NVLink per-GPU bandwidth: Gen5 (Blackwell) → Gen6 (Vera Rubin, shipping spec); Gen4 (Hopper) was 900 GB/s
1.8 → 3.0 TB/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 | Named NVIDIA generation/platform, bidirectional aggregate convention; not delivered collective bandwidth, cache coherence or arbitrary mixed-platform performance. NVIDIA announced 3.6 TB/s per GPU and 260 TB/s per NVL72 in its Jan–Mar 2026 technical blogs and at Hot Chips (Aug 2026); its product pages and the first OEM shipping release (Sep 2026) list 3 TB/s and 216 TB/s. The NVLink Fusion page still quotes 3.6 TB/s per XPU. |
| As of | 2026-09 |
| Source | NVIDIA Vera Rubin NVL72 and NVLink product pages (checked 2026-09-29); Supermicro Vera Rubin NVL72 shipping release, 2026-09-23. |
| Review | checking…review by 2027-01-31 · standard cadence |
| Recorded changes | last 2026-10-05 · 6 revisions tracked |
| Claim id | nvlink-per-gpu-bandwidth-gen5-blackwell-gen6-2 |
Where the guide uses it
← Full numbers register — every date-stamped figure in the guide, with revision history.