HPE Ethernet 100Gb 2-Port QSFP28 MCX516A-CDAT Adapter for HPE

The HPE Ethernet 100Gb 2-Port QSFP28 MCX516A-CDAT Adapter (P21927-B21) is built on the Mellanox ConnectX-5 ASIC and delivers dual-port 100 Gigabit Ethernet connectivity via QSFP28 interfaces. Designed and validated for HPE ProLiant Gen10 and Gen10 Plus server platforms, this adapter supports both Ethernet and RoCE (RDMA over Converged Ethernet) operation, enabling low-latency, high-throughput networking for demanding data-center environments. The adapter is a PCIe 3.0 x16 device and supports hardware offloads including stateless TCP/UDP/IP checksum offload, large send offload (LSO), and receive-side scaling (RSS).

The MCX516A-CDAT variant is optimized for 100GbE operation on both ports simultaneously, making it suitable for environments requiring maximum bandwidth density per PCIe slot. Support for RDMA over Converged Ethernet (RoCEv2) allows applications to bypass the OS network stack and communicate directly with remote memory, dramatically reducing CPU overhead and latency for storage and inter-node communication workloads. The adapter integrates with standard data-center switching fabrics and is compatible with industry-standard network management frameworks.

As an HPE-branded and factory-integrated option, the adapter is tested and qualified within HPE's server firmware and driver ecosystem, ensuring compatibility with HPE iLO, Smart Array, and ProLiant system software stacks. It ships with HPE-supplied drivers and is supported under HPE's standard server support agreements, simplifying lifecycle management in enterprise deployments. Target environments include high-performance computing clusters, large-scale virtualization, NVMe-oF storage fabrics, and AI/ML training infrastructure.

Ideal for

• NVMe over Fabrics (NVMe-oF) storage networking, leveraging RoCEv2 for low-latency, CPU-efficient access to all-flash storage arrays
• High-density server virtualization requiring 100GbE uplinks to handle aggregated VM traffic from HPE ProLiant hosts in VMware vSphere or Microsoft Hyper-V environments
• AI and machine learning distributed training clusters where inter-GPU RDMA communication over RoCEv2 reduces synchronization latency between nodes
• High-performance computing (HPC) workloads demanding consistent low-latency MPI message passing across compute nodes connected via 100GbE leaf-spine fabrics
• Large-scale data analytics platforms (Hadoop, Spark) requiring high-throughput east-west bandwidth between compute and storage tiers
• Software-defined networking and NFV deployments on ProLiant servers where SR-IOV support enables direct hardware access for virtualized network functions

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.