quickstart →

x64arm64 at 2 vCPU4 vCPU8 vCPU

docker pullsecondslower is betterarm64 · 4 vCPU

RunsOn tmpfs and Namespace Apple silicon tie for the fastest docker pull.

RunsOn tmpfs to Ubicloud: 4.0× apart.

  1. RunsOntmpfs median of 2 machines 4.5s
  2. NamespaceApple silicon 4x16 · 16 GB 4.7s
  3. RunsOn median of 3 machines 4.9s
  4. Avrea 16 GB 5.5s
  5. Warpbuild 16 GB 8.7s
  6. Blacksmith 16 GB 12.9s
  7. Namespace median of 2 machines 14.1s
  8. GitHub 16 GB 15.6s
  9. Ubicloud standard · 16 GB 18.1s
  10. WarpbuildApple silicon 6 vCPU · xfast (Apple M4 Pro) · 14 GB 8.6s not ranked
  11. AWS CodeBuild no arm64 runner
  12. boat no arm64 runner

The same image pulled on every runner, registry to disk: the download, then the layers unpacked, so the disk and the CPU count too.

Full network benchmark

The network benchmark for arm64, 4 vCPU →