Mellanox CX556A ConnectX-5 VPI 2P QSFP28 Ethernet/InfiniBand NIC MCX556A-ECAT (Low Profile Bracket)

Ethernet / InfiniBand PCIe 3.0 x16 Network Adapter Card – MCX556A-ECAT

The NVIDIA Mellanox ConnectX-5 VPI Adapter Card (MCX556A-ECAT) is a high-performance dual-port network interface card designed for 100GbE Ethernet and EDR InfiniBand connectivity. Built for modern data centers, cloud infrastructure, and high-performance computing (HPC) environments, it delivers ultra-low latency, advanced hardware offloads, and protocol flexibility via Mellanox’s Virtual Protocol Interconnect (VPI) technology.

Part Number Identification

• MCX556A-ECAT – Manufacturer part number

• CX556A – ConnectX-5 adapter family designation

General Specifications

• Device Type: Network Adapter (VPI – Virtual Protocol Interconnect)

• Ports: 2 × QSFP28

• Maximum Data Rate:

  • Up to 100 Gb/s per port

  • Supports Ethernet and InfiniBand modes

• Host Interface: PCI Express 3.0 x16

  • Backward compatible with earlier PCIe generations

  • Operates in PCIe Gen 4.0 systems (slot dependent)

• Form Factor: Plug-in card

  • Available with full-height (standard) and low-profile brackets

• Typical Power Consumption: ~23W

Supported Protocols

InfiniBand

• SDR, DDR, QDR, FDR, EDR (up to 100 Gb/s)

Ethernet

• 1GbE, 10GbE, 25GbE, 40GbE, 50GbE, 100GbE

Standards Compliance

• IEEE 802.x

• InfiniBand Trade Association (IBTA) standards

Key Features & Accelerations

• VPI (Virtual Protocol Interconnect):
  Automatically detects and operates over Ethernet or InfiniBand

• Ultra-Low Latency:
  Sub-600 nanosecond latency for latency-sensitive workloads

• RDMA Support:

  • RoCE (RDMA over Converged Ethernet)

  • Native InfiniBand RDMA

• NVMe over Fabrics (NVMe-oF):
  Enables high-throughput, low-latency storage access

• GPUDirect:
  Direct data movement between NIC and GPU memory

• Virtualization:

  • SR-IOV support

  • Hardware-accelerated switching and offloads

• CPU Offload Engine:
  Reduces host CPU utilization for networking and storage workloads

Operating System Compatibility

• Microsoft Windows Server

• Linux distributions (RHEL, CentOS, Ubuntu, FreeBSD)

• VMware ESXi

(OS support depends on driver and firmware versions)

Ideal Use Cases

• High-Performance Computing (HPC)

• Cloud and hyperscale data centers

• Big data analytics

• Machine learning and AI workloads

• High-speed storage and NVMe-oF deployments