Skip to content

Latest commit

 

History

History
127 lines (78 loc) · 3.44 KB

virtualgl-installation.md

File metadata and controls

127 lines (78 loc) · 3.44 KB
Raw
layout title nav_order description permalink
default
Install VirtualGL+TurboVNC to Enable 3D Acceleration for Remote Access
110
Guide to install VirtualGL+TurboVNC for new servers.
/virtualgl-installation

⛔️️ WARNING: This will require a server reboot and will break the Nvidia driver for all users if not done properly. Make sure you know what you are doing. Otherwise, don't do anything!

Background Knowledge

Understanding what X server is

Understanding what VirtualGL is

Part A: TurboVNC + VirtualGL Installation Guide

Uninstall Current VNC (TurboVNC for example):

dpkg -r turbovnc

Reference: https://virtualgl.org/vgldoc/2_1_3/#hd005006

Uninstall Nvidia-*:

sudo apt-get remove --purge '^nvidia-.*'
  • nvidia-smi to check whether nvidia driver has been uninstalled

Install VirtualGL:

Follow the Reference Link

Install Nvidia-driver

Follow the Reference Video

Disable "HardDPMS" in xorg.conf:

sudo nvidia-xconfig -a --allow-empty-initial-configuration

sudo vim /etc/X11/xorg.conf

Delete all lines about "DPMS"

Add Option "HardDPMS" "false" to the bottom of all monitor, device and screen sections (within those section blocks)

Reference: https://virtualgl.org/Documentation/HeadlessNV

VirtualGL Configuration:

Follow the Reference Link

Tips:
  • Make sure you successfully shut down the display manager of your computer
  • In option 1 (Configure server for use with VirtualGL (GLX + EGL back ends)) : Select "No" for all those three questions

TurboVNC Installation:

dpkg -i turbovnc*.deb

Reference: https://virtualgl.org/vgldoc/2_1_3/#hd005001

Final Check:

Open one terminal:

watch -n 0.2 nvidia-smi

Open another terminal:

vglrun glxgears

Make sure a new process using GPU called glxgears appear in the first terminal.

You can also manually starta a X-server by:

sudo Xorg -noreset +extension GLX +extension RANDR +extension RENDER -logfile ./xdummy.log -config ./xorg.conf :20 &

This starts the X server at display :20.

Then, you can run vglrun -d :20 glxgears to run the program at the newly started X server at :20.

Part B: GPU Access on a remote Linux server in ThreeDWorld

Install TDW on a remote Linux server

Comments:

  • For /etc/X11/xorg.conf, no need to remove ServerLayout and Screen section
  • No need to run x-server by sudo nohup Xorg :<Display server name> -config /etc/X11/xorg.conf &

TDW Run:

  • Download the latest build of TDW and extract the zip file into TDW

  • Create Python script. Make sure to set launch_build=False to Controller

  • In one shell window, run the controller:

    python3 my_controller.py

  • In the second shell window, cd to TDW, the location of the build (TDW.x86.64)

    vglrun ./TDW.x86_64 -port=1071