dumping tool.

I'm not sure what xorg options you need to set to manually specify modelines. To override the EDID (once dumped), I did this:

Section "Screen"
	Option     "ConnectedMonitor"   "DFP-0,DFP-1"
	Option      "IgnoreDisplayDevices" "CRT, TV"
	Option     "CustomEDID" "DFP-0:/etc/X11/245B.edid; DFP-1:/etc/X11/214T.edid"
	...

I didn't need any further options. This is obviously for my dual-screen setup, adjust as needed.

[erlend]

EDID works over VGA connectors, too. Also, nvidia-settings can dump the EDID of all connected monitors. If you only get the EDID of the internal LCD, that's a limitation of your dumping tool.

Ok, I've just used switchxres on the Mac to capture the EDID... not really sure what to think of it though.

DDC block report generated by SwitchResX version 3.8.7 for display
VGA/SVGA Display (2)


-----------------------------------------------------
------------------- RAW DATA ------------------------
-----------------------------------------------------
      0  1  2  3  4  5  6  7  8  9  A  B  C  D  E  F
-----------------------------------------------------
0  | 00 FF FF FF FF FF FF 00 00 00 17 07 00 00 00 00
1  | 0A 0A 01 01 0C 1E 17 BE E8 2D C9 A0 57 47 98 27
2  | 12 48 4C 00 00 00 01 01 01 01 01 01 01 01 01 01
3  | 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01
4  | 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01
5  | 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01
6  | 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01
7  | 01 01 01 01 01 01 00 00 00 00 00 00 00 00 00 02

-----------------------------------------------------
 <  00FFFFFF FFFFFF00 00001707 00000000 0A0A0101 0C1E17BE E82DC9A0 57479827 12484C00 00000101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010101 01010000 00000000 00000002	> 

-----------------------------------------------------
	Valid EDID block: checksum passed

-----------------------------------------------------
------------------- MAIN EDID BLOCK -----------------
-----------------------------------------------------

EDID Version........1.1
Manufacturer........@@@
Product Code........5895 (1707) (0717) 
Serial Number.......00000000

Manufactured........Week 10 of year 2000
Max H Size..........30 cm
Max V Size..........23 cm
Gamma...............2.90

Display Supported Features:
---------------------------
	Power Management: Active off	Power Management: Suspend	Power Management: Standby

Display type:
------------- 
	RGB color display
	Display is non continuous frequency
	Default color space is not sRGB standard


Input signal & sync:
-------------------- 
Analog input with: 	0.700V / 0.300V
	Separate Sync
	Composite Sync


Color info:
----------- 
Red x = 0.625  Green x = 0.280  Blue x = 0.155  White x = 0.283
Red y = 0.342  Green y = 0.595  Blue y = 0.070  White y = 0.298

Established Timings:
-------------------- 

Manufacturer Reserved Timings:
------------------------------ 

Standard Timing Identification:
------------------------------- 

Monitor Description blocks:
--------------------------- 
	Descriptor #0 is Empty descriptor

	Descriptor #1 is Empty descriptor

	Descriptor #2 is Empty descriptor

	Descriptor #3 is Empty descriptor

My first thought is that the "first week of 2000" seems incorrect, but that's maybe just not been set. There doesn't seem to be any monitor descriptor timings.

[rohcQaH]

alright, that needs to be binary, not a text-dump. I've converted it, get it here, save it anywhere you like, then add it to your xorg.conf as mentioned above.

It looks like that EDID is incomplete, but anyway, please try. If it doesn't work, please attach xorg.conf and Xorg.0.log.

A new GPU won't solve that problem by the way.

[erlend]

alright, that needs to be binary, not a text-dump. I've converted it, get it here, save it anywhere you like, then add it to your xorg.conf as mentioned above.

Wow, thanks for that doing that. Here's my Xorg.0.log

http://pastebin.com/f7b553389

and my xorg.conf (a garbled mess, I appologisse for that and for the ati comments: I used to have an ati card a long time ago)

http://pastebin.com/f45b680a5

The EDID binary you gave me has had an effect - the DPI has changed (and now fonts in X look slightly funny).

[rohcQaH]

Looks like indeed the monitor is at fault, the EDID is broken. Windows somehow manages to avoid that by manually adding modelines. Yay for products tested on windows only.

Please find the technical specifications of your monitor, if they're equal to my Samsung 245B I'll send you my edid as a substitute.

Samsung 245B:
native resolution: 1920 x 1200
Hz: 56 - 75
kHz: 30 - 81

edit: oh, here they are:

Philips 230WP7NS:
native resolution: 1920 x 1200
Hz: 56 - 85 Hz
kHz: 30 - 94 kHz

It should accept every timing my 245B can do. I'll send you the edid when I'm back home (ETA: 8 hours)

[erlend]

Looks like indeed the monitor is at fault, the EDID is broken. Windows somehow manages to avoid that by manually adding modelines. Yay for products tested on windows only.

Please find the technical specifications of your monitor, if they're equal to my Samsung 245B I'll send you my edid as a substitute.


edit: oh, here they are:

It should accept every timing my 245B can do. I'll send you the edid when I'm back home (ETA: 8 hours)

I think this might be related to the issue, http://www.microsoft.com/whdc/device...edid_over.mspx
Windows users presumably install some driver which contains the EDID and modelines.

The monitor is listed here,

http://www.p4c.philips.com/cgi-bin/dcbint/cpindex.pl?ctn=230WP7NS/00&slg=en&scy=GB

and the driver for Windows is there for download. Can I extract the EDID from the driver .inf or .cab file?

Thanks for your help.

Erlend