GUI lost on setting up GPU computing

The GUI for the screen connected to my server (Dell T550) no longer works after setting up GPU computing for a computational chemistry application (Maestro, schrodinger.com). I have an Nvidia Ampere A2 graphics card. I am currently using Rocky 9.3 as operating system (was the same with 9.2), GPU computing all seems to work fine. Apparently I shouldn’t have lost the GUI. Anyone with a similar experience? Software provider indicated this shouldn’t happen. It only happens when Rocky OS boots up.
Thanks
Paul

That one does not look like a “graphics card” as in a thing you connect a display to.
Yes, it is a “GPU”, but more “an accelerator”.

I have installed Maestro into several systems, but never done anything additional for “GPU computing”. Not for Maestro. Furthermore, my systems have NVidia Quadro or GeForce graphics card, where same card does both graphics for display and GPGPU.

If the whatever you did did reconfigure your desktop session to use A2 instead of whatever produces display in your system, then obviously that affects the GUI.

Hi - thanks

I followed their instructions for the set up to use Desmond on Maestro. Support

The display had worked previous to this.

The NVIDIA A2 GPU is not a supported by schrodinger, but Desmond runs fine with it.

Paul

Did you install NVidia driver and CUDA “for this”, i.e. did you run display without NVidia’s proprietary driver previously?

How did you install the NVidia driver?

It is possible that the installation of NVidia driver configured the GUI to use the A2, which is obviously wrong since you can’t connect displays to the A2.


On system that does not have NVidia drivers:

$ ls /etc/X11
applnk  fontpath.d  mwm  xinit  Xmodmap  xorg.conf.d  Xresources  Xsession.d
$ ls /etc/X11/xorg.conf.d
00-keyboard.conf

(These are at least for X11. I don’t know how Wayland is configured.)

Hi, thanks,

I did install the Nvidia driver and CUDA. I had run the display without the driver previously.

This is the output of

$nvidia-smi

NVIDIA-SMI 545.23.08 Driver Version: 545.23.08 CUDA Version: 12.3 |

-----------------------------------------±---------------------±---------------------+

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 A2 Off | 00000000:98:00.0 Off | Off |

0% 58C P0 22W / 60W | 538MiB / 16380MiB | 0% Default |

N/A

$ ls /etc/X11

applnk fontpath.d xinit Xmodmap xorg.conf.d Xresources Xsession.d

ls /etc/X11/xorg.conf.d

00-keyboard.conf 10-nvidia.conf

The Rocky does have nouveau driver for NVidia, but since your nvidia-smi does not show more than A2, it is more likely that your system has some other GPU, like IGP in CPU.

I would start by

mv /etc/X11/xorg.conf.d/10-nvidia.conf /etc/X11/xorg.conf.d/10-nvidia.conf-off
systemctl reboot

That might not be sufficient, since driver installation tends to modify the kernel command-line too,
and since GUI does do some auto-detecting when no explicit config exist and now there is additional driver, it might still default to A2 as output device. As said, I’ve never had the need to create explicit config for case like yours, so don’t know what it requires.