SuperNIC delivering up to 800 Gb/s for InfiniBand or Ethernet GPU-to-GPU networking.
The NVIDIA ConnectX-8 SuperNIC delivers up to 800 Gb/s of network bandwidth per adapter for GPU-to-GPU communication in large AI clusters. It supports both InfiniBand (XDR) and Ethernet, and is the standard compute-fabric NIC in Blackwell Ultra systems.
A PCIe Gen6 host interface with up to 48 lanes and an integrated PCIe switch let one ConnectX-8 serve GPUs, CPUs and storage without extra switch chips.
Highest-bandwidth NIC for AI compute fabrics.
XDR 800G InfiniBand by default, or 2× 400GbE Ethernet.
Up to 48 lanes, backward compatible with Gen5 and Gen4.
High-bandwidth, low-latency transfers tuned for generative AI.
PCIe cards, OCP 3.0 and mezzanine options for different servers.
| Maximum bandwidth | 800 Gb/s |
|---|---|
| InfiniBand | XDR 800 / 400 / 200 / 100 Gb/s |
| Ethernet | 2× 400GbE; 400 / 200 / 100 / 50 / 25 GbE |
| InfiniBand compliance | IBTA v1.7 |
| Host interface | PCIe Gen6, up to 48 lanes (Gen5 / Gen4 compatible) |
|---|---|
| Form factors | PCIe HHHL 1× OSFP; PCIe HHHL 2× QSFP112; dual ConnectX-8 mezzanine; OCP 3.0 TSFF 1× OSFP |
End-to-end 800 Gb/s InfiniBand platform for the largest AI training clusters.
Ethernet platform purpose-built for AI, from single-cluster fabrics to multi-site AI factories with Spectrum-XGS.
Data processing unit that offloads networking, storage and security from host CPUs using NVIDIA DOCA.
Tell us your workload and timeline for the NVIDIA ConnectX-8 SuperNIC. We reply within one business day with a configuration, pricing and lead time.