Black sreen with 2xT1000 on xubuntu 20.04 and CUDA 12.1

Graphics / Linux Linux
1

Good afternoon!
Another one classic problem with black screen on loading system Xubuntu 20.04. It is clean install.
We use:
motherboard: Asus Sage W299 10G
CPU: Intel(R) Core™ i9-10900X CPU @ 3.70GHz
GPU: 2xNvidia T1000 (in 1 and 3 slots)
Drivers: CUDA 12.1 and 530.30.02
After installation see only black screen. In Xorg.log
[ 14.130] () NVIDIA(G0): Stereo disabled by request
[ 14.130] () NVIDIA(G0): NVIDIA SLI Mosaic mode selected.
[ 14.130] (WW) NVIDIA(G0): Invalid Multi-GPU option: ‘true’; using single GPU rendering.
[ 14.130] (WW) NVIDIA(G0): Base Mosaic is available only on screen 0. Disabling Base
[ 14.130] (WW) NVIDIA(G0): Mosaic.
[ 14.130] (EE) NVIDIA(G0): Only one X screen is supported when Base Mosaic is enabled.
[ 14.130] (EE) NVIDIA(G0): Disabling this screen.
[ 14.130] (EE) NVIDIA(G0): Failing initialization of X screen
[ 14.130] (II) UnloadModule: “nvidia”
[ 14.130] (II) UnloadSubModule: “wfb”
[ 14.130] (II) UnloadSubModule: “fb”
[ 14.130] (EE) Screen(s) found, but none have a usable configuration.
[ 14.130] (EE)
Fatal server error:
[ 14.130] (EE) no screens found(EE)

[ 14.130] (EE) Please also check the log file at “/var/log/Xorg.0.log” for additional information.
[ 14.130] (EE)
[ 14.131] (EE) Server terminated with error (1). Closing log file.

I saved xorg.conf in X11 as previous for two videocards in motherboard. But it is doesn’t work.
It is very common problem with it I often come across.
I attach needed log files: dmesg, X.log, syslog, lightdm.log, kern.log. And collect info by nvidia-bug-report.log.gz

Please, help me with it problem.
bugreport.tar (59.3 MB)

2

First, please make sure

sudo cat /sys/module/nvidia_drm/parameters/modeset

returns “N”
Then please delete your /etc/X11/xorg.conf
and create a new /etc/X11/xorg.conf only containing

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    BusID          "PCI:104:0:0"
    Option         "SLI" "Mosaic"
EndSection

then reboot. In case it still doesn’t work, please create a new nvidia bug-report.log.

3

Good afternoon
Yes, checked it.

cat /sys/module/nvidia_drm/parameters/modeset

return N
When I delete xorg.xonf file, system load, but only 1 card work. I’ll try your advice. But as I read, option sli mosaic need to write in section screen. Didn’t try it previously.

4

They may be specified either in the Screen or Device sections of the X config file.
https://download.nvidia.com/XFree86/Linux-x86_64/460.67/README/xconfigoptions.html

1 Like
5

Good afternoon
Could you help me? Same problem. And same system, but i need to change videocards. Now it is P2200 in 1 and 5 pcie slots.
If I don’t create xorg.conf system load. If i change mosaic and save xorg.conf on next boot black screen( If I try to enable second videocard on next boot blackscreen too(

cat /sys/module/nvidia_drm/parameters/modeset

return N
create a new /etc/X11/xorg.conf only containing

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    BusID          "PCI:104:0:0"
    Option         "SLI" "Mosaic"
EndSection

but error in xorg0.log because it is need basemosaic option.

By the way, if i run nvidia-bug-report.sh it is freeze and i couldn’t send you this info. nvidia-smi same problem, freeze.

6

arch.tar (10.1 MB)

Logs in attachment

In dmesg
[ 11.146586] nvidia-modeset: ERROR: GPU:1: Failed to query display engine channel state: 0x0000927c:0:0:0x00000057
[ 11.146601] nvidia-modeset: ERROR: GPU:1: Failed to query display engine channel state: 0x0000927c:1:0:0x00000057
[ 11.146610] nvidia-modeset: ERROR: GPU:1: Failed to query display engine channel state: 0x0000927c:2:0:0x00000057
[ 11.146618] nvidia-modeset: ERROR: GPU:1: Failed to query display engine channel state: 0x0000927c:3:0:0x00000057
[ 11.146651] BUG: unable to handle page fault for address: ffffbad3c13f1048
[ 11.146655] #PF: supervisor write access in kernel mode
[ 11.146658] #PF: error_code(0x0002) - not-present page

7

This rather looks like the second P2200 is either broken or incorrectly seated. Please check if it works alone in the slot of the other one.

8

Thank you for your quick reply! I’ll check it now and will write result

9

Checked. Two cards are work. I try to use firstly one card that we think was broken in slot 1. It is good.
Next i insert two cards in slots 1 & 5 as prev it was problem so black screen at loading.
Next as eject all pcie cards that it is not videocards. same problem.
Next I changed slot 5 to 2. So videocards in 1 & 2 slots. Same black screen and error message.
Found next when i delete xorg.conf and system start (only one card work). I open terminal, open nvidia-settings and it layout see what 2 cards present, but displays from 2nd card are disabled. Ok I click on advanced, then as previously set BasaMosaic see what can enable display from 2nd card to screen0. apply but nothing happen. but! It teminal see logs from nvidia-setting program Package xorg-server was not found in the pkg-config search-path … this package of course installed. And if i save config xorg.conf and reboot black screen( Try to change GPU-name to GPU-0, GPU-1 as prev. but no result(
maybe same advices? I can check it and show logs. Hope you help me. Motherboard Asus Sage W299 10G
and linux kernel 5.15.67-lowlevel.

10

Ok, please check if setting kernel parameter
iommu=off
makes it work.

11

Good afternoon!
Hope what it help with it, but no result( I add in grub loader when check kernel, iommu=off but black screen. In Xorg0.log last lines:

[ 10.187] (II) systemd-logind: logind integration requires -keeptty and -keeptty was not provided, disabling logind integration
[ 10.188] (II) xfree86: Adding drm device (/dev/dri/card0)
[ 10.188] (II) xfree86: Adding drm device (/dev/dri/card1)
[ 10.198] () OutputClass “nvidia” ModulePath extended to “/usr/lib/x86_64-linux-gnu/nvidia/xorg,/usr/lib/xorg/modules”
[ 10.198] () OutputClass “nvidia” ModulePath extended to “/usr/lib/x86_64-linux-gnu/nvidia/xorg,/usr/lib/x86_64-linux-gnu/nvidia/xorg,/usr/lib/xorg/modules”
[ 10.205] (–) PCI: (104@0:0:0) 10de:1c31:1028:131b rev 161, Mem @ 0xd7000000/16777216, 0xc0000000/268435456, 0xd0000000/33554432, I/O @ 0x0000b000/128, BIOS @ 0x???/524288
[ 10.205] (–) PCI:*(105@0:0:0) 10de:1c31:1028:131b rev 161, Mem @ 0xd5000000/16777216, 0xa0000000/268435456, 0xb0000000/33554432, I/O @ 0x0000a000/128, BIOS @ 0x???/131072
[ 10.206] (II) “glx” will be loaded. This was enabled by default and also specified in the config file.
[ 10.206] (II) LoadModule: “dbe”
[ 10.206] (II) Module “dbe” already built-in
[ 10.206] (II) LoadModule: “extmod”
[ 10.206] (II) Module “extmod” already built-in
[ 10.206] (II) LoadModule: “glx”
[ 10.207] (II) Loading /usr/lib/xorg/modules/extensions/libglx.so
[ 10.216] (II) Module glx: vendor=“X.Org Foundation”
[ 10.216] compiled for 1.20.13, module version = 1.0.0
[ 10.216] ABI class: X.Org Server Extension, version 10.0
[ 10.216] (II) LoadModule: “nvidia”
[ 10.216] (II) Loading /usr/lib/x86_64-linux-gnu/nvidia/xorg/nvidia_drv.so
[ 10.223] (II) Module nvidia: vendor=“NVIDIA Corporation”
[ 10.223] compiled for 1.6.99.901, module version = 1.0.0
[ 10.223] Module class: X.Org Video Driver
[ 10.224] (II) NVIDIA dlloader X Driver 530.30.02 Wed Feb 22 03:49:34 UTC 2023
[ 10.224] (II) NVIDIA Unified Driver for all Supported NVIDIA GPUs
[ 10.229] (II) Loading sub module “fb”
[ 10.229] (II) LoadModule: “fb”
[ 10.229] (II) Loading /usr/lib/xorg/modules/libfb.so
[ 10.230] (II) Module fb: vendor=“X.Org Foundation”
[ 10.230] compiled for 1.20.13, module version = 1.0.0
[ 10.230] ABI class: X.Org ANSI C Emulation, version 0.4
[ 10.230] (II) Loading sub module “wfb”
[ 10.230] (II) LoadModule: “wfb”
[ 10.230] (II) Loading /usr/lib/xorg/modules/libwfb.so
[ 10.231] (II) Module wfb: vendor=“X.Org Foundation”
[ 10.231] compiled for 1.20.13, module version = 1.0.0
[ 10.231] ABI class: X.Org ANSI C Emulation, version 0.4
[ 10.858] () NVIDIA(0): Depth 24, (–) framebuffer bpp 32
[ 10.858] (==) NVIDIA(0): RGB weight 888
[ 10.858] (==) NVIDIA(0): Default visual is TrueColor
[ 10.858] (==) NVIDIA(0): Using gamma correction (1.0, 1.0, 1.0)
[ 10.858] (II) Applying OutputClass “nvidia” options to /dev/dri/card1
[ 10.858] () NVIDIA(0): Option “Stereo” “0”
[ 10.858] () NVIDIA(0): Option “nvidiaXineramaInfoOrder” “DFP-4”
[ 10.858] () NVIDIA(0): Option “SLI” “off”
[ 10.858] () NVIDIA(0): Option “MultiGPU” “Off”
[ 10.858] () NVIDIA(0): Option “BaseMosaic” “on”
[ 10.858] () NVIDIA(0): Option “AllowEmptyInitialConfiguration”
[ 10.858] () NVIDIA(0): Stereo disabled by request
[ 10.858] () NVIDIA(0): NVIDIA SLI disabled.
[ 10.858] () NVIDIA(0): NVIDIA Multi-GPU disabled.
[ 10.858] () NVIDIA(0): Option “MetaModes” “GPU-0.DP-4: nvidia-auto-select +0+0, GPU-0.DP-0: nvidia-auto-select +3840+0, GPU-1.DP-4: nvidia-auto-select +7680+0”
[ 10.858] () NVIDIA(0): Enabling 2D acceleration
[ 10.858] (II) Loading sub module “glxserver_nvidia”
[ 10.858] (II) LoadModule: “glxserver_nvidia”
[ 10.858] (II) Loading /usr/lib/x86_64-linux-gnu/nvidia/xorg/libglxserver_nvidia.so
[ 10.932] (II) Module glxserver_nvidia: vendor=“NVIDIA Corporation”
[ 10.933] compiled for 1.6.99.901, module version = 1.0.0
[ 10.933] Module class: X.Org Server Extension
[ 10.933] (II) NVIDIA GLX Module 530.30.02 Wed Feb 22 03:51:52 UTC 2023
[ 10.934] (II) NVIDIA: The X server supports PRIME Render Offload.
[ 11.719] (II) NVIDIA(GPU-0): NVIDIA Base Mosaic enabled.

12

Good afternoon!
I got again server with [2xT1000 on xubuntu 20.04 and CUDA 12.1] but with other motherboard. I install clean xubuntu 20.04 lts. And i forgot to say what i install not only nvidia-driver, I install cuda with nvidia driver. It is install latest cuda 12.1 and nvidia driver 530.30.02. Same problem. If I set BaseMosaic true and save config, it is black screen after reboot. If i save config and set basemosaic false, system load but i couldn’t use two cards in one layout(
Hope You will help me. Our maybe you advice me how correctly downgrade cuda & driver version when it is installed. For example I need to install driver 515 remember it is work correct.

13

Guys, good afternoon!
Still have this problem. I didn’t understand why 2xT1000 couldn’t work in base mosaic mode. linux xubuntu 20.04.6 with cuda 11.7 and driver 515.43.04 . Please help. Can show all need information.

upd: And one more information. When I remove xorg.conf (as previously) system start. And i check this case connect display to 2nd card and open nvidia-setting and saw what display show and hide in loop in 2-3 second. If i disconnect it from card and plug again.

upd2: and in Xorg0.conf saw
[ 301.544] randr: failed to create shared pixmap
[ 301.544] (EE) NVIDIA(G0): Failed to configure Reverse PRIME on output DP-1-5.