NVIDIA Mellanox 980-9I510-00NS00 Network Device | Product Launch and Key Feature Overview

NVIDIA Mellanox 980-9I510-00NS00 Network Device | Product Launch and Key Feature Overview

As AI training clusters and HPC workloads continue to push east-west traffic demands to new extremes, data center interconnect is accelerating from 400G toward 800G. The NVIDIA Mellanox 980-9I510-00NS00 — a twin-port OSFP 800G multimode transceiver — provides network engineers and architects with a validated, high-density path for scaling GPU fabric and InfiniBand infrastructure without ripping out existing cabling.

Why This Product Matters Now

Modern GPU clusters no longer simply move packets — they coordinate distributed compute where microseconds determine job completion time. Legacy short-reach optics often become the bottleneck when scaling to hundreds of nodes with NDR 400G per port. The 980-9I510-00NS00 addresses this by packing two 400G NDR channels into a single OSFP module, doubling per-slot density while preserving the 50m multimode reach that many existing AI data halls are already cabled for.

Key Technical Features

• 800G Twin-Port OSFP: Eight channels of 100G-PAM4 electrical modulation feed two 4-channel optical interfaces, enabling 2×400G straight connections or 2×200G breakout via splitter fiber .
• 50m Multimode Reach: 850nm VCSEL transmitters support up to 30m over OM3 and 50m over OM4 fiber, with dual MPO-12/APC connectors .
• Thermal Flexibility: Finned-top variants target air-cooled switches, while flat-top versions suit DGX-H100 or liquid-cooled deployments .
• Power and Compliance: Maximum 17W power draw with a 1.5W low-power sleep mode; compliant with OSFP MSA and CMIS 4.0, hot-pluggable and RoHS compliant .
• NVIDIA Validation: Listed in NVIDIA documentation as the 50m multimode SKU for NDR multi-mode switch-side connectivity .

For teams evaluating the 980-9I510-00NS00 network product, the specifications align closely with NDR InfiniBand and 800G Ethernet breakout architectures. Engineers comparing the 980-9I510-00NS00 datasheet against competing modules will find the twin-port design particularly relevant for leaf-spine fabrics where port count and cable management dominate rack-scale decisions. The 980-9I510-00NS00 data center high-speed networking role is further reinforced by its compatibility with the QM97XX NDR switch family, where it serves as the designated 50m multimode optic .

Deployment Scenarios

The module fits naturally into AI factory back-end fabrics, where GPU servers connect to NDR leaf switches over OM4 multimode trunks. Its dual-MPO interface also supports 2×400G breakout to next-generation compute nodes, giving IT managers a migration lever between 400G and 800G port economics. Third-party compatible alternatives exist for budget-sensitive pilots, but NVIDIA-validated units remain the reference for production fabric certification.

Summary

The NVIDIA Mellanox 980-9I510-00NS00 is not a general-purpose networking appliance — it is a purpose-built optical building block for 800G-era AI and HPC fabrics. By combining twin 400G NDR ports, 50m multimode reach, and NVIDIA switch-level validation in a single OSFP package, it removes a key physical-layer constraint from high-density data center high-speed networking designs. Architects specifying the next leaf-spine refresh should treat this SKU as a known-good starting point for short-reach 800G connectivity.