Comprehensive Guide to Installing NVIDIA HPC SDK on Linux Arm Servers

The NVIDIA HPC SDK is a powerful toolset designed for high-performance computing (HPC) applications. It includes compilers, libraries, and tools that are optimized for NVIDIA’s architecture. This guide will walk you through the installation process for various Linux distributions on Arm servers, focusing on both CUDA version 12.3 and bundled versions with CUDA 12.3 and 11.8.

Installation Instructions for Linux Arm Server (CUDA 12.3)

Using Tar File

For users preferring a tarball installation, follow these steps:

1. Download the Tar File:

   wget https://developer.download.nvidia.com/hpc-sdk/24.1/nvhpc_2024_241_Linux_aarch64_cuda_12.3.tar.gz

2. Extract the Files:

   tar xpzf nvhpc_2024_241_Linux_aarch64_cuda_12.3.tar.gz

3. Run the Installer:

   nvhpc_2024_241_Linux_aarch64_cuda_12.3/install

Make sure to execute the install command with the necessary permissions, especially if you’re installing into directories like /opt.

Using DNF for RHEL/Rocky Linux

If you are on a RHEL or Rocky Linux system, you can utilize the dnf package manager:

1. Add the Repository:

   sudo dnf config-manager --add-repo https://developer.download.nvidia.com/hpc-sdk/rhel/nvhpc.repo

2. Install the SDK:

   sudo dnf install -y nvhpc-24.1

Using Yum for RHEL/Rocky Linux

For users who prefer yum, the process is similar:

More Read

Discover SyGra Studio: Your Gateway to Exceptional Creative Solutions

Optimizing Performance: Efficiently Scaling the Polars GPU Parquet Reader

How AI is Transforming Creativity: Inspiring Artists and Industrialists to Reimagine Their Crafts

Mastering Data Filtering: Overcoming Common Challenges in Data Management

Key Takeaways and Highlights from PyTorch Community Sessions

1. Add the Repository:

   sudo yum-config-manager --add-repo https://developer.download.nvidia.com/hpc-sdk/rhel/nvhpc.repo

2. Install the SDK:

   sudo yum install -y nvhpc-24.1

Using Zypper for SLES/SUSE

For SLES or openSUSE users, you can install using zypper:

1. Add the Repository:

   sudo zypper addrepo https://developer.download.nvidia.com/hpc-sdk/sles/nvhpc.repo

2. Install the SDK:

   sudo zypper install -y nvhpc-24.1

Using APT for Ubuntu

For Ubuntu installations, follow these steps:

1. Add the GPG Key:

   curl https://developer.download.nvidia.com/hpc-sdk/ubuntu/DEB-GPG-KEY-NVIDIA-HPC-SDK | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-hpcsdk-archive-keyring.gpg

2. Add the Repository:

   echo 'deb [signed-by=/usr/share/keyrings/nvidia-hpcsdk-archive-keyring.gpg] https://developer.download.nvidia.com/hpc-sdk/ubuntu/arm64 /' | sudo tee /etc/apt/sources.list.d/nvhpc.list

3. Update and Install:

   sudo apt-get update -y
   sudo apt-get install -y nvhpc-24-1

Installation Instructions for Linux Arm Server (CUDA 12.3 and 11.8)

If you need a version that includes both CUDA 12.3 and 11.8, follow similar steps as outlined above, adjusting your commands accordingly.

Using Tar File for Multi-CUDA Installation

1. Download the Multi-CUDA Tar File:

   wget https://developer.download.nvidia.com/hpc-sdk/24.1/nvhpc_2024_241_Linux_aarch64_cuda_multi.tar.gz

2. Extract the Files:

   tar xpzf nvhpc_2024_241_Linux_aarch64_cuda_multi.tar.gz

3. Run the Installer:

   nvhpc_2024_241_Linux_aarch64_cuda_multi/install

Using DNF for Multi-CUDA Installation on RHEL/Rocky

1. Add the Repository:

   sudo dnf config-manager --add-repo https://developer.download.nvidia.com/hpc-sdk/rhel/nvhpc.repo

2. Install the Multi-CUDA SDK:

   sudo dnf install -y nvhpc-cuda-multi-24.1

Using Yum for Multi-CUDA Installation on RHEL/Rocky

1. Add the Repository:

   sudo yum-config-manager --add-repo https://developer.download.nvidia.com/hpc-sdk/rhel/nvhpc.repo

2. Install the Multi-CUDA SDK:

   sudo yum install -y nvhpc-cuda-multi-24.1

Using Zypper for Multi-CUDA Installation on SLES/SUSE

1. Add the Repository:

   sudo zypper addrepo https://developer.download.nvidia.com/hpc-sdk/sles/nvhpc.repo

2. Install the Multi-CUDA SDK:

   sudo zypper install -y nvhpc-cuda-multi-24.1

Using APT for Multi-CUDA Installation on Ubuntu

1. Add the GPG Key:

   curl https://developer.download.nvidia.com/hpc-sdk/ubuntu/DEB-GPG-KEY-NVIDIA-HPC-SDK | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-hpcsdk-archive-keyring.gpg

2. Add the Repository:

   echo 'deb [signed-by=/usr/share/keyrings/nvidia-hpcsdk-archive-keyring.gpg] https://developer.download.nvidia.com/hpc-sdk/ubuntu/arm64 /' | sudo tee /etc/apt/sources.list.d/nvhpc.list

3. Update and Install:

   sudo apt-get update -y
   sudo apt-get install -y nvhpc-24-1-cuda-multi

Additional Resources

For additional information about the NVIDIA HPC SDK, including checksums for installers and patches, please refer to the official documentation. This guide provides a comprehensive overview of installation methods across various Linux distributions for both CUDA versions, ensuring you can easily set up your environment for high-performance computing tasks.

Source: Original Article