Monitor is not detected at all, and ConnectedMonitor / UseDisplayDevice do not help.

[netllama]

Thanks. With your binary EDID, I'm able to reproduce this problem, and have opened bug 273796.

[alaric]

This comment is for other users that might read this who have the same problem(s).

To summarize:

1) The (CRT) monitor, connected to the DVI output, is not detected at all.
2) The EDID is garbled.

These are two seperate problems. The first problem stems from the fact that a CRT connected to the DVI plug can be seens as being a DFP. You can solve this by using the Option "ConnectedMonitor" "CRT-1" (The '1' means the DVI plug for the card that I own).

An additional problem was that, because the EDID was garbled, the above solution did not work with driver 1.0-8776. It does work with driver 1.0-9631 however.

Nevertheless, in order to solve the garbled EDID (which is the result of the driver thinking that it's EDID version 2.0, while it's not) is to use Option "CustomEDID".

The following Device section gives me a perfect result:

# The DVI output plug on the FX5200 card with the VMP17
# (IIyama Vision Master Pro17 - MT-9017T) connected (using a DVI->VGA
# adaptor). The VMP17 gives a wrong EDID header, therefore the EDID is
# replaced with one from the file VMP17edid.bin. Moreover, this EDID
# (apart from the header directly from the monitor) contains wrong
# frequencies (too low), therefore also EdidFreqs and EdidMaxPClkCheck
# are being ignored.
Section "Device"
        Identifier      "FX5200_DVI_VMP17"
        Driver          "nvidia"
        BusID           "PCI:1:0:0"
        Screen          0
        Option          "IgnoreDisplayDevices" "DFP, TV"
        Option          "ConnectedMonitor" "CRT-1"
        Option          "CustomEDID" "CRT-1:/etc/X11/VMP17edid.bin"
        Option          "ModeValidation" "CRT-1: NoEdidMaxPClkCheck"
        Option          "HorizSync"   "CRT-1: 27-96"
        Option          "VertRefresh" "CRT-1: 50-160"
EndSection

The used file /etc/X11/VMP17edid.bin is attached as VMP17edid.bin.gz and is the same as what my IIyama Vision Master Pro17 sends to the card when connected to the VGA plug (see startx -- -logverbose 6) but with the first 8 bytes replaced with "00 ff ff ff ff ff ff 00" and byte 128 replaced to fix the checksum.

Now going to get TwinView working again,
Alaric

[alaric]

Lonni,

my IIyama Vision Master Pro410 (A702HT, manufactured July 2000) suffers from the exact same problems. However, because in this case the raw EDID starts with 0x00, the driver didn't mistake it for an EDID version 2.0 and simply dediced to ignore the EDID (cannot read EDID for device DFP-0).

Again, when I replace the first 8 bytes, I get sane results. So, I am starting to get the feeling that the nvidia drivers could do better: apparently there are more monitors out there that don't start with 00 ff ff ff ff ff ff 00.

I've attached the raw EDID of this monitor too (it looks quite different).

Alaric