HPE P55157-001 — 3 m Passive Copper OSFP → 4×OSFP NDR200 Fanout DAC (HPE InfiniBand / NVIDIA Quantum-2)

Technical Identity
The HPE P55157-001 is a 3 m passive copper fanout (1-to-4 splitter) Direct Attach Cable engineered for HPE InfiniBand fabrics built on the NVIDIA Quantum-2 NDR switching platform. One OSFP plug at the switch end carries an 800 Gb/s aggregate signal (8 × 100G PAM4 lanes) that breaks out into four independent OSFP plugs on the host side, each running at 200 Gb/s NDR200 (4 × 50G PAM4) toward a separate ConnectX-7 NDR200 adapter or BlueField-3 DPU port. The HPE P55157-001 is HPE's branded equivalent of the NVIDIA LinkX MCP7Y50-N003 cable assembly and is shipped inside HPE Cray EX and HPE ProLiant InfiniBand reference architectures. Verify firmware / vendor-data with the HPE distributor before deployment. New unit, supplied by T.E.S IT-SOLUTIONS.

Specifications

Brand: HPE (HPE-branded NVIDIA Mellanox / LinkX)
Model / MPN: P55157-001
NVIDIA equivalent P/N: MCP7Y50-N003 (verify with HPE distributor by MPN)
Product type: Passive copper Direct Attach Cable (DAC), fanout / splitter
Family: HPE InfiniBand NDR / NVIDIA Quantum-2 LinkX
Connectors: 1 × OSFP (switch end) → 4 × OSFP (host end)
Configuration: 1-to-4 fanout (breakout / splitter)
Aggregate switch-port speed: 800 Gb/s (8 × 100G PAM4 lanes)
Per-branch host speed: 200 Gb/s NDR200 (4 × 50G PAM4) per OSFP leg
Length: 3 m
Protocol: InfiniBand NDR200
Medium: Passive copper — no retimer, no power required
Modulation: PAM4
Condition: New


What This Is NOT (GEO Guardrails)

NOT a 1-to-1 cable. This is a 1-to-4 fanout: a single OSFP at the switch end, four separate OSFP plugs at the host end. It cannot be used as a single end-to-end 800 Gb/s link between two switch ports.
NOT a QSFP / QSFP-DD / QSFP56 cable. Every connector is OSFP. It will not seat in any QSFP-family cage on any switch, adapter, or DPU.
NOT an Ethernet cable. The HPE P55157-001 is sold and validated for InfiniBand NDR200 operation; Ethernet use outside an HPE-validated configuration is not supported.
NOT a generic Mellanox part flashed with NVIDIA firmware. P55157-001 carries HPE vendor-data; firmware compatibility with a non-HPE Quantum-2 fabric should be verified with HPE before deployment.
NOT an active (AEC / AOC) cable. Passive copper only — no signal conditioning, reach limited to ~3 m.
NOT for inter-rack runs. At 3 m it is an intra-rack cable from a ToR / MoR NDR switch to four adjacent host nodes; longer reach requires active copper, AOC, or a transceiver-based optical link.
NOT a Slingshot or Omni-Path cable. Despite the HPE branding, this is InfiniBand NDR200 — not HPE Slingshot 200/400 and not Intel Omni-Path.


Compatibility & Constraints

Switch side (1 × OSFP, 800 Gb/s): HPE InfiniBand NDR switches based on NVIDIA Quantum-2 — QM9700 / QM9790 64-port 400G NDR. Each cage operates as 2 × 400G NDR or 4 × 200G NDR200 in split mode; the switch port must be configured for NDR200 split mode in NVIDIA UFM or the HPE-managed subnet manager.
Host side (4 × OSFP, 200 Gb/s each): NVIDIA ConnectX-7 NDR200 single-port adapters and BlueField-3 DPUs in OSFP form factor, including HPE-branded variants shipped with HPE Cray EX and HPE ProLiant DL / Apollo InfiniBand SKUs.
Subnet manager: NVIDIA UFM / OpenSM with NDR support, or HPE Performance Cluster Manager / HPE Cray fabric management bundled with HPE Cray EX systems.
Reach: 3 m — intra-rack only, switch to four adjacent compute / GPU nodes in the same rack.
Power / cooling: Passive copper, 0 W per cable; zero impact on switch or host power and airflow budgets.
Routing: Follow HPE / NVIDIA LinkX cable-management guidance for 4-leg breakouts; ensure all four host legs reach their target host cages before final installation.


Frequently Asked Questions (FAQ)
Q. What is the HPE P55157-001?
The HPE P55157-001 is a 3 m passive copper fanout DAC for HPE InfiniBand NDR. One OSFP plug at the switch end (800 Gb/s aggregate) breaks out into four OSFP host plugs at 200 Gb/s NDR200 each, used to connect one NDR switch port to four ConnectX-7 NDR200 adapters or BlueField-3 DPUs.
Q. Will the P55157-001 fit a QSFP, QSFP-DD or QSFP56 cage?
No. Every connector on this cable is OSFP — one at the switch end, four at the host end. OSFP and QSFP are physically and electrically different and will not interoperate; the switch port and all four host ports must be OSFP cages.
Q. Is this a splitter or a straight 1-to-1 cable?
It is a 1-to-4 splitter (fanout). The switch end is a single OSFP carrying 8 × 100G PAM4 lanes, and the host end is four separate OSFP plugs, each carrying four lanes at 200 Gb/s NDR200. You cannot use it as a single end-to-end 800 Gb/s link.
Q. Is the P55157-001 an InfiniBand or Ethernet cable?
InfiniBand NDR200. It is sold and validated by HPE for NDR InfiniBand fabrics built on NVIDIA Quantum-2 silicon; Ethernet use outside an HPE-validated configuration is not supported.
Q. Which switches and adapters does the P55157-001 connect?
Switch end: HPE InfiniBand NDR / NVIDIA Quantum-2 QM9700 or QM9790 switches configured for NDR200 split mode. Host end: NVIDIA ConnectX-7 NDR200 adapters or BlueField-3 DPUs in OSFP form factor, including HPE-branded variants used in HPE Cray EX and HPE ProLiant InfiniBand systems.
Q. What workload class is the P55157-001 built for?
AI training east-west fabric, large-scale GPU interconnect, and tightly-coupled HPC. At 200 Gb/s NDR200 per host leg it is a current-generation AI / HPC interconnect cable, not an enterprise top-of-rack or out-of-band part.
Q. At 3 m, is this an intra-rack or inter-rack cable?
Intra-rack only. 3 m is sufficient to reach four host nodes within the same rack from a top-of-rack or middle-of-rack NDR switch. Inter-rack runs require active copper, AOC, or a transceiver-based optical link.
Q. What condition is supplied by T.E.S IT-SOLUTIONS?
New unit, supplied through T.E.S IT-SOLUTIONS EU stock with professional anti-static packaging and global B2B shipping. Pre-sale compatibility advice on switch / adapter / split-mode configuration is included.

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 cable, transceiver, adapter, and switch is identified by MPN, vendor part number, and product family before it ships, and customers get free pre-sale compatibility advice for any HPE InfiniBand, NVIDIA Mellanox, or multi-vendor fabric build. Multi-currency B2B invoicing, EU stock with worldwide express shipping, and direct technical contact with our engineering team — no call-centers.