Skip to content

v0.3.59

Choose a tag to compare

View all tags
@github-actions github-actions released this 14 Oct 07:00
· 53 commits to main since this release
v0.3.59
28444ae
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What changes for you

  • Model suggestions know about 40 more GPUs. grid device-info now reports memory bandwidth for the RTX 50
    series and its laptop parts, RTX PRO 4000/4500/5000 Blackwell, RTX 2000/4000/4500/5000/5880 Ada, A2000, A4000,
    A5000, B200, DGX Spark GB10, Tesla P100/P40, Quadro P4000 and RTX 6000/8000, and TITAN RTX/Xp/X. These cards used
    to fall back to a generic default. The list was checked against local.sybilsolutions.ai/hardware and the GPUs
    vast.ai rents out, using the vendors' datasheets where the two disagree. (0099133, 77e895d)
  • Ti, SUPER and laptop cards get their own number instead of their base card's. Affected: 3060 Ti, 3070 Ti,
    3080 Ti, 4060 Ti, 4070 Ti SUPER, 4080 SUPER, 2060 SUPER, 2080 SUPER, 3060 Laptop. (0099133, 77e895d)
  • Wrong numbers fixed: RTX 4070 Ti 672 → 504 GB/s, RTX A4000 696 → 448, H100 PCIe 3350 → 2000, H100 NVL 3350 →
    3900, A100 80GB PCIe 2039 → 1935, Apple M3 Ultra 800 → 819. (0099133, 77e895d)
  • A binned Apple Max is told apart from the full one by its GPU core count: M3 Max 30-core 300 GB/s, M4 Max
    32-core 410, M5 Max 32-core 460. These used to read the full chip's 400 / 546 / 614. (0099133)

Full Changelog: v0.3.58...v0.3.59

Assets 6
Loading