NVIDIA/Mellanox MMS4X00-NM-HGX (980-9I302-00NM00) Compatible 800GBASE-2xDR4 OSFP RHS/Flat Top 8x100G PAM4 1310nm 500m SMF DOM Dual MPO-12/APC Optical Transceiver Module for ConnectX-8/ConnectX-9 SuperNIC & NVIDIA Rubin NVL8 Server

$789.00

SKU: 6D4E38CAF8D9

Product Highlights

  • Maximum Power Consumption: ≤16.5W.
  • Professional Compatibility Verification and Optical/Electrical Parameter Test Reports Available.
  • Compliant with the OSFP MSA for high-speed, hot-pluggable optical connectivity.
  • RoHS Compliant and Lead-Free.
  • DDM (Digital Diagnostic Monitoring) Supported.
  • Designed for high-bandwidth multi-plane networking. When deployed with compatible C8180 SuperNICs and supported system configurations, it can support advanced network topologies, including quad-plane designs.
  • 30-Day Free Return, 1-Year Free Replacement, 3-Year Warranty, and Lifetime After-Sales Technical Support.

NVIDIA/Mellanox MMS4X00-NM-HGX (980-9I302-00NM00) Compatible 800GBASE-2xDR4 OSFP RHS/Flat Top 8x100G PAM4 1310nm 500m SMF DOM Dual MPO-12/APC Optical Transceiver Module

The NVIDIA/Mellanox MMS4X00-NM-HGX (980-9I302-00NM00) Compatible OSFP-800G-2xDR4 optical transceiver is designed for high-bandwidth networking in next-generation AI data centers and accelerated computing infrastructures. Featuring a twin-port OSFP architecture, it provides an aggregate data rate of 800Gbps through two independent 400G DR4 optical channels.

Operating at a nominal wavelength of 1310nm, the module supports transmission distances of up to 500m over single-mode fiber (SMF). Each optical channel uses four 100G PAM4 lanes and a dedicated MPO-12/APC interface, enabling two independent 400G optical links through a single twin-port OSFP module. This dual-channel architecture provides flexible connectivity for high-performance Ethernet and InfiniBand network deployments.

The module is designed for compatible NVIDIA networking platforms and HGX Rubin NVL8 server infrastructure. It can be integrated into supported system architectures involving ConnectX-8 or ConnectX-9 SuperNICs, subject to the specific host configuration, port mapping, firmware, and optical connectivity requirements.

The OSFP housing is available in configurations intended for different host thermal designs. Flat Top and RHS-compatible implementations should be selected according to the equipment’s cage structure, heatsink arrangement, and airflow requirements. The NVIDIA reference ordering information for MMS4X00-NM-HGX specifies a Flat Top version for HGX Rubin NVL8 systems.

DDM functionality enables monitoring of supported module operating parameters, helping network administrators assess optical link conditions and simplify maintenance. With a maximum power consumption of ≤16.5W, this module provides a high-density optical connectivity option for AI clusters, accelerated computing platforms, and large-scale data center networks.

Specifications

Part Number OSFP-800G-2DR4F Vendor Name SFPPlaza
Form Factor Twin-port OSFP Max. Data Rate 800Gbps
Wavelength 1310nm Max. Cable Distance 500m
Connector Two MPO-12/APC Media SMF
Transmitter Type EML Receiver Type PIN
Max. Power Consumption ≤16.5W Protocols 800G Ethernet Compliant
DDM/DOM Supported Commercial Temperature Range 0 to 70°C (32 to 158°F)
Modulation (Electrical) 8×106.25Gb/s PAM4 Modulation (Optical) Two ports 4×106.25Gb/s PAM4

Highlights

Designed for NVIDIA HGX Rubin NVL8 Infrastructure

The MMS4X00-NM-HGX compatible transceiver is designed for high-bandwidth optical connectivity in NVIDIA HGX Rubin NVL8 server environments. It supports demanding network architectures that connect accelerated computing resources with high-performance switching infrastructure. Typical deployment scenarios include:
  • NVIDIA HGX Rubin NVL8 AI server networking.
  • GPU-to-switch and server-to-switch optical interconnections.
  • Large-scale distributed AI training and inference clusters.
  • High-performance computing and accelerated data processing.
  • High-bandwidth Ethernet and InfiniBand network fabrics.

Twin-Port Architecture for Flexible Network Topologies

The two independent optical ports provide flexibility for multi-link network designs. Depending on the supported system architecture, the module can be deployed in switch-to-switch connections or in configurations connecting multiple SuperNICs. This allows network designers to organize bandwidth and physical links across different network planes.

Multi-Plane Networking with C8180 SuperNICs

When integrated into a validated C8180-based system, the twin-port optical architecture can form part of advanced multi-plane network topologies, including quad-plane designs where supported by the platform. The complete topology depends on switch port allocation, NIC configuration, firmware, and the correct crossover or breakout fiber assemblies.

Flexible OSFP Thermal Configurations

Different host platforms use different OSFP cooling arrangements. Flat Top and RHS-compatible options address different mechanical and thermal integration requirements. Before ordering, confirm whether the target equipment uses a module-mounted heatsink, a host-side heatsink, or a specific Riding Heatsink configuration.

Professional Compatibility and Deployment Support

Compatibility verification and optical/electrical parameter testing help reduce integration risks in complex AI network environments. Professional technical support is available to assist with module selection, fiber connectivity, host compatibility, and troubleshooting.

100% Compatibility Validated for Arista, NVIDIA, Cisco & Broadcom RoCE Networks

SFPPlaza optical transceivers, AOCs, and DACs undergo rigorous compatibility and interoperability testing with leading RoCE networking platforms, including Arista 7060XE7 and 7060X6 Series data center switches, NVIDIA Spectrum-4 SN5000 Series switches, Cisco N9100 and Nexus 9364E-SG2 Series switches, and Broadcom BCM57708 and BCM957608 NICs. Designed for high-performance RoCE and AI data center networks, our connectivity solutions deliver reliable high-speed data transmission, stable link performance, and seamless integration across demanding network environments.

NVIDIA/Mellanox (Ethernet) 920-9N42F-00RI-3C1, MSN4700-WS2FC, 900-9X81E-00EX-ST0, MCX75310AAS-NEAT, B3240, MCX623105AC-VDAT
Arista 7060XE7-128PE, 7060X6-64PE, 720R4-128PE, 7060DX5-32, 7050DX4-32S, 7280DR3A-54
Juniper QFX5240-64QD, QFX5700, QFX5220-32CD, QFX5120-48T, QFX5110-32Q
Cisco N9164E-NS4-O, N9364E-SG2-Q, Nexus 9364D-GX2A, Nexus 9504, 8501-SYS-MT, Nexus 3432D-S, N9K-X9400-16W(200G)
Broadcom P1800GO, BCM957608-P1400GDF00, BCM957508-N1200G
HUAWEI CE9875-64EO, CloudEngine 16804, CloudEngine 9865-4C, CloudEngine S6780-H4Z, CloudEngine 6885-48YS8CQ-T
H3C S9827-64EP, S9857-24DH8EP, S9855-32D, S9820-8C
Nokia 7220 IXR-H5-32D, 7220 IXR-H2

Questions & Answers

OSFP-800G-2xDR4: Frequently Asked Questions

Q: What type of optical fiber cable is required for the 800G twin-port 2xDR4 optical transceiver?

A: This module uses two MPO-12/APC optical interfaces, each carrying a 4-channel 400G DR4 link. For direct connection between compatible twin-port transceivers, use two single-mode MPO-12/APC crossover fiber cables with the correct polarity and lane mapping. OS2 single-mode fiber is suitable for the specified 500m link reach, subject to the complete link budget and cable specifications. For NVIDIA-compatible deployments, select the appropriate 4-channel MPO-12/APC crossover cable assembly.

Q: What is the difference between Flat Top and Finned Top OSFP 800G 2xDR4 optical modules?

A: The main difference is the module housing and thermal integration. A Flat Top module has a flat upper surface and is intended for host systems with a compatible heatsink or cooling structure. A Finned Top module incorporates fins on the module housing to provide additional heat-dissipation surface area. The correct version depends on the host cage design, heatsink arrangement, airflow, and thermal requirements. Always verify the exact OSFP housing specification supported by your NIC, switch, or server platform before installation.

Q: Can this 800G 2xDR4 module be used with NVIDIA ConnectX-8 C8180 SuperNICs?

A: This module is designed for compatible NVIDIA 800G networking architectures and HGX Rubin NVL8 deployments. In C8180-based configurations, the twin-port module may be used on a compatible switching platform to establish optical links to ConnectX-8 SuperNICs. The C8180 adapter itself uses a single-port OSFP cage with a specific RHS thermal configuration, so do not assume that this twin-port Flat Top module can be installed directly into the NIC. Verify the system topology, host port configuration, cable mapping, and thermal requirements before deployment.

Q: Does this module support quad-plane networking with NVIDIA C8180 SuperNICs?

A: Quad-plane networking may be supported as part of a validated system-level network architecture using compatible C8180 SuperNICs, switches, and optical cabling. The twin-port 2xDR4 module provides two independent 400G optical links. The resulting network topology and link stability depend on the host platform, switch configuration, firmware, and physical cabling. Confirm the supported quad-plane configuration with the system design before deployment.

Q: Can this module connect to two separate 400G DR4 optical ports?

A: Yes. The module provides two independent 400G DR4 optical channels through two MPO-12/APC interfaces. Each interface can connect to a compatible 400G DR4 optical port using a suitable single-mode MPO-12/APC crossover cable. Ensure that the optical lane mapping, connector polarity, link budget, and host port configuration are compatible.
x