Specifications
Product
NVIDIA/Mellanox MMS4A50-XM 1.6T OSFP224 2xFR4 XDR optical transceiver
Primary PN
MMS4A50-XM
NVIDIA Order PN
980-9IAS0-00XM00
Form Factor
OSFP224 (OSFP1600), IHS / closed finned top
Aggregate Rate
1.6 Tb/s
Port Configuration
2 × 800G FR4
Electrical Lanes
8 × 200 Gb/s PAM4
Wavelength
CWDM (4-lambda per 800G port)
Reach
2 km over single-mode fibre
Optical Connector
Dual duplex LC
DSP
Broadcom Sian3 BCM83628, 3 nm
FEC
No additional optical-side FEC required
Application
NVIDIA Quantum-X800 air-cooled switches; XDR spine links
Protocols
InfiniBand XDR, Ethernet 800GbE
Thermal
Integrated finned heat sink for air-cooled cages
Compliance
OSFP MSA, CMIS
Why FR4 over DR8
Duplex LC structured cabling at 2 km instead of high-count MPO trunks
Availability
Volume — quote for allocation
Overview
The MMS4A50-XM is the reach-optimised member of the NVIDIA 1.6T OSFP224 family. Instead of eight parallel single-mode fibres on an MPO trunk, it multiplexes four CWDM wavelengths per 800G port onto duplex LC, reaching 2 km. That makes it the right choice wherever the AI fabric crosses structured single-mode cabling — head-end to hall, or between buildings — rather than terminating inside a cable-rich pod.
It carries the same 8 × 200G PAM4 electrical interface and Broadcom Sian3 DSP as MMS4A00-XM, so the switch cage, firmware profile and thermal design are identical; only the optical interface and reach differ. In practice the two are mixed in one fabric, and the engineering task is simply to record which switch ports carry DR8 (MPO) and which carry FR4 (LC).
QS Compute supplies MMS4A50-XM alongside MMS4A00-XM, MMS4C10-XM and the full 1.6T OSFP224 cable portfolio, so a single purchase order covers an entire XDR leaf-spine build regardless of the fibre plant it lands on.
Key Benefits
2 km reach covers cross-hall and structured-cabling runs the DR8 variant cannot. Duplex LC drops the fibre count per link dramatically versus MPO trunks, which matters when a cable tray is already full. Shared DSP and cage design with MMS4A00-XM keeps spares and firmware profiles to one family. IHS finned top matches the Quantum-X800 air-cooled switch cage.
Applications
Cross-hall and inter-building XDR spine links; fabrics built on existing duplex single-mode cable plants; long-reach storage and management fabrics between AI halls; and mixed DR8/FR4 builds where reach, not fibre count, is the constraint.
Request a Quote — NVIDIA MMS4A50-XM (980-9IAS0-00XM00) — 1.6T OSFP224 2XFR4 XDR TRANSCEIVER
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →