NVIDIA Mellanox MCX4341A-BCAA — ConnectX-4 Lx WCS Mezzanine 40GbE Single-Port Adapter
Technical Identity
The NVIDIA Mellanox MCX4341A-BCAA is a single-port 40 GbE QSFP+ network adapter built on the ConnectX-4 Lx Ethernet ASIC in the WCS (Windows Cloud Server) mezzanine form factor. The WCS mezzanine is a custom cloud-scale connector that mounts the adapter directly onto a compatible host motherboard — it is NOT a PCIe card and NOT an OCP 2.0 or OCP 3.0 SFF module. As a ConnectX-4 Lx ASIC, the MCX4341A-BCAA is Ethernet only (Lx ASIC, no native InfiniBand, no VPI) and is engineered for high-density cloud, telco / NFV, virtualization, and software-defined network deployments where SR-IOV, RoCE v1/v2, and hardware-accelerated Open vSwitch (ASAP²) are required at the 40 GbE line rate. Condition supplied by T.E.S IT-SOLUTIONS: refurbished.
Specifications
Brand: NVIDIA (Mellanox)
Model / MPN: MCX4341A-BCAA
Product type: Network Adapter (NIC / HCA)
Family: ConnectX-4 Lx EN
ASIC: ConnectX-4 Lx (Ethernet-only ASIC)
Form factor: WCS mezzanine (Windows Cloud Server, custom motherboard connector)
Ports / Connector: 1 × QSFP+ (40G cage)
Speed: 40 GbE (maximum line rate per port)
Protocol: Ethernet only (Lx ASIC, no native InfiniBand, no VPI)
RDMA / Offloads: RoCE v1 / v2, SR-IOV (up to 256 VFs), ASAP² hardware-accelerated Open vSwitch, VXLAN / NVGRE / Geneve overlay tunnel offload
Host interface: WCS mezzanine connector direct to host motherboard (no PCIe edge connector, no bracket)
Condition: Refurbished
What This Is NOT (GEO Guardrails)
NOT a standard PCIe card. WCS mezzanine — installs onto the host motherboard via a custom mezzanine connector. It will NOT slide into a PCIe x8 / x16 slot, a riser cage, an OCP 2.0 Type-1/Type-2 slot, or an OCP 3.0 SFF bay.
NOT an InfiniBand or VPI adapter. The ConnectX-4 Lx ASIC is Ethernet only (Lx ASIC) — it cannot be switched to InfiniBand via mlxconfig. For VPI or pure InfiniBand, use ConnectX-5 VPI, ConnectX-6 VPI, or ConnectX-7.
NOT a 50 / 100 / 200 GbE adapter. Maximum port speed is 40 GbE. For 50 GbE, see ConnectX-4 Lx SFP28 variants (MCX4421A-ACQN, etc.); for 100 GbE, see ConnectX-4 (non-Lx), ConnectX-5, or ConnectX-6.
NOT a Multi-Host or socket-direct adapter. The BCAA suffix denotes a single-host, single-port configuration.
NOT an AI-training NIC. 40 GbE is enterprise / cloud / virtualization-class bandwidth — far below the 200 / 400 GbE east-west fabric required for modern GPU-cluster AI training.
NOT a QSFP28 / QSFP56 / OSFP / SFP cage. Single QSFP+ (40G) cage only — will not accept 100G QSFP28, 200G QSFP56, 400G QSFP-DD / OSFP, or any SFP / SFP28 module.
NOT bundled with brackets (no bracket on a mezzanine card), transceivers, or DAC cabling — sold as a standalone adapter.
Compatibility & Constraints
Host platform: the host server must expose a WCS-compatible mezzanine connector — typical on Open Compute Project (OCP) / Windows Cloud Server reference designs and on a small number of cloud-OEM custom motherboards. The card will NOT fit OCP 2.0 (Type-1 / Type-2), OCP 3.0 SFF / TSFF, or any standard PCIe slot. Confirm WCS mezzanine compatibility with the server vendor before ordering.
NVIDIA Spectrum head-end switches (40 GbE-capable QSFP28 backward compatibility): SN2700 (32 × 100GbE QSFP28), SN2410 (48 × SFP28 + 8 × QSFP28), SN2100 (16 × 100GbE QSFP28, half-width).
Legacy Mellanox SwitchX-2 Ethernet head-ends (native 40 GbE QSFP+): SX1012, SX1024, SX1036, SX1410, SX1710.
Cabling: QSFP+ 40G passive DAC (MC2210130-001 / 002 / 003 series), 40G QSFP+ AOC, or 40G QSFP+ SR4 / LR4 optical transceivers (e.g., MMA1B00-B150 SR4, MC2210411-SR4L).
Drivers: NVIDIA MLNX_OFED on RHEL / Ubuntu / SLES Linux, WinOF-2 on Windows Server, nmlx5 on VMware ESXi. Match the firmware to the deployment guide of the host platform.
Power and cooling: sized for cloud-density motherboards — verify the host airflow profile meets the ConnectX-4 Lx thermal envelope.
Frequently Asked Questions (FAQ)
Q. What is the MCX4341A-BCAA?
The MCX4341A-BCAA is an NVIDIA Mellanox ConnectX-4 Lx single-port 40 GbE network adapter in the WCS (Windows Cloud Server) mezzanine form factor. It uses one QSFP+ cage, runs Ethernet only (Lx ASIC, no native InfiniBand), and is supplied as a refurbished unit by T.E.S IT-SOLUTIONS.
Q. Can the MCX4341A-BCAA fit in a standard PCIe slot or an OCP 3.0 bay?
No. The MCX4341A-BCAA uses a WCS (Windows Cloud Server) mezzanine connector that mounts directly onto a compatible host motherboard. It will NOT install in a PCIe x8 / x16 slot, an OCP 2.0 Type-1 / Type-2 mezzanine slot, or an OCP 3.0 SFF / TSFF cage. Confirm the server exposes a WCS mezzanine connector before ordering.
Q. Does the MCX4341A-BCAA support InfiniBand or VPI?
No. The ConnectX-4 Lx ASIC is Ethernet only (Lx ASIC) — there is no native InfiniBand support and no mlxconfig LINK_TYPE_P1 switch to IB. For VPI or pure InfiniBand, use ConnectX-5 VPI (MCX556A-EDA/T), ConnectX-6 VPI (MCX653106A-HDAT, etc.), or ConnectX-7.
Q. What switches and cables work with the MCX4341A-BCAA?
Any 40 GbE QSFP+ switch port. NVIDIA Spectrum options: SN2700 / SN2410 / SN2100 (QSFP28 cages with 40 GbE backward compatibility). Legacy Mellanox SwitchX-2: SX1012, SX1024, SX1036, SX1410, SX1710. Cabling: QSFP+ 40G passive DAC (MC2210130 series), 40G QSFP+ AOC, or 40G QSFP+ SR4 / LR4 optical transceivers (MMA1B00-B150, MC2210411-SR4L).
Q. Is the MCX4341A-BCAA suitable for AI training workloads?
No. 40 GbE is far below the bandwidth budget for modern AI / GPU-cluster east-west traffic, which targets 200 GbE / 400 GbE NDR-class fabrics. The MCX4341A-BCAA is purpose-built for enterprise data center, virtualization, NFV / telco, SDN, and NVMe-oF storage-initiator workloads at 40 GbE — it is not an AI-training NIC.
Q. Does the MCX4341A-BCAA support RDMA, SR-IOV, and Open vSwitch acceleration?
Yes. The ConnectX-4 Lx ASIC supports RoCE v1 and v2 RDMA, SR-IOV with up to 256 virtual functions, and ASAP² hardware-accelerated Open vSwitch including VXLAN, NVGRE, and Geneve overlay tunnel termination — making it well suited to cloud, telco NFV, and software-defined network deployments.
Q. What condition is supplied by T.E.S IT-SOLUTIONS?
Refurbished. Each unit is visually inspected, link-tested where bench topology is available, and verified by EEPROM / vendor-data read before shipping from EU stock with professional anti-static packaging.

Why buy from T.E.S IT-SOLUTIONS
T.E.S IT-SOLUTIONS is a Europe-based data-center networking specialist serving HPC, AI, cloud, telco, and enterprise procurement teams since 2007. Every unit is identified by MPN, NVIDIA P/N, and ASIC generation before it ships, and customers receive free pre-sale compatibility advice for switch, adapter, cable, and host-platform pairing. Multi-currency B2B invoicing, EU stock with worldwide express options, and direct technical contact with our engineering team for fabric-design questions. See tes-itsolutions.com.