This repo is a clone based on Plumbee's original code for the older g2 Nvidia GRID (Cloud Graphics) AWS Ec2 instances. Shout out to pigiuz for his original demo using the g2 instances located here: https://github.com/plumbee/nvidia-hw-accelerated-box
This is intended to be a demo setup to run hardware accelerated content on AWS g3 (Nvidia Tesla M60) GPU Hardware accelerated instances https://aws.amazon.com/blogs/aws/new-next-generation-gpu-powered-ec2-instances-g3/
These scripts must be run on a M60 instances running Ubuntu 16.04 (HVM).
Among the other packages these scripts will install
- nvidia drivers v384.81 http://www.nvidia.com/download/driverResults.aspx/124722/en-us
- VirtualGL v2.5.2 http://www.virtualgl.org/
- TurboVNC http://www.turbovnc.org/
- docker https://www.docker.com/
- nvidia-docker https://github.com/NVIDIA/nvidia-docker
- nvidia-docker-compose https://github.com/eywalker/nvidia-docker-compose
- a barebone Mate desktop https://mate-desktop.org/
SSH into your EC2 box, once in just run the following
git clone https://github.com/rybruscoe/revup-nvidia-gpu-renderbox-linux
cd revup-nvidia-gpu-renderbox-linux
sudo su
source setup.shKeep in mind that:
- The installation will reboot the machine once done.
- At the startup the VNC server will start on DISPLAY :1.
VNC password is dynamically set to the first 8 characters of your EC2 instance id.
VNC starts on DISPLAY :1, VirtualGL content will render offscreen and the resulting image will be copied to DISPLAY :1.
VNC receives connections from port 5900 + Display. Given that DISPLAY is :1 the port will be 5901.
VirtualGL provides a 'vglrun' executable which enables (m)any GUI application to access the GPU when needed. The command below will run glxgears with hardware acceleration
vglrun glxgearsYes, it's possible, check this out https://github.com/plumbee/nvidia-virtualgl