Skip to Content
GuidesОцінка робочої станції Ubuntu NVIDIA GPU

Оцінка робочої станції Ubuntu NVIDIA GPU

Цей посібник готує рецепт робочої станції Ubuntu 24.04 для NVIDIA tooling. Звичайний KVM test guest може не мати NVIDIA device, тому package і configuration checks не є physical GPU qualification.

Prompt

Create an Ubuntu 24.04 GNOME workstation for NVIDIA compute evaluation.
Include the gpu-nvidia, cuda, docker, Python, Git, SSH, and Jupyter features.
Do not choose “latest”; show the exact driver, CUDA toolkit, container runtime,
repositories, signing keys, and supported GPU architecture in the plan. Create
a locked-password user named mldev. Add package and service tests for the VM,
then list nvidia-smi, CUDA sample, container-GPU, suspend/resume, and workload
tests that must run on the target workstation.

Recipe shape

{
  "name": "ubuntu24-nvidia-workstation-evaluation",
  "base_image": "ubuntu-24.04",
  "hardware": {
    "platform": "pc",
    "architecture": "x86_64",
    "gpu": "nvidia",
    "min_cpu_cores": 8,
    "min_memory_gb": 32,
    "min_storage_gb": 128,
    "nic_count": 1
  },
  "os": {
    "features": [
      "desktop",
      "gpu-nvidia",
      "cuda",
      "docker",
      "python",
      "git",
      "ssh",
      "jupyter"
    ],
    "users": [
      {
        "username": "mldev",
        "groups": ["sudo", "docker"],
        "shell": "/bin/bash"
      }
    ],
    "desktop_settings": {
      "color_scheme": "prefer-dark",
      "power": {
        "idle_delay": 600,
        "sleep_inactive_ac_timeout": 0,
        "sleep_inactive_ac_type": "nothing"
      }
    }
  }
}

Hardware block , declared intent. Він не робить NVIDIA device visible для test VM. Перегляньте Docker-group privilege і не expose Jupyter без authentication і network policy.

Compatibility review

Перед build зафіксуйте:

  • exact GPU model і compute capability;
  • kernel, NVIDIA driver, CUDA toolkit і user-space library compatibility;
  • source repository і signing-key provenance;
  • Secure Boot/module-signing behavior, якщо used;
  • display-GPU versus compute-only expectations; та
  • license і redistribution constraints для кожного proprietary component.

Evidence split

Virtual build може довести package inventory, executables, configuration files, desktop launch і non-GPU services. Лише target hardware може довести:

  • nvidia-smi sees intended device without errors;
  • compiled CUDA sample produces expected result;
  • pinned container image runs with --gpus і sees device;
  • thermals, power limits, suspend/resume і reboot stable; та
  • real training або inference workload completes within memory і performance requirements.

Retain both evidence sets і label by environment. Do not report working GPU stack from package presence alone.

Оцінка образу ROS 2Організації