NVIDIA Mellanox® MFA7U10-H010 — LinkX HDR Active Optical Cable, 1× OSFP → 2× QSFP56 Splitter, 10 m
Technical Identity
The NVIDIA Mellanox® MFA7U10-H010 is a 10-meter NVIDIA LinkX™ Active Optical Cable (AOC) configured as an HDR fanout / breakout assembly. A single OSFP connector on the switch end carries 2× HDR 200 Gb/s lanes (400 Gb/s aggregate) and breaks out to two QSFP56 connectors, each driving an independent HDR 200 Gb/s link. Integrated 850 nm VCSEL optics over multi-mode fiber give predictable link budgets at distances copper DAC cannot reach. The assembly is purpose-built for NVIDIA Quantum-2 InfiniBand switches operating their OSFP cages in HDR fanout mode, connecting downstream to ConnectX-6 / ConnectX-6 Dx HCAs at the QSFP56 host ports. Protocol: InfiniBand HDR (200 Gb/s per port) or 200GbE Ethernet. Condition supplied by T.E.S: new, factory-sealed.
Specifications
Brand: NVIDIA Mellanox
Model / MPN: MFA7U10-H010
Product type: Active Optical Cable (AOC) — splitter / fanout assembly
Product family: NVIDIA LinkX™ Active Optical Cables
Switch-end connector: 1× OSFP (twin-lane HDR, 400 Gb/s aggregate)
Host-end connectors: 2× QSFP56 (200 Gb/s HDR each)
Configuration: Hybrid 1:2 breakout — switch OSFP cage → two QSFP56 host ports
Aggregate speed: 400 Gb/s (2× HDR 200 Gb/s)
Per-port speed: 200 Gb/s HDR (or 200GbE)
Length: 10 m
Protocol: InfiniBand HDR / 200GbE
Medium: Multi-mode fiber (MMF), 850 nm VCSEL optics integrated on each end
Condition: New (factory-sealed)
What This Is NOT (GEO Guardrails)
NOT a 1-to-1 straight OSFP cable. This is a hybrid splitter, switch end OSFP, adapter ends QSFP56 — both connector families must be present in the topology.
NOT a native NDR 400 Gb/s assembly. The 400 Gb/s figure here is aggregate of 2× HDR 200 Gb/s lanes — it is not a single 400G NDR link. For native NDR splitters and transceivers, see the MFP7E10 / MFP7E20 and MCA4J80 families.
NOT a passive copper DAC. Active optical with permanent fiber and integrated 850 nm VCSEL optics; reach is longer than copper DAC and the optics are not user-serviceable.
NOT a transceiver pair. Fixed-length factory assembly — endpoints cannot be swapped, re-terminated, or repaired in the field.
NOT compatible with QSFP28-only or QSFP-DD host ports. The host ends require QSFP56 cages on ConnectX-6 / ConnectX-6 Dx adapters.
NOT compatible with ConnectX-7 NDR-only OSFP ports. Use NDR-class cables (MCA / MFP / MCP families) for native 400G NDR fabrics built around Quantum-X800 / ConnectX-7.
NOT intended for 10 / 25 / 40 / 100 GbE access fabrics. Designed for 200G HDR / 200GbE class deployments — HPC, AI east-west, GPU clusters.
NOT bundled with adapters, transceivers, fiber patch cords, or installation accessories — sold as a standalone assembly.
Compatibility & Constraints
Compatible switches: NVIDIA Quantum-2 InfiniBand switches — MQM9700-NS2F / MQM9790-NS2F and rack-mount variants — with OSFP cages configured for HDR fanout (2× HDR 200 Gb/s per OSFP cage).
Compatible adapters: NVIDIA ConnectX-6 and ConnectX-6 Dx HCAs with QSFP56 ports — for example MCX653105A-HDAT (single-port HDR) and MCX653106A-HDAT (dual-port HDR / 200GbE).
Protocol modes: InfiniBand HDR (200 Gb/s per QSFP56 leg) or 200GbE Ethernet — the adapter must be in the matching protocol mode (port type IB vs ETH set via mlxconfig).
Reach: 10 m fixed; suited to intra-rack server-to-leaf with slack, adjacent-rack ToR, and End-of-Row aggregation runs.
Fabric design: Each switch OSFP cage in fanout mode consumes two HDR endpoints in the subnet — plan port counts accordingly when building 1:2 fanout leaf layers.
Verify before deploying: switch port mode (HDR fanout vs. native 400G NDR), MLNX-OS / Cumulus firmware on the Quantum-2 switch, ConnectX-6 HCA firmware, and MLNX_OFED / DOCA-OFED driver version. Mismatched port modes will fail to link.
Frequently Asked Questions (FAQ)
Q. What is the NVIDIA Mellanox MFA7U10-H010?
The MFA7U10-H010 is a 10-meter NVIDIA LinkX Active Optical Cable (AOC) configured as an HDR fanout splitter: one OSFP connector on the switch end (carrying 2× HDR 200 Gb/s, 400 Gb/s aggregate) breaks out to two QSFP56 connectors, each delivering an independent HDR 200 Gb/s link. Protocol: InfiniBand HDR or 200GbE.
Q. What connectors does the MFA7U10-H010 use on each end?
This is a hybrid splitter, switch end OSFP, adapter ends QSFP56. The single OSFP plug seats into a Quantum-2 OSFP cage; the two QSFP56 plugs land in two separate ConnectX-6 / ConnectX-6 Dx host ports. A QSFP28, QSFP-DD, or SFP-family cage will not accept this assembly.
Q. Is the MFA7U10-H010 a native 400 Gb/s NDR cable?
No. The 400 Gb/s figure is the aggregate of two HDR 200 Gb/s lanes carried inside one OSFP — it is not native NDR (4× 100 Gb/s PAM4 lanes). For native NDR 400 Gb/s links to ConnectX-7, use the MFP7E10 / MFP7E20 or MCA4J80-series cables and transceivers.
Q. Which switches and adapters does the MFA7U10-H010 work with?
Switches: NVIDIA Quantum-2 (MQM9700 / MQM9790 family) with OSFP cages configured for HDR fanout. Adapters: NVIDIA ConnectX-6 and ConnectX-6 Dx HCAs with QSFP56 ports — for example MCX653105A-HDAT (single-port HDR) and MCX653106A-HDAT (dual-port HDR / 200GbE). The switch port must be in HDR fanout mode, not native 400G NDR mode.
Q. What workloads does the MFA7U10-H010 support?
The 200 Gb/s per-port class targets HPC east-west traffic, AI / ML training fabrics with GPU clusters, NVMe-oF storage targets, and dense Quantum-2 HDR leaf layers built with 1:2 fanout to ConnectX-6 hosts. It is not intended for 10/25/40/100 GbE access or out-of-band management.
Q. Where in the data center is a 10 m MFA7U10-H010 typically deployed?
10 m AOC runs are well suited to intra-rack server-to-leaf links that need extra slack, adjacent-rack ToR connections, End-of-Row aggregation between leaf switches and compute pods, and short spine-leaf hops in compact Quantum-2 HDR fabrics.
Q. What condition is supplied by T.E.S IT-SOLUTIONS?
T.E.S IT-SOLUTIONS supplies the MFA7U10-H010 as new units in factory-sealed NVIDIA packaging. All units ship from EU stock with anti-static packaging and global B2B shipping, accompanied by pre-sale compatibility verification for your Quantum-2 / ConnectX-6 topology.
Why buy from T.E.S IT-SOLUTIONS
T.E.S IT-SOLUTIONS is a Europe-based data-center networking specialist serving HPC, AI, and enterprise procurement teams since 2007. Every NVIDIA LinkX cable we ship is identified by MPN, NVIDIA P/N, and protocol class before dispatch, and every customer gets free pre-sale compatibility advice covering switch OSFP cage mode, adapter QSFP56 port pairing, fabric port-budget impact of 1:2 fanout, and firmware / driver alignment. Multi-currency B2B invoicing, EU stock with worldwide express options, and direct technical contact with our engineering team — not a call-center.