NVIDIA reference read target for vision training; no single per-GPU floor — B300 Standard is 0.31 GB/s/GPU and Enhanced ~0.98; 4–10 GB/s/GPU for multimodal and checkpoint-heavy runs

Name: NVIDIA reference read target for vision training; no single per-GPU floor — B300 Standard is 0.31 GB/s/GPU and Enhanced ~0.98; 4–10 GB/s/GPU for multimodal and checkpoint-heavy runs
License: https://aidatacenterguide.com/terms#intellectual-property

~4 GB/s/GPUobserved

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	Published reference-architecture guidance, not a measured requirement. NVIDIA's own tiers span 0.31–5 GB/s/GPU depending on the tier and reference platform, and NVIDIA states the tiers assume a mix of workloads. Derive the target from the loader's measured consumption rate; use these as named comparison cases.
As of	2025
Source	NVIDIA DGX SuperPOD storage RA; DDN AI400X2-Turbo baseline
Review	checking…review by 2027-03-01 · standard cadence
Recorded changes	last 2026-06-29
Claim id	nvidia-reference-read-target-for-vision

Where the guide uses it

9.1 Storage in the AI Lifecycle: Why It Determines GPU Efficiency

← Full numbers register — every date-stamped figure in the guide, with revision history.