Per-GPU scale-out bandwidth on the NVIDIA Rubin platform — a per-GPU platform figure delivered as two ConnectX-9 SuperNICs per GPU (each an 800 Gb/s part, the family's first with single-port 800GbE); double the ConnectX-8 generation's 800 Gb/s per GPU
up to 1.6 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. |
|---|---|
| Caveat | 1.6 Tb/s is the Rubin per-GPU platform bandwidth (2x800G ConnectX-9), not a single-NIC port: the ConnectX-9 hardware manual lists 800 Gb/s SKUs (C9180 single-OSFP, 800GbE-capable). Rubin partner systems 2H 2026. |
| As of | 2026 |
| Source | NVIDIA Rubin platform blog + ConnectX-9 SuperNIC User Manual · Rubin platform blog and CES 2026 coverage: 1.6 Tb/s is the per-GPU scale-out figure, delivered as two 800 Gb/s ConnectX-9 SuperNICs per GPU. The ConnectX-9 hardware manual lists 800 Gb/s SKUs (C9180 single-OSFP, 800GbE-capable); NVIDIA's newsroom release names CX-9 but omits a bandwidth figure and the datasheet is cookie-walled. |
| Review | checking…review by 2027-03-06 · fast cadence |
| Recorded changes | last 2026-08-13 · 2 revisions tracked |
| Claim id | nvidia-connectx-9-supernic-bandwidth-rubin |
Where the guide uses it
← Full numbers register — every date-stamped figure in the guide, with revision history.