Nvidia driver 455.23.04 (beta?): lost 4K display

zebulon · 2020-09-23 05:21:00

Hi,

I do not know if I am the only one, but I lost the capability of displaying 4K via HDMI with latest nvidia-dkms driver 455.23.04 (screen is a Samsung 4K TV). "xrandr -q" only detects resolution up to 1920x1080@60Hz, whereas I was able to display 3840x2160@60Hz before.

Weird thing, when searching for drivers at https://www.nvidia.com/Download/Find.aspx?lang=en-us, it says beta if you search for RTX20 series and WHQL when searching for RTX30 series. I will try to revert to 450.66 to see if I can restore the 4k display capability. Anyone else experiencing the issue?

Last edited by zebulon (2020-09-23 05:22:49)

V1del · 2020-09-23 08:43:57

Works fine here on a GTX TITAN X. What's in your xorg log? Is the driver setup correctly apart from this (does acceleration work, does glxinfo -B /nvidia-smi return the correct renderer? Occasionally there could be race conditions or the like leading to incorrect xorg initialization which could show up as "errors" without being a real issue)

Last edited by V1del (2020-09-23 08:45:04)

zebulon · 2020-09-23 11:44:43

Thanks V1del. Will investigate a bit further when at home.

zebulon · 2020-09-24 04:00:36

I checked again, I cannot see any error in Xorg log. Also, I tried to readjust file /etc/X11/xorg.conf.d/20-nvidia.conf using:

Section "Monitor"
    # HorizSync source: edid, VertRefresh source: edid
    Identifier     "Monitor0"
    VendorName     "Unknown"
    ModelName      "SAMSUNG"
    HorizSync       15.0 - 185.0
    VertRefresh     24.0 - 120.0
    Option         "DPMS"
    Option         "UseEdidDpi" "False"
    Option         "DPI" "96 x 96"
EndSection

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BoardName      "GeForce RTX 2070 SUPER"
EndSection

Section "Screen"
    Identifier     "Screen0"
    Device         "Device0"
    Monitor        "Monitor0"
    DefaultDepth    24
    Option         "Stereo" "0"
    #Option         "metamodes" "1920x1080 +0+0; 1280x720 +0+0"
    Option         "metamodes" "3840x2160_60 +0+0"
    #Option         "metamodes" "2560x1440_40 +0+0; 1920x1080_120 +0+0"
    Option         "SLI" "Off"
    Option         "MultiGPU" "Off"
    Option         "BaseMosaic" "off"
    Option "ForceFullCompositionPipeline" "True"
    SubSection     "Display"
        Depth       24
    EndSubSection
EndSection

As you can see, I used only the metamode for 4Kresolution, but it still starts at 1920x1020.

One thing: I am using DP-1 port, not HDMI, I do not know if this has incidence.

zebulon · 2020-09-24 04:03:14

Also as requested:

[eric@arch Steam]$ nvidia-smi 
Thu Sep 24 06:02:17 2020       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.23.04    Driver Version: 455.23.04    CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce RTX 207...  Off  | 00000000:01:00.0  On |                  N/A |
|  0%   39C    P8    18W / 215W |    181MiB /  7979MiB |     16%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      2000      G   /usr/lib/Xorg                      88MiB |
|    0   N/A  N/A      4407      G   /bin/kwin_x11                       2MiB |
|    0   N/A  N/A      4483      G   /usr/bin/plasmashell               48MiB |
|    0   N/A  N/A      4555      G   /bin/amarok                         3MiB |
|    0   N/A  N/A      4688      G   ...akonadi_archivemail_agent        3MiB |
|    0   N/A  N/A      4716      G   .../akonadi_mailfilter_agent        3MiB |
|    0   N/A  N/A      4750      G   ...n/akonadi_sendlater_agent        3MiB |
|    0   N/A  N/A      4751      G   ...nadi_unifiedmailbox_agent        3MiB |
+-----------------------------------------------------------------------------+

[eric@arch Steam]$ glxinfo -B
name of display: :0
display: :0  screen: 0
direct rendering: Yes
Memory info (GL_NVX_gpu_memory_info):
    Dedicated video memory: 8192 MB
    Total available memory: 8192 MB
    Currently available dedicated video memory: 7794 MB
OpenGL vendor string: NVIDIA Corporation
OpenGL renderer string: GeForce RTX 2070 SUPER/PCIe/SSE2
OpenGL core profile version string: 4.6.0 NVIDIA 455.23.04
OpenGL core profile shading language version string: 4.60 NVIDIA
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile

OpenGL version string: 4.6.0 NVIDIA 455.23.04
OpenGL shading language version string: 4.60 NVIDIA
OpenGL context flags: (none)
OpenGL profile mask: (none)

OpenGL ES profile version string: OpenGL ES 3.2 NVIDIA 455.23.04
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.20

zebulon · 2020-09-24 04:04:04

V1del wrote:

Works fine here on a GTX TITAN X. What's in your xorg log? Is the driver setup correctly apart from this (does acceleration work, does glxinfo -B /nvidia-smi return the correct renderer? Occasionally there could be race conditions or the like leading to incorrect xorg initialization which could show up as "errors" without being a real issue)

Are you connected via DP or HDMI please?

seth · 2020-09-24 06:28:56

Please post your entire Xorg log and the output of "xrandr -q".
Then move away the xorg.conf, restart X11 and post an updated xorg log and "xrandr -q" output.

zebulon · 2020-09-24 06:53:45

seth wrote:

Please post your entire Xorg log and the output of "xrandr -q".
Then move away the xorg.conf, restart X11 and post an updated xorg log and "xrandr -q" output.

Will do this tonight. However yesterday I have been able to play a bit with my /etc/X11/xorg.conf.d/20-nvidia.conf file.
Adding options such as:

Section "Monitor"
   Option         "UseEdid" "FALSE"
EndSection

Section "Device"
   Option         "ModeValidation" "AllowNonEdidModes"
EndSection

led to detection of further resolutions, but not up to 3840.
Otherwise everything works fine (acceleration and OpenGL).

I think there is a problem with the Edid detection with latest drivers via DisplayPort. If I use no xorg.conf file, then there is a problem detecting the Dpi (I also used Option "UseEdidDPI" "FALSE" and force DPI to 96x96 for a long time). I will try what you suggest. Also I may try to downgrade the nvidia drivers to check if this is an issue with them.

zebulon · 2020-09-24 06:55:02

V1del wrote:

Works fine here on a GTX TITAN X. What's in your xorg log? Is the driver setup correctly apart from this (does acceleration work, does glxinfo -B /nvidia-smi return the correct renderer? Occasionally there could be race conditions or the like leading to incorrect xorg initialization which could show up as "errors" without being a real issue)

Question for you: is your 3840 resolution explicitely defined as a modeline in your xorg.conf, or as a "metamode" or do you have it automatic (undeclared)?

Last edited by zebulon (2020-09-24 07:33:48)

zebulon · 2020-09-27 11:11:40

So I reverted to nvidia 450.66 dkms drivers and got the 4K resolution detection back with xrandr -q. Definitely a bug in the new 455 drivers.
I will try to collect logs without any xorg.conf file to see if we can find out something.

rek2 · 2020-09-27 15:04:29

zebulon wrote:

Hi,
I do not know if I am the only one, but I lost the capability of displaying 4K via HDMI with latest nvidia-dkms driver 455.23.04 (screen is a Samsung 4K TV). "xrandr -q" only detects resolution up to 1920x1080@60Hz, whereas I was able to display 3840x2160@60Hz before.
Weird thing, when searching for drivers at https://www.nvidia.com/Download/Find.aspx?lang=en-us, it says beta if you search for RTX20 series and WHQL when searching for RTX30 series. I will try to revert to 450.66 to see if I can restore the 4k display capability. Anyone else experiencing the issue?

I have similar problem, I updated last night and after that my resolution is lower! to a default something is wrong with the linux+nvidia

also notice nvidia-settings stop working, said it was unable to find something
basically is like nvidia is not able to get the monitor capable resolutions
also opengl for mozilla hubs STOP working... my games seen to work(less fps) but works.. mozilla hubs that is webgl on chromium is not working at al.. I used toget 60 fps now is like 2-5 fps.

reverted driver and everything is working

Last edited by rek2 (2020-09-27 15:08:07)

V1del · 2020-09-27 15:18:44

@rek2 Your particular issue just sounds like the usual race condition with your graphics driver not being ready before the graphical target is loaded, this can generally be fixed by adding the nvidia modules to your initramfs: https://wiki.archlinux.org/index.php/NV … de_setting

This is a race condition, it can and can not happen with every driver on each bootup.

rek2 · 2020-09-27 15:50:58

V1del wrote:

@rek2 Your particular issue just sounds like the usual race condition with your graphics driver not being ready before the graphical target is loaded, this can generally be fixed by adding the nvidia modules to your initramfs: https://wiki.archlinux.org/index.php/NV … de_setting
This is a race condition, it can and can not happen with every driver on each bootup.

This makes no sense, I never had t o do anything before with this particular system76 PC, just updated and thats it. once in a while I have a issue with something in nvidia getting updated and not linux, in that case I just downgrade but in that case X wont even start... this is diff, this is the first time I ever had this with this card.
it has to be something new on nvidia+linux since they both updated at the same time, but since I downgraded and worked I did not try more things.
and I remember even video was choppy on chromium not just webgl.

❯ nvidia-smi -L
GPU 0: GeForce GTX 1080 Ti (UUID: GPU-36b39cb3-eaca-5e6e-ee3d-7d3cbc492564)

[updated] upgraded again, rebooted and same problem... now im back to older driver that works perfectly.
BTW thanks for there reply, and trying to help. I am fine for now since it works with the other driver, I try again next week.

Last edited by rek2 (2020-09-27 16:09:35)

zebulon · 2020-09-27 16:23:27

rek2 wrote:

I have similar problem, I updated last night and after that my resolution is lower! to a default something is wrong with the linux+nvidia

also notice nvidia-settings stop working, said it was unable to find something
basically is like nvidia is not able to get the monitor capable resolutions
also opengl for mozilla hubs STOP working... my games seen to work(less fps) but works.. mozilla hubs that is webgl on chromium is not working at al.. I used toget 60 fps now is like 2-5 fps.
reverted driver and everything is working

Yes, you can see detected resolutions using xrandr -q command. For me, the 4K resolution only shows with 450.66 driver, not 455.

zebulon · 2020-09-27 16:56:22

With 450.66:

$ xrandr -q
Screen 0: minimum 8 x 8, current 1920 x 1080, maximum 32767 x 32767
DP-0 disconnected (normal left inverted right x axis y axis)
DP-1 connected primary 1920x1080+0+0 (normal left inverted right x axis y axis) 1210mm x 680mm
   3840x2160     30.00 +  59.94    50.00    29.97    25.00    23.98    23.98  
   4096x2160     59.94    50.00    29.97    25.00    24.00    23.98  
   1920x1080    119.88*  100.00    60.00    59.94    50.00    29.97    25.00    23.98  
   1680x1050     59.95  
   1600x900      60.00  
   1440x900      59.89  
   1366x768      59.79  
   1280x1024     75.02    60.02  
   1280x800      59.81  
   1280x720      60.00    59.94    50.00  
   1152x864      75.00  
   1024x768      75.03    70.07    60.00  
   800x600       75.00    72.19    60.32  
   720x576       50.00  
   720x480       59.94  
   640x480       75.00    72.81    59.94  
HDMI-0 disconnected (normal left inverted right x axis y axis)
DP-2 disconnected (normal left inverted right x axis y axis)
DP-3 disconnected (normal left inverted right x axis y axis)
DP-4 disconnected (normal left inverted right x axis y axis)
DP-5 disconnected (normal left inverted right x axis y axis)

With 455.23.04:

$ xrandr -q
Screen 0: minimum 8 x 8, current 1920 x 1080, maximum 32767 x 32767
DP-0 disconnected (normal left inverted right x axis y axis)
DP-1 connected primary 1920x1080+0+0 (normal left inverted right x axis y axis) 1872mm x 1053mm
   1920x1080     60.00*+  59.94    50.00    29.97    25.00    23.98  
   1680x1050     59.95  
   1600x900      60.00  
   1440x900      59.89  
   1366x768      59.79  
   1280x1024     75.02    60.02  
   1280x800      59.81  
   1280x720      60.00    59.94    50.00  
   1152x864      75.00  
   1024x768      75.03    70.07    60.00  
   800x600       75.00    72.19    60.32  
   720x576       50.00  
   720x480       59.94  
   640x480       75.00    72.81    59.94  
HDMI-0 disconnected (normal left inverted right x axis y axis)
DP-2 disconnected (normal left inverted right x axis y axis)
DP-3 disconnected (normal left inverted right x axis y axis)
DP-4 disconnected (normal left inverted right x axis y axis)
DP-5 disconnected (normal left inverted right x axis y axis)

Last edited by zebulon (2020-09-27 17:02:07)

seth · 2020-09-27 17:27:31

rek2 wrote:

V1del wrote:
@rek2 Your particular issue just sounds like the usual race condition with your graphics driver not being ready before the graphical target is loaded, this can generally be fixed by adding the nvidia modules to your initramfs: https://wiki.archlinux.org/index.php/NV … de_setting
This is a race condition, it can and can not happen with every driver on each bootup.
This makes no sense, I never had t o do anything before

Simply Explained: Multithreaded Programs (also multitasking)

Edit: hope everyone got how race conditions work ;-)

Last edited by seth (2020-09-28 15:25:51)

V1del · 2020-09-27 17:31:16

Your case does really look like a regression I suggest your run a nvidia-bug-report.sh as root and post that together with the two xrandr -q to the nvidia linux boards or email linux-bugs [ at ] nvidia [ dot ] com with a relevant attachment.

@rek2 that's fine, but simply from your description this does make sense, it has all the signs of broken acceleration and something like vesa being used instead of the actual nvidia drivers (in contrast to zerbulon's case where it sounds like everything else is working properly just the resolution detection being off), without logs this is indeed just a guess, but it's a guess based on a lot of experience with similar issues. If you want to keep it that way for now no one is stopping you

zebulon · 2020-09-28 12:52:52

V1del wrote:

Your case does really look like a regression I suggest your run a nvidia-bug-report.sh as root and post that together with the two xrandr -q to the nvidia linux boards or email linux-bugs [ at ] nvidia [ dot ] com with a relevant attachment.

Thanks I have done that and will return with the outcome. Looks there are quite a few bugs with this driver on the Nvidia developer forum.
Now you are right about my situation: driver 455.23.04 works perfectly well otherwise in my case. There is no other issue than the mere detection of Edid modes higher than Full HD. Reading again rek2's post, it is indeed a different matter.

zebulon · 2020-09-29 06:07:34

By the way, 455.23.04 is indeed a beta driver as confirmed on the nvidia boards https://forums.developer.nvidia.com/t/c … ases/28500
It probably should not have been packaged into the stable repository yet.

V1del · 2020-09-29 07:51:36

It's also a hardware enablement driver for the new generation of graphics cards, which is imo a valid point to package it up on a rolling release, there was at least one thread a few days ago where someone set up with a new 30xx card and they need this driver.

And you seem to be the only person so far with this particular issue here, so it's likely not an as far reaching of a regression. FWIW regarding your question a while ago, my resolution is auto detected and that's still the case in my particular setup.

Last edited by V1del (2020-09-29 07:56:02)

zebulon · 2020-09-29 07:55:44

V1del wrote:

It's also a hardware enablement driver for the new generation of graphics cards, which is imo a valid point to package it up on a rolling release.

That is a fair point, you are talking about the 3000 series, which have only been recently released.
That said there are specific beta packages for nvidia drivers on the AUR (nvidia-beta and nvidia-beta-dkms) and it makes sense to keep it that way, especially if the beta drivers contain serious bugs. They could have gone to the Testing repository as well. It is also true that nvidia website was not clear about the beta status, as I pointed out earlier, only the nvidia Linux forums were.

Last edited by zebulon (2020-09-29 07:59:09)

zebulon · 2020-09-29 07:58:42

I have posted nvidia-bug-reports with logverbose level 6 for both 450 and 455 versions of the drivers, the thread can be seen here: https://forums.developer.nvidia.com/t/l … -04/155729

zebulon · 2020-09-29 08:24:39

I had a look at my bug report files and found out something from the Xorg.0.log files:

For 450.66:

[     5.036] (--) NVIDIA(GPU-0): SAMSUNG (DFP-1): connected
[     5.036] (--) NVIDIA(GPU-0): SAMSUNG (DFP-1): Internal TMDS
[     5.036] (--) NVIDIA(GPU-0): SAMSUNG (DFP-1): 600.0 MHz maximum pixel clock
...
[     5.101] (==) NVIDIA(0): No modes were requested; the default mode "nvidia-auto-select"
[     5.101] (==) NVIDIA(0):     will be used as the requested mode.
[     5.102] (==) NVIDIA(0): 
[     5.103] (II) NVIDIA(0): Validated MetaModes:
[     5.103] (II) NVIDIA(0):     "DFP-1:nvidia-auto-select"
[     5.103] (II) NVIDIA(0): Virtual screen size determined to be 3840 x 2160

For 455.23.04:

[     5.308] (--) NVIDIA(GPU-0): SAMSUNG (DFP-1): connected
[     5.308] (--) NVIDIA(GPU-0): SAMSUNG (DFP-1): Internal TMDS
[     5.308] (--) NVIDIA(GPU-0): SAMSUNG (DFP-1): 165.0 MHz maximum pixel clock
...
[     5.358] (==) NVIDIA(0): No modes were requested; the default mode "nvidia-auto-select"
[     5.358] (==) NVIDIA(0):     will be used as the requested mode.
[     5.358] (==) NVIDIA(0): 
[     5.360] (II) NVIDIA(0): Validated MetaModes:
[     5.360] (II) NVIDIA(0):     "DFP-1:nvidia-auto-select"
[     5.360] (II) NVIDIA(0): Virtual screen size determined to be 1920 x 1080

There is a discrepancy between the maximum pixel clock values (165 vs 600 MHz), and am wondering if this in the end affects the virtual screen size as determined by nvidia-auto-select.

zebulon · 2020-09-29 08:55:34

Also, xrandr --verbose (line 8285 and 8378 of the bug reports for 455 and 450, repectively) shows that both versions read the same EDID information, however the computed modes are different, maxing out at 1920 for 455, and with different maximum clock frequencies.

zebulon · 2020-10-04 14:25:49

I am now in contact with nvidia developers who have been very prompt at returning to me. It turns out the new driver version rejects 4K resolutions due to the maximum pixel clock calculation on the Display Port 1 on the TV. Version 450.66 got 600 MHz whereas 453 gets 165 MHz. The reason behind this is still under investigation.

Arch Linux

#1 2020-09-23 05:21:00

Nvidia driver 455.23.04 (beta?): lost 4K display

#2 2020-09-23 08:43:57

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#3 2020-09-23 11:44:43

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#4 2020-09-24 04:00:36

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#5 2020-09-24 04:03:14

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#6 2020-09-24 04:04:04

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#7 2020-09-24 06:28:56

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#8 2020-09-24 06:53:45

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#9 2020-09-24 06:55:02

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#10 2020-09-27 11:11:40

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#11 2020-09-27 15:04:29

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#12 2020-09-27 15:18:44

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#13 2020-09-27 15:50:58

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#14 2020-09-27 16:23:27

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#15 2020-09-27 16:56:22

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#16 2020-09-27 17:27:31

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#17 2020-09-27 17:31:16

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#18 2020-09-28 12:52:52

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#19 2020-09-29 06:07:34

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#20 2020-09-29 07:51:36

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#21 2020-09-29 07:55:44

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#22 2020-09-29 07:58:42

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#23 2020-09-29 08:24:39

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#24 2020-09-29 08:55:34

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

#25 2020-10-04 14:25:49

Re: Nvidia driver 455.23.04 (beta?): lost 4K display

Board footer