NVIDIA Mellanox NDR 400G OSFP to 2x200G Passive Copper Splitter DAC 1.5m

The Mellanox NDR 400G OSFP to 2x200G Passive Copper Splitter DAC (MCP7Y00-H001R30L) is a 1.5-metre passive Direct Attach Copper breakout cable designed to interconnect a single NDR 400G OSFP port on a switch or network adapter to two separate 200G OSFP or HDR/NDR endpoints. The cable splits one 400G electrical lane bundle into two independent 200G channels, enabling high-density, low-latency copper connectivity without the optical components or power overhead associated with active transceivers. It complies with the InfiniBand NDR signalling standard and supports 4x100G PAM4 electrical lanes per 400G side, making it fully compatible with NVIDIA Quantum-2 NDR switches and ConnectX-7 series adapters.

As a passive copper assembly, the MCP7Y00-H001R30L introduces no additional signal processing latency and draws no active power from the host ports, a key advantage in thermally dense GPU server racks where power budgets are tightly managed. The OSFP form factor on both the trunk and leaf ends supports the high-current requirements of 400G and 200G NDR electrical interfaces. The 1.5-metre reach is optimised for within-rack and adjacent top-of-rack switch connections, making it a practical solution for direct server-to-switch cabling in high-performance computing clusters.

This splitter DAC is particularly well suited to GPU-to-GPU and GPU-to-switch direct interconnect scenarios in AI training and HPC fabric deployments where bandwidth, density, and simplicity of cabling are paramount. By converting a single 400G uplink port into two 200G connections, network architects can double the number of connected endpoints per switch port, significantly improving fabric density and reducing the total number of switch ports required in large-scale GPU cluster builds.

Ideal for

• AI and deep learning training clusters requiring high-density, low-latency GPU-to-switch InfiniBand NDR interconnects within a single rack
• HPC compute nodes connecting dual-port 200G ConnectX-7 adapters to a single 400G NDR switch port to maximise switch port utilisation
• Top-of-rack NDR 400G switch deployments where passive breakout cabling eliminates the need for active optical modules and reduces per-port power draw
• Hyperscale data centre fabric spine-to-leaf or leaf-to-server cabling where short-reach copper DAC assemblies replace more expensive active optical solutions
• GPU-direct RDMA storage fabrics interconnecting NVMe-over-Fabrics storage nodes to NDR compute switches at 200G per storage initiator
• In-rack disaggregated computing architectures splitting a single high-speed fabric port to service two independent 200G accelerator nodes simultaneously

Technical specifications

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