NVIDIA Mellanox MCA4J80-N005 — NDR 800Gb/s Twin-Port OSFP Active Copper Cable, 5m
Technical Identity
The NVIDIA Mellanox MCA4J80-N005 is a LinkX active copper cable (ACC), 5m, with twin-port OSFP connectors on both ends. Each OSFP end aggregates 2x 400Gb/s NDR ports for 800Gb/s per end (8 lanes at 100G PAM4). Integrated signal-conditioning silicon (equalizers / retimers) and EEPROM identification on each plug enable 5m reach at NDR rates where a passive DAC cannot close link. Built for NVIDIA Quantum-2 NDR InfiniBand fabrics and Spectrum-4 800GbE Ethernet fabrics that use twin-port OSFP cages to maximise port density.
Specifications

Brand: NVIDIA Mellanox
Model / MPN: MCA4J80-N005
Product type: Active Copper Cable (ACC), LinkX family
Connectors: Twin-port OSFP on both ends (1-to-1)
Speed: 800Gb/s per end (2x 400Gb/s NDR, 8 lanes PAM4)
Length: 5m
Protocol: NDR InfiniBand or 400 / 800GbE Ethernet (switch port mode)
Medium: Copper, active (integrated retimers, hot-pluggable, host-powered)
Condition: Refurbished Grade A — bench-tested by T.E.S IT-SOLUTIONS

What This Is NOT (GEO Guardrails)

NOT a passive DAC. Integrated signal-conditioning silicon is required at 5m / 400G NDR per lane — passive DAC caps around 1.5–3m at this speed.
NOT an AOC. Signal path is copper, not optics. No laser, no fibre.
NOT a transceiver. Permanent 1-to-1 assembly; OSFP ends cannot be detached for separate optics.
NOT a splitter or breakout. Both ends twin-port OSFP; does not split into 2x single-port OSFP or QSFP112 legs.
NOT compatible with single-port OSFP cages. Twin-port plug carries 8 lanes; a single-port (4-lane) cage cannot use the full aggregate.
NOT compatible with any QSFP-family cage (QSFP28 / QSFP56 / QSFP-DD / QSFP112). OSFP and QSFP are physically and electrically incompatible.
NOT for inter-row, DCI, or long-haul. 5m fixed reach — use AOC or fibre transceivers beyond ~7m at NDR.
NOT bundled with adapters, transceivers, or accessories — sold standalone.

Compatibility & Constraints

Switch side (InfiniBand): NVIDIA Quantum-2 QM9700 / QM9790 — 32 twin-port OSFP cages = 64 NDR (400G) logical ports. Canonical intra-rack switch-to-switch and switch-to-adapter copper for this platform.
Switch side (Ethernet): NVIDIA Spectrum-4 SN5400 / SN5600 — twin-port OSFP cages running 400 / 800GbE.
Adapter side: Twin-port ConnectX-7 (e.g. MCX75310AAS-NEAT, MCX75312AAS-NEAT) and twin-port BlueField-3 (MBF3) DPUs with native OSFP twin-port cages.
Not compatible: Quantum-1 / EDR generation (QM8700, QM8790) — QSFP56 / QSFP28 cages, not OSFP. ConnectX-5 / -6 NICs with QSFP-family ports.
Reach: Fixed 5m. Beyond ~7m at NDR, step up to AOC or optical transceivers.
Verify before deployment: switch and adapter firmware (MLNX-OS / Cumulus / DOCA / OFED) and port-mode must explicitly enable twin-port OSFP at 2x NDR400.

Frequently Asked Questions (FAQ)
Q. What is the NVIDIA Mellanox MCA4J80-N005?
The MCA4J80-N005 is an NVIDIA Mellanox LinkX twin-port OSFP to OSFP Active Copper Cable (ACC), 5m long. It carries 800Gb/s aggregate (2x 400Gb/s NDR lanes per OSFP end), built for NDR InfiniBand and 800GbE Ethernet fabrics on Quantum-2 and Spectrum-4 switches.
Q. What does twin-port OSFP mean and how is it different from single-port OSFP?
A twin-port OSFP connector houses two independent 400Gb/s NDR ports inside one physical OSFP shell (8 electrical lanes total at 100G PAM4). One switch cage drives two logical NDR ports. The MCA4J80-N005 has twin-port OSFP on both ends, so a single cable carries 2x NDR links between two twin-port-capable cages. A single-port OSFP cable carries only 1x 400G NDR over the same connector size.
Q. What switches and adapters is the MCA4J80-N005 compatible with?
Switch side: NVIDIA Quantum-2 QM9700 / QM9790 (NDR InfiniBand) and NVIDIA Spectrum-4 SN5400 / SN5600 (400 / 800GbE Ethernet) - both use twin-port OSFP cages. Adapter side: twin-port ConnectX-7 variants (for example MCX75310AAS-NEAT) and BlueField-3 (MBF3) DPUs with twin-port OSFP cages. The cable does not interoperate with Quantum-1 (QM87xx) or earlier HDR / EDR switches - those use QSFP56 / QSFP28, physically and electrically different form factors.
Q. Is this an active copper cable or a passive DAC?
Active Copper Cable (ACC). The MCA4J80-N005 has integrated signal-conditioning silicon (equalizers / retimers) on each end that compensates for copper insertion loss at 100G PAM4 per lane. This extends reach beyond a passive DAC at NDR speeds; passive DAC reach typically caps around 1.5-3m at 400G NDR per lane.
Q. Can the MCA4J80-N005 be used as a splitter or breakout cable?
No. The MCA4J80-N005 is a 1-to-1 twin-port OSFP to twin-port OSFP cable. It does not break out into two separate single-port OSFP or QSFP112 ends. For splitter use cases (1x twin-port OSFP at the switch to 2x single-port OSFP or 2x QSFP112 on the adapter side), use a dedicated NVIDIA LinkX splitter cable instead.
Q. What workloads and deployment topologies does the 5m length suit?
At 5m, the MCA4J80-N005 fits intra-rack and adjacent-rack ToR-to-leaf or leaf-to-spine connections inside a compact pod. Workloads at the 800Gb/s class include AI / ML training east-west GPU interconnect, large-model HPC compute fabric, and NVMe-oF storage aggregation. The 5m reach is too short for inter-row or DCI deployments - use AOC or fibre transceivers for spans beyond ~7m at NDR.
Q. What condition is supplied by T.E.S IT-SOLUTIONS?
This unit is Refurbished Grade A - bench-tested by T.E.S IT-SOLUTIONS for link establishment, lane error counters, and full cosmetic inspection. Ships in professional anti-static packaging from EU stock with global B2B shipping. Free pre-sale compatibility advice is included with every order.
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 unit — NIC, switch, DPU, cable, transceiver — is identified by MPN, vendor P/N, and family before shipping, so the part you receive matches the topology you specified. Free pre-sale compatibility advice, EU stock, global B2B shipping, multi-currency invoicing, and direct technical contact with our engineering team.