🪴 Quartz 4.0

Search

SearchSearch

Installing Prerequisites

Aug 03, 20245 min read

Prerequisites

Warning

For Windows users, WSL2 is needed to run LEIP SDK. We currently do not support native Windows.

As a general supportability statement, any OS that has Docker installed with access to a CUDA-enabled GPU can utilize LEIP SDK.

For our testing, we have used:

  • Linux-based system (Ubuntu 20.04 or 22.04)
  • NVIDIA GPU (from CUDA 10 to CUDA 12).
  • Docker 1.19 or greater

Ubuntu

Windows (WSL2)

  • How to Install WSL2
    • If you’re running Windows 10 version 2004 and higher (Build 19041 and higher) or Windows 11, the easiest way to install WSL2 is by opening a PowerShell as administrator and type wsl --install. This will default to the Ubuntu distro.
    • You can change the WSL distro choice by typing wsl -l -o to view the list of available distros online, then typing wsl --install -d <Distribution Name> to install the chosen distro.
    • For the manual install refer to this page.

Step 1: Installing NVIDIA Drivers

Skip this section if...

You can skip this step if your environment has a CUDA-enabled GPU installed properly.

Simply type nvidia-smi into a terminal to verify. You should be able to see this output if successful.

Ubuntu (and other Linux-Based Distros)

  1. Based on your Linux distro, update your package index files. For Ubuntu, that would be sudo apt update
  2. Simply go to this link, select your architecture, distro, the latest CUDA version, and select “deb (network).”
  3. Follow the “Driver Installer” instructions.

Windows (WSL2)

  1. Go to this link.
  2. Fill out your OS and GPU details, click search.
  3. Click download.
  4. Run the .exe and install

Successful Output (Step 1)

Note

You should also take note of the CUDA Version on the top right corner of the output. This version does not represent your CUDA version installed on the system, but is listed to state what CUDA version your drivers can support up to.

To verify a successful installation, simply type nvidia-smi into your terminal. You should see a similar output below.

$ nvidia-smi
Tue Jun 18 09:18:44 2024
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 555.52.01              Driver Version: 555.99         CUDA Version: 12.5     |
|-----------------------------------------+------------------------+----------------------+
| 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 GeForce RTX 3090        On  |   00000000:0A:00.0  On |                  N/A |
|  0%   40C    P8             28W /  350W |    2424MiB /  24576MiB |      4%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A        29      G   /Xwayland                                   N/A      |
+-------

Step 2: Installing Docker

Skip this section if...

If you are able to type docker in the terminal and return back this output, you have Docker installed.

Ubuntu

  1. Refer to this link (Install using the apt repository) to install Docker via apt.
  2. Enable the GPU for usage inside Docker. To do this, install nvidia-container-runtime by following this specific section (“Installing with Apt”).
  3. Now, we need to set Docker to leverage the NVIDIA runtime. Following only the first two steps in this specific section (“Configuring Docker).

Other Linux Distros

It is possible to run containers with other Docker alternatives, such as containerd, but installation is currently not supported by us.

Windows (WSL2)

  1. The easiest way to accomplish this is to install Docker Desktop, which will enable Docker automatically with your NVIDIA GPU.

Successful Output (Step 2)

Tip

If you cannot use docker without sudo, follow these instructions (Manage Docker as a non-root user) and restart your computer to allow your user to use docker without sudo privileges.

Type docker in your terminal and verify if you can see a similar output as the one below.

$ docker

Usage:  docker [OPTIONS] COMMAND

A self-sufficient runtime for containers

Common Commands:
	...

Management Commands:
	...

Swarm Commands:
  swarm       Manage Swarm

Commands:
	...

Global Options:
	...

Run 'docker COMMAND --help' for more information on a command.[[]]

For more help on how to use Docker, head to https://docs.docker.com/go/guides/

Step 3: Enable GPU Usage in Docker

To enable Docker to use NVIDIA GPUs, we need to install nvidia-container-toolkit.

  1. Visit the “Installation” section of this link
  2. Follow the “Configuration” section in the same link. You do not have to follow the “Rootless Mode,” only the first two steps.

Step 4: Verify GPU Usage by Docker

The easiest test to verify if you have GPU access in a Docker container is by running the command below:

docker run --gpus all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -gpu -benchmark

Successful Output (Step 4)

Run "nbody -benchmark [-numbodies=<numBodies>]" to measure performance.
        -fullscreen       (run n-body simulation in fullscreen mode)
        -fp64             (use double precision floating point values for simulation)
        -hostmem          (stores simulation data in host memory)
        -benchmark        (run benchmark to measure performance)
        -numbodies=<N>    (number of bodies (>= 1) to run in simulation)
        -device=<d>       (where d=0,1,2.... for the CUDA device to use)
        -numdevices=<i>   (where i=(number of CUDA devices > 0) to use for simulation)
        -compare          (compares simulation results running once on the default GPU and once on the CPU)
        -cpu              (run n-body simulation on the CPU)
        -tipsy=<file.bin> (load a tipsy model file for simulation)

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.

> Windowed mode
> Simulation data stored in video memory
> Single precision floating point simulation
> 1 Devices used for simulation
GPU Device 0: "Turing" with compute capability 7.5

> Compute 7.5 CUDA device: [NVIDIA GeForce RTX 2060]
30720 bodies, total time for 10 iterations: 69.751 ms
= 135.299 billion interactions per second
= 2705.976 single-precision GFLOP/s at 20 flops per interaction

Step 5: Install LEIP

Refer to Installing LEIP SDK.

Graph View

Backlinks