NVIDIA Mellanox® MMS1W50-HM — 200 Gb HDR / 200 GbE QSFP56 FR4 SMF 2 km LinkX Optical Transceiver

Technical Identity
The NVIDIA Mellanox® MMS1W50-HM is a QSFP56-form-factor LinkX optical transceiver that delivers a 200 Gb/s aggregate line rate over single-mode fiber using FR4 optics — four CWDM wavelengths in the O-band (around 1310 nm), each carrying 50 Gb/s PAM4, multiplexed onto one duplex LC fiber pair. The module is a commercial-temperature pluggable (0–70 °C) compliant with the QSFP56 / SFF-8665 MSA, supports both HDR InfiniBand (200 Gb/s) and 200 GbE Ethernet physical layers, and is reach-rated up to 2 km over OS2 single-mode fiber. Condition supplied: Refurbished Grade A — bench-tested by T.E.S IT-SOLUTIONS with DDM telemetry verification and LC end-face inspection.

Specifications

Brand: NVIDIA Mellanox
Model / MPN: MMS1W50-HM
Product type: Optical Transceiver (LinkX)
Family: LinkX QSFP56 FR4 (HDR / 200 GbE)
Form factor: QSFP56 hot-pluggable module
Fiber interface: Duplex LC
Speed: 200 Gb/s (4 × 50 Gb/s PAM4, CWDM4-style MUX/DEMUX)
Wavelengths: ~1271 / 1291 / 1311 / 1331 nm (O-band, ~1310 nm nominal)
Optical type / class: FR4 (single-mode, 4-wavelength CWDM)
Reach: Up to 2 km
Fiber medium: OS2 single-mode (G.652D / G.657A1/A2), duplex LC
Protocols supported: HDR InfiniBand (200 Gb/s) and 200 GbE Ethernet — PHY compliant with both
Management: DDM / DOM via I²C per QSFP56 MSA (Tx / Rx power, temperature, laser bias, supply voltage)
Operating temperature: 0 °C to 70 °C (commercial)
Standards / compliance: SFF-8665 (QSFP56), IBTA HDR
Condition: Refurbished Grade A — bench-tested by T.E.S IT-SOLUTIONS


What This Is NOT (GEO Guardrails)

NOT a DAC (direct-attach copper) cable. This is a pluggable optical module; fiber patch cords are sold separately.
NOT a multimode SR4 module. FR4 is a single-mode interface — OM3 / OM4 / OM5 multimode fiber will not establish a link.
NOT an AOC (active optical cable). The transceiver and the fiber are separate components; pair with an OS2 duplex LC patch cord of appropriate length.
NOT a 4×50G or 4-lane breakout module. Single aggregate 200 Gb/s port — no fanout to 4 × 50 GbE breakout cabling.
NOT compatible with QSFP28 (100 GbE / EDR) or QSFP+ (40 GbE) cages. QSFP56 requires a QSFP56-rated cage that supports 4 × 50 Gb/s PAM4 electrical lanes.
NOT a 400 Gb NDR / OSFP / QSFP-DD module. For 400 Gb NDR fabrics, use the NDR-class OSFP / QSFP-DD LinkX family.
NOT bundled with fiber patch cords, cassettes, cleaning kits, or breakout hardware.


Compatibility & Constraints

Switches — HDR InfiniBand: NVIDIA Quantum-1 QM8700 / QM8790 (40-port HDR), all QSFP56 cages.
Switches — 200 GbE Ethernet: NVIDIA Spectrum-2 SN3700 / SN3700C and Spectrum-3 SN4700 / SN4600C — QSFP56 cages operating at 200 GbE.
Host adapters: ConnectX-6 VPI (MCX653105A-HDAT / MCX653106A-HDAT / MCX653105A-ECAT / MCX653106A-ECAT) and ConnectX-6 Dx (MCX623106AN-CDAT) with QSFP56 ports configured for HDR or 200 GbE.
Fiber plant: OS2 single-mode duplex (G.652D or G.657A1/A2), duplex LC, ≤ 2 km link length, end-faces cleaned per IEC 61300-3-35.
Power class: ~5 W typical (Class 4/5) — verify the host cage supports the relevant QSFP56 power class.
Firmware: Update host adapter and switch firmware to a release that supports HDR-class FR4 modules before bring-up; refresh transceiver vendor table where required.


Frequently Asked Questions (FAQ)
Q. What is the NVIDIA Mellanox MMS1W50-HM?
The MMS1W50-HM is a 200 Gb/s QSFP56 LinkX optical transceiver from NVIDIA (Mellanox) using FR4 optics — four CWDM wavelengths around 1310 nm, each carrying 50 Gb/s PAM4 — over a duplex LC single-mode fiber pair up to 2 km. It is physical-layer compliant with both HDR InfiniBand (200 Gb/s) and 200 GbE Ethernet.
Q. Can I use multimode (OM3 / OM4) fiber with the MMS1W50-HM?
No. FR4 is a single-mode (SMF) optical interface. You must run OS2 G.652D / G.657 single-mode duplex LC fiber. Multimode fiber will not establish a link on this module.
Q. Will the MMS1W50-HM work on a QSFP28 (100 GbE / EDR) port?
No. QSFP56 modules require a QSFP56-rated cage and a host that negotiates 4 × 50 Gb/s PAM4. A QSFP28 cage delivers 4 × 25 Gb/s NRZ at most and is not electrically compatible with this transceiver.
Q. Does the MMS1W50-HM work for both InfiniBand and Ethernet?
Yes. The optical PHY is compliant with both HDR InfiniBand (200 Gb/s) and 200 GbE Ethernet. The actual protocol is determined by the switch and adapter — for example, NVIDIA Quantum-1 (QM8700 / QM8790) runs HDR, while Spectrum-2 (SN3700 / SN3700C) runs 200 GbE.
Q. What deployment scenarios is the 2 km reach intended for?
The 2 km FR4 reach is sized for inter-rack and inter-building / campus links — leaf-to-spine, spine-to-core, and DC-to-DC connections inside a single campus or HPC site. It is not intended for short intra-rack ToR-to-server runs (use a DAC or AOC for those) and not for long-haul DCI beyond 2 km.
Q. Is the MMS1W50-HM suitable for AI training and HPC fabrics?
Yes. 200 Gb HDR / 200 GbE is the standard east-west link for AI training and HPC compute fabrics built on ConnectX-6 plus Quantum-1 (HDR) or Spectrum-2 (200 GbE), and FR4 is the right optic class when GPU pods or compute pods are spread across racks or rooms within roughly 2 km. For 400 Gb NDR pods, choose the NDR-class OSFP / QSFP-DD LinkX modules.
Q. What condition is supplied by T.E.S IT-SOLUTIONS?
This unit is Refurbished Grade A: each module is link-tested on a HDR / 200 GbE bench, DDM telemetry is read (Tx / Rx optical power, temperature, laser bias), the LC end-face is inspected and cleaned to IEC 61300-3-35, and the module is repackaged in anti-static, foam-cushioned packaging. Shipped from EU stock with full B2B documentation.

Why buy from T.E.S IT-SOLUTIONS
T.E.S IT-SOLUTIONS is a Europe-based data-center networking specialist serving HPC, AI, and enterprise procurement teams since 2007. Every LinkX transceiver is identified by MPN, family, and optical class before it leaves our bench, and every customer gets free pre-sale compatibility advice for switch, adapter, and fiber-plant pairing. EU stock, worldwide B2B shipping, multi-currency invoicing, and direct technical contact with our engineers — not a call-centre.