observed async checkpoint drain rates across 40 production runs; no correlation to model size
50–200 GB/sobserved
| Value kind | observed — Reported measurements, counts, and specifications keep the precision and scope stated by their source; an exact specification is not treated as a range. |
|---|---|
| Scope | VAST's global drain rate showed no clear correlation to model size across 40 runs on more than 1,024 GPUs; per-GPU checkpoint bandwidth fell as models grew. |
| As of | 2025 |
| Source | VAST Data (Optimizing Checkpoint Bandwidth) |
| Review | checking…review by 2027-04-20 · fast cadence |
| Recorded changes | last 2026-06-29 |
| Claim id | observed-async-checkpoint-drain-rates-across-40 |
Where the guide uses it
← Full numbers register — every date-stamped figure in the guide, with revision history.