upon boot up only 30 Hz on 4k display connected via DP

Hi,

upon booting I only manage to have 30Hz refresh rate. In the session I can manually force the screen ASUS PB27U to DP 1.2 via OSD. Then with nvidea-settings I can choose 4k resolution and choose and successfully apply 60Hz. When I keep this settings (and also save them in Xorg.conf) upon reboot the screen goes blank. Resetting the display to DP 1.1 via OSD and rebooting results in 30Hz refresh rate again.

Maybe this sniplet from Xorg.0.log is informative:

[ 8.159] (–) NVIDIA(GPU-0): AUS ASUS PB27U (DFP-2): connected
[ 8.159] (–) NVIDIA(GPU-0): AUS ASUS PB27U (DFP-2): Internal DisplayPort
[ 8.159] (–) NVIDIA(GPU-0): AUS ASUS PB27U (DFP-2): 960.0 MHz maximum pixel clock
[ 8.159] (–) NVIDIA(GPU-0):
[ 8.163] (WW) NVIDIA(0): No valid modes for “3840x2160_60+0+0”; removing.
[ 8.163] (WW) NVIDIA(0):
[ 8.163] (WW) NVIDIA(0): Unable to validate any modes; falling back to the default mode
[ 8.163] (WW) NVIDIA(0): “nvidia-auto-select”.
[ 8.163] (WW) NVIDIA(0):
[ 8.163] (II) NVIDIA(0): Validated MetaModes:
[ 8.163] (II) NVIDIA(0): “DFP-2:nvidia-auto-select”

I also manually added this to Xorg.conf section Monitor:

ModeLine       "3840x2160"   533.25   3840 3888 3920 4000   2160 2163 2168 2222 +hsync -vsync

No success.

The GPU is NVIDIA GPU Quadro K420 (GK107GL)

Any help is appreciated!

Please set up an xorg.conf and add

Option "ModeDebug" "true"

inside the device section.
Afterwards, reboot with DP 1.2 enabled, then run nvidia-bug-report.sh as root and attach the resulting .gz file to your post. Hovering the mouse over an existing post of yours will reveal a paperclip icon.
https://devtalk.nvidia.com/default/topic/1043347/announcements/attaching-files-to-forum-topics-posts/

Dear generix,

please find attached the bug report. With DP1.2 active I cannot login as there is a blank screen. The bug report was established with DP1.1.
nvidia-bug-report.log.gz (1.26 MB)

[     8.373] (WW) NVIDIA(GPU-0):     Mode is rejected: DP Bandwidth check failed.

Maybe just a bad DP-cable?

I was also wondering about this, but is rather says there is a disagreement with the monitor capabilities …

Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (II) NVIDIA(GPU-0): --- Building ModePool for AUS ASUS PB27U (DFP-2) ---
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):   Validating Mode "3840x2160_60":
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     Mode Source: EDID
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     3840 x 2160 @ 60 Hz
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):       Pixel Clock      : 533.25 MHz
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):       HRes, HSyncStart : 3840, 3888
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):       HSyncEnd, HTotal : 3920, 4000
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):       VRes, VSyncStart : 2160, 2163
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):       VSyncEnd, VTotal : 2168, 2222
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):       Sync Polarity    : +H -V
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     The EDID for AUS ASUS PB27U (DP-1) contradicts itself:
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     mode "3840x2160" is specified in the EDID; however, the
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     EDID's valid HorizSync range (135.000 kHz) would exclude
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     this mode's HorizSync (133.313 kHz); ignoring HorizSync
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     check for mode "3840x2160".
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     Mode is rejected: DP Bandwidth check failed.
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     GPU extended capability check failed.
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):     Mode "3840x2160_60" is invalid.
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (WW) NVIDIA(GPU-0):
Nov 14 18:52:46 InaUbuntu /usr/lib/gdm3/gdm-x-session[1939]: (II) NVIDIA(GPU-0):   Validating Mode "3840x2160_30":

The cable is new and 60 Hz is working when monitor is in DP 1.2 and the higher refresh rate it initiated manually.

I replaced the cable with the one from the office (in fact running 4k 60 Hz on an identical monitor). Sadly there is no change. Same errors in Xorg.0.log and screen remains blank when monitor is DP1.2 upon boot.

Is there anotherway to override the DP Bandwidth check ? I tried to use:

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BoardName      "Quadro K420"
    Option         "ModeDebug" "true"
    Option         "ModeValidation" "NoDisplayPortBandwidthCheck"
EndSection

But this resulted in blank screen for DP1.2 and DP1.1. I had to use failsafe graphics to continue.

Rather looks like the link just can’t be trained to high enough speeds to carry the 4k@60 on initial boot. Does power cycling the monitor make it available?

Sorry, I do not understand. How can I test this?

Boot with DP1.2 enabled to black screen, then unplug and replug the monitor’s power cord.

Ok tried this, did not work.

I made the following interesting observation. When the monitor is set to DP1.2 the screen goes blank after boot-up messages from the BIOS but BEFORE I can see the GRUB screen. I thought that it went on since I I could not see anything but in fact I can neither PING nor SSH to the “blank” machine.

It seems it got stuck already on GRUB. I double-checked syslog and in fact I only see two intial kernel lines:

Nov 20 20:51:09 InaUbuntu kernel: [ 0.000000] Linux version 5.0.0-36-generic (buildd@lgw01-amd64-060) (gcc version 7.4.0 (Ubuntu 7.4.0-1ubuntu1~18.04.1)) #39~18.04.1-Ubuntu SMP Tue Nov 12 11:09:50 UTC 2019 (Ubuntu 5.0.0-36.39~18.04.1-generic 5.0.21)

The first one came from first boot-up when I checked the SSH access and the second when I set back to DP1.1 again.

Any ideas?

You’re using old style bios boot, maybe playing with grub’s
GRUB_GFXMODE
GRUB_GFXPAYLOAD_LINUX
options setting lower console resolution helps.

I tried yesterday lower resolutions and also the setting console. With DP1.2 GRUB does not start, with DP1.1 no problem.

When I boot with DP1.1 into grub and then adjust the monitor to DP1.2 it runs smoothly to the 60@4k with an empty xorg.conf. The Payload Option I have not yet tested but this probably does not affect the start of GRUB.

The BIOS messages are shown in DP1.2,somehow GRUB fails to use the screen.

Can you boot a live-dvd in efi mode?

The motherboard is quite old (MSI MS-7641/760GM-P21(FX)). I have never seen an option for EFI. In fact I looked yesterday whether there could be any tweaks impacting GRUB. I do have also the newest BIOS (from 2015).

From the OP logs, the monitor’s EDID has a mistake:
Valid HSync Range : 135.0 KHz - 135.0 KHz

And unless you get lucky, no video mode is going to hit that value.

Typical 4K (3840x2160) :-

  • 60Hz RVB hsync=133.25 KHz
  • 60Hz no RVB hsync=134.18 KHz;

Try:
Option “ModeValidation” “NoHorizSyncCheck”

HTH.

The log is from a situation with DP1.1.

The blank screen is caused by a failure even to load GRUB2: When GRUB2 is loaded it somehow checks for the graphic adapter and fails to proceed but only if the display is in DP1.2 (if GRUB2 would work correctly apart from the graphics it would boot with the default setting but this is not the case).

Strangely the BIOS messages are seen under all circumstances but GRUB2 is sensitive to the DP1.2 mode of the monitor.

I would guess this is a bug of the interplay GRUB2/nvidia. With GRUB2 probing for the graphics adapter …

A similar cases were observed with other nvidia card:
https://www.reddit.com/r/linuxquestions/comments/c1r3pu/grub_wont_load_when_connected_to_monitor_via/
https://forum.level1techs.com/t/odd-displayport-grub2-boot-issue/136745

I did file a bug in launchpad: https://bugs.launchpad.net/ubuntu/+source/grub2/+bug/1853735

No answer yet.

As the problem is observed with nvidia cards only maybe it is worth filing a bug with nvidia also.

Just another data point…
My old BIOS booting *buntu 18.04LTS on core2duo with gtx1050ti has been working fine with DP1.2 4K monitor BDM4350 for last year.
The desktop boots to 50Hz but works fine at 24, 50 & ~60Hz refresh for video playback.
I have no idea what the initial GRUB modes are as the monitors firmware is hopeless. It assumes 4K==60Hz & the monitor EDID only has 60Hz modes (as I recall).