NVIDIA MMS4X00-NS 800G Twin-Port OSFP Transceiver: 2x400G NDR InfiniBand &

The NVIDIA MMS4X00-NS (finned‑top version) and MMS4X00-NS-FLT (flat‑top) are twin‑port 800Gb/s transceivers that utilize two MPO‑12/APC optical connectors, each delivering 400Gb/s using 100G‑PAM4 modulation. This innovation allows a single OSFP cage to support two physical ports, effectively doubling front‑panel density in Quantum‑2 InfiniBand and Spectrum‑4 Ethernet switches. The transceiver auto‑detects the switch protocol (InfiniBand or Ethernet) and operates seamlessly across both environments. With built‑in digital diagnostics (DDM), low‑power sleep mode (1.5W), and CMIS 4.0 compliance, the MMS4X00‑NS provides reliable, energy‑efficient 800G optical interconnect for modern data centers.

The MMS4X00-NS leverages 100G‑PAM4 modulation on 8 parallel single‑mode lanes per port, with two independent transceiver engines inside a single OSFP module. NVIDIA’s patented vertical MPO‑12/APC connector arrangement fits two full‑duplex 400G interfaces within the standard OSFP footprint. The module supports adaptive Tx equalization, programmable Rx output amplitude, and comprehensive digital diagnostic monitoring (temperature, voltage, bias current, TX/RX power). It also features host and line side loopback as well as PRBS pattern checking, enabling field‑level troubleshooting and link optimization.

• Switch‑to‑Switch at 800Gb/s: Two MMS4X00-NS transceivers (one in each Quantum‑2/Spectrum‑4) linked with two straight MFP7E30 fiber cables for 800G bi‑directional trunk.
• Switch to Dual 400G Adapters/DPUs: Connect one twin‑port transceiver to two ConnectX‑7 (OSFP) or BlueField‑3 (QSFP112) 400G cards using straight fiber cables.
• Switch to Four 200G Links: Using 1:2 splitter cables (MFP7E40‑N0xx) to provide four 200Gb/s connections from a single OSFP cage.
• DGX H100 Integration: Flat‑top version (MMS4X00-NS-FLT) mates with Cedar‑7 mezzanine cards inside NVIDIA DGX H100, reducing transceiver count by half compared to previous generations.
• Long‑Reach Options: For 500m switch‑to‑switch, order MMS4X00‑NS‑NM (purple stripe) for extended single‑mode reach.

Fully compatible with NVIDIA end‑to‑end InfiniBand and Ethernet ecosystems. Validated for use in:

• Quantum‑2 NDR InfiniBand switches (air‑cooled, finned‑top required)
• Spectrum‑4 400GbE switches (32/64 OSFP cages)
• ConnectX‑7 adapters (OSFP and QSFP112 form factors)
• BlueField‑3 DPUs (QSFP112)
• DGX H100 & liquid‑cooled platforms (flat‑top version)

Note: Parameters not publicly specified in the reference document are labeled as “Not publicly specified" – please confirm before ordering for custom configurations.

• Port density doubled: Two 400G ports per physical OSFP cage, reducing switch count and cabling complexity.
• Lower system power per gigabit: Shared module electronics and low‑power sleep mode dramatically cut idle energy.
• Future‑ready 800G capacity: Seamless migration from 400G to 800G infrastructure without changing switch hardware.
• Unified inventory: One transceiver SKU for both InfiniBand and Ethernet – simplifies spares and deployment.
• Native NVIDIA ecosystem validation: Rigorous production testing ensures out‑of‑box reliability and performance.

Hong Kong Starsurge Group provides full technical pre‑sales and post‑sales support, including compatibility verification, firmware advisory, and global logistics. Our team offers customized cable assembly recommendations and RMA services. For volume orders, dedicated account management and 24‑hour response available.

• Always install dust caps when no fiber cable is inserted; up to 80% of link issues originate from contaminated connectors.
• Clean both transceiver receptacle and fiber connector before mating. Use approved dry cleaning methods – never apply liquids.
• Observe ESD handling precautions. The module is sensitive to electrostatic discharge.
• Do not exceed absolute maximum ratings (case temperature >70°C or supply >3.6V).
• Use only NVIDIA‑recommended crossover Type‑B cables to maintain polarity and signal integrity.

Hong Kong Starsurge Group Co., Limited is a technology‑driven provider of network hardware, IT services, and system integration solutions. Founded in 2008, the company serves customers worldwide with products including network switches, NICs, wireless access points, controllers, cables, and related networking equipment. Backed by an experienced sales and technical team, Starsurge supports industries such as government, healthcare, manufacturing, education, finance, and enterprise. The company also offers IoT solutions, network management systems, custom software development, multilingual support, and global delivery. With a customer‑first approach, Starsurge focuses on reliable quality, responsive service, and tailored solutions that help clients build efficient, scalable, and dependable network infrastructure.

• Confirm switch model (Quantum‑2 / Spectrum‑4) and cooling type (air/liquid) to select finned or flat‑top.
• Calculate fiber length: up to 100m for standard, choose -NM for up to 500m.
• Determine required breakout: straight cables (2x400G) or 1:2 splitter (4x200G).
• Ensure MPO‑12/APC connectors (angled polish, green shell) – not compatible with UPC.
• Verify that both fiber cables in a twin‑port link are the same type (both straight or both splitter).
• Check that ConnectX‑7/BlueField‑3 firmware supports NDR200 or NDR400 speeds.

• NVIDIA MMS4X00-NS-FLT – Flat‑top twin‑port 800G for DGX H100
• NVIDIA MMS1X00-NS400 – Single‑port 400G OSFP transceiver (DR4)
• MFP7E30‑Nxxx – Straight MPO‑12/APC crossover fiber cable (1m to 100m)
• MFP7E40‑N0xx – 1:2 splitter MPO‑12/APC to dual MPO‑12/APC, up to 50m
• Quantum‑2 Switch Platform – 64 ports of 400G NDR InfiniBand

• NVIDIA Cable Management Guidelines and FAQ (Application Note)
• OSFP MSA Specification Rev. 1.12 (www.osfpmsa.org)
• CMIS 4.0 / 5.0 Management Interface Standard
• NVIDIA InfiniBand NDR Deployment Best Practice.