NVIDIA ConnectX-7 netwerkadapter PCIe 5.0 x16 200Gb/s NDR200 IB QSFP112

The NVIDIA ConnectX-7 network adapter delivers dual-port 200Gb/s connectivity over PCIe 5.0 x16, supporting both InfiniBand and Ethernet with ultra-low latency and advanced offloads for demanding data center workloads.

Advantages

Engineered for high-performance computing and AI clusters, the ConnectX-7 provides unmatched throughput with 200Gb/s per port, hardware-based RDMA, and GPUDirect acceleration. Its OCP 3.0 SFF form factor ensures seamless integration into modern servers, while advanced congestion control and adaptive routing guarantee predictable performance under heavy loads. With support for up to 256 virtual functions per port, it enables efficient resource sharing in virtualized environments.

Suitable for

• High-performance computing (HPC) clusters requiring low-latency interconnects
• AI and machine learning training platforms leveraging GPUDirect RDMA
• Software-defined storage and NVMe-oF deployments
• Virtualized data centers needing SR-IOV and robust security features
• Cloud and enterprise networks demanding 200GbE bandwidth

Key specifications

• Controller: NVIDIA ConnectX-7
• Host interface: PCIe 5.0 x16
• Ports: 2x QSFP112
• Data rate: 200Gb/s (NDR200 InfiniBand / 200GbE Ethernet)
• Protocol support: InfiniBand, Ethernet (VPI)
• Offloads: RoCE, NVMe-oF, GPUDirect RDMA, DPDK, SR-IOV, VirtIO
• Form factor: OCP 3.0 SFF
• Power consumption: 25W typical
• Operating temperature: 0°C to 55°C
• OS support: Windows, Linux, VMware ESXi, FreeBSD

Recommendations

For data centers migrating to 200Gb/s fabrics, the ConnectX-7 offers a future-proof investment with its dual-mode VPI capability and comprehensive offload engine. Its OCP 3.0 design simplifies deployment in compatible servers, while advanced management features like NC-SI and PLDM streamline operations at scale.