HW & SW Requirements#

This part describes the minimal hardware and software requirements needed for the successful platform installation.

Note

Please contact your Intel® Geti™ account representative or technical support personnel if you have any questions.

HW Requirements#

The platform can be installed on a machine with the following minimum hardware:

  • CPU for workstations: Intel® Core™ i7, Intel® Core™ i9 or Intel® Xeon® scalable processors family capable of running 20 concurrent threads (in case of using the default K3s) or 48 concurrent threads (in case of using pre-installed K8s).

Note

From Intel® Core™ family, we recommend the following CPU series:

  • 13th gen (Raptor Lake): Intel® Core™ i7 13700 series and Intel® Core™ i9 13900 series

  • 12th gen (Alder Lake): Intel® Core™ i9 12900 series

  • CPU for cloud deployments: CPUs capable of running min. 24 concurrent threads for K3s or min. 48 concurrent threads for K8s (so for example, on AWS EC2 instances, this requirement would be translated to min. 24 vCPUs for K3s or min. 48 vCPUs for K8s).

  • GPU: min. one, two, or max. three NVIDIA GPUs with min. 16GB of memory (e.g. RTX 4080, RTX 3090, RTX 6000, RTX 8000, Tesla A100, Tesla V100, Tesla P100, or Tesla T4; other NVIDIA GPUs in a similar series are likely also compatible if they meet minimum memory requirements; however, the full range of devices is not fully tested and not specifically supported); we recommend 24GB of memory for more stable training & optimization.

  • Memory: min. 64 GB RAM (128 GB recommended) per GPU

  • Disk Space: min. 1 TB (2 TB recommended) available space on the root partition

Installation of the platform on a multi-node configuration of Kubernetes cluster is not supported.

Note

Use the workstation where the Intel® Geti™ platform is installed and hosted solely for the Intel® Geti™ platform. Do not run any other software in parallel on that same machine. To learn more, go to the Best Practices page.

Warning

Contact your business representative before changing the hardware configuration after installation. The license key is bound to hardware - changing the configuration will make the license key invalid.

SW Requirements#

The platform can be installed on Ubuntu 20.04 LTS or Ubuntu 22.04 LTS.

Warning

It is strongly advised to install the Intel® Geti™ platform on a fresh OS.

  1. The supported OS on the machine where the installer is run is Ubuntu 20.04 LTS or Ubuntu 22.04 LTS. If, when running platform_installer, you get a message “Error loading Python lib”, it means that you are trying to run the installer on an older, unsupported system.

  2. Ensure the Secure Boot option is disabled in BIOS. Otherwise, the license component will not be able to identify the machine and you will not be able to apply for the license.

  3. Ensure that /dev/mem is accessible by the root user. You can check it by running sudo head /dev/mem and observing if the command returns permission errors.

  4. The user’s account used to install the platform must be included in the sudoers lists. Ensure it is present in /etc/sudoers.

  5. The machine, on which you are running the installer, must have access to the Internet in order to download Docker images.

  6. A dedicated folder to store the platform artifacts must be available.

  7. Ensure that the entire directory path, starting from the root to the location where you plan to extract the installer and from which you plan to run the platform installation, does not contain spaces in any of its names. This is due to installer limitations that can cause errors if spaces are present in any part of the directory path.

  8. All GPU cards installed on the machine will be used to run training jobs. Ensure all of them are accelerators capable of running CUDA based code.

  9. GPU NVIDIA drivers must be installed on the machine. The recommended version is 525 (nvidia-driver-525). For NUC installation the recommended version is also 525.

  10. If curl is installed on the machine, ensure it can access the Internet when run as sudo.

  11. Geti installer installs K3S. However, if you choose to proceed with the installation on an existing Kubernetes cluster, the following requirements must be met:

  • Kubernetes cluster configuration:

    • Kubernetes version: 1.25.x.

    • Installation method: kubeadm, setup guide can be found here.

    • Container runtime: containerd version 1.6.8 or higher.

    • Nodes number: 1.

  • The metrics-server configuration:

    • Version: v0.6.x.

    • The component metrics-server should have the label k8s-app=metrics-server.

    • Installation guide can be found here.

  • NGINX Ingress Controller v1.8.1 must be deployed on the cluster:

    • To ensure correct installation, please use our deploy.yaml file, which can be found in the package under platform/installer/nginx_ingress/deploy.yaml. Please note that this file includes some configuration changes relative to the original file.

  • The NVIDIA GPU device plugin must be installed with the following components:

    • nvidia-container-toolkit, which can be installed by following the instructions available here.

    • containerd must be configured using the instructions provided here.

    • nvidia-device-plugin must be installed according to the instructions provided here.

  1. For the best platform performance and stability, it is advised to add the following entries to /etc/security/limits.conf:

    root - nofile 65536 * - nofile 65536

  2. In case of installation/upgrade on existing Kubernetes cluster, kubectl must be available on the machine from which the installer is executed.

Cloud Deployment Requirements#

The Intel® Geti™ platform needs a static IP address to work and cloud providers offer different means to ensure that. For example, in AWS this can be achieved by using Elastic IP address when creating a VM instance (see how this can be done in AWS docs).

If for any reason the IP address has been changed, please follow the instructions in the Changing platform’s IP address section of the installation guide.

Note

Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.