Lost GUI / HDMI Output on DGX after purging and reinstalling NVIDIA drivers (NVML error resolution)

Infrastructure & Networking Software And Drivers Mellanox OFED
, ,
1

Hello NVIDIA Support Team,

I am requesting assistance to restore the graphical desktop (GUI) output via HDMI on my NVIDIA DGX system. The system is currently accessible via SSH, but the physical display output is completely lost (shows “No Signal” / Check Device Power after the initial boot sequence).

Here is the exact sequence of events that led to this state:

  1. Initial Issue: My Docker containers (using NVIDIA runtime) failed to start with the following error: nvidia-container-cli: initialization error: nvml error: driver not loaded

  2. Failed Update Attempt: I attempted to resolve this by running sudo ubuntu-drivers autoinstall, but encountered unmet dependencies and conflicts specifically with libnvidia-common and libnvidia-common-580/590.

  3. Purge and Reinstall: To clear the conflict, I executed a deep purge: sudo apt-get purge '^nvidia-.*' '^libnvidia-.*' sudo apt-get autoremove --purge After clearing the packages, I successfully reinstalled the drivers and rebooted.

  4. Current state
    this is the nvdia-smi output:
    ±----------------------------------------------------------------------------------------+
    | NVIDIA-SMI 580.126.09 Driver Version: 580.126.09 CUDA Version: 13.0 |
    ±----------------------------------------±-----------------------±---------------------+
    | 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 NVIDIA GB10 Off | 0000000F:01:00.0 On | N/A |
    | N/A 35C P8 5W / N/A | Not Supported | 0% Default |
    | | | N/A |
    ±----------------------------------------±-----------------------±---------------------+

    ±----------------------------------------------------------------------------------------+
    | Processes: |
    | GPU GI CI PID Type Process name GPU Memory |
    | ID ID Usage |
    |=========================================================================================|
    | No running processes found |
    ±----------------------------------------------------------------------------------------+
    juandi@spark-9fd1:~$

    Also tried to access the GRUB by implementing these changes via ssh: Change GRUB_TIMEOUT_STYLE=hidden to GRUB_TIMEOUT_STYLE=menu

    Change GRUB_TIMEOUT=0 to GRUB_TIMEOUT=10

    But when booting it just shows nvidia logo, a bunch of lines and then a circling points before going blank again.