NVIDIA Mellanox LinkX 400G OSFP to QSFP112 4x100G Active Optical Cable

The NVIDIA Mellanox LinkX MFP7E20-N003 is a 400G OSFP to QSFP112 breakout active optical cable (AOC) designed to interconnect 400G OSFP-ported switches and routers with four independent 100G QSFP112-ported endpoints over a single fiber assembly. The cable employs an active optical architecture that integrates electrical-to-optical conversion circuitry within each connector housing, enabling reach and signal integrity performance that copper direct-attach cables cannot achieve at comparable distances. The assembly ships as a 3-meter length and is built around OM3/OM4-class multimode fiber, supporting 4x100G NRZ electrical interfaces on both the OSFP and the four QSFP112 breakout ends.

Targeted at high-density AI and machine learning fabric deployments, the MFP7E20-N003 addresses the structured cabling requirements of GPU-to-switch and switch-to-switch interconnects where rack separation exceeds the practical range of passive copper. The OSFP host-side connector supports the full 400G aggregate bandwidth expected by platforms such as the NVIDIA Quantum-2 and Spectrum-4 switch families, while the four QSFP112 fan-out connectors allow direct attachment to compute nodes, storage appliances, or downstream leaf switches operating at 100G per port. This breakout topology maximises port utilisation and reduces the total number of switch ports required in spine-leaf AI fabric designs.

As an NVIDIA Mellanox LinkX branded product, the cable is validated against NVIDIA's end-to-end ecosystem, enabling plug-and-play operation with NVIDIA NIC and switch hardware without additional optical module qualification steps. The active optical design eliminates the electromagnetic interference sensitivity and distance limitations inherent to passive copper breakout assemblies, making the MFP7E20-N003 well suited for structured cabling in new 400G and 800G AI supercomputing fabrics, cloud data centres, and high-performance computing installations where cable management, bend radius, and long-term thermal reliability are operational priorities.

Ideal for

• Connecting 400G OSFP switch ports to four independent 100G QSFP112 GPU server NICs in a spine-leaf AI training fabric
• Top-of-rack to end-of-row structured cabling runs in hyperscale data centres where copper direct-attach cables cannot span the required distance
• Breakout interconnects between NVIDIA Quantum-2 InfiniBand or Spectrum-4 Ethernet switches and 100G compute or storage nodes
• High-density GPU cluster pod cabling in 400G AI supercomputing installations requiring consistent signal integrity across all breakout lanes
• Switch-to-switch 400G uplink connections in multi-tier leaf-spine topologies with physical separation exceeding copper reach
• Structured pre-terminated fiber deployments in enterprise HPC environments where cable management, low weight, and small bend radius are required

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.