{{ 'fb_in_app_browser_popup.desc' | translate }} {{ 'fb_in_app_browser_popup.copy_link' | translate }}
{{ 'in_app_browser_popup.desc' | translate }}
The NVIDIA MCX623106AN-CDAT ConnectX-6 Dx EN is a dual-port 100GbE PCIe 4.0 x16 SmartNIC. Equipped with QSFP56 interfaces, zero-touch RoCE, and hardware ASAP offloads, this non-crypto variant delivers sub-microsecond latency for cloud, storage, and AI workloads.
全店,Free Shipping
商品存貨不足,未能加入購物車
您所填寫的商品數量超過庫存
{{'products.quick_cart.out_of_number_hint'| translate}}
{{'product.preorder_limit.hint'| translate}}
每筆訂單限購 {{ product.max_order_quantity }} 件
現庫存只剩下 {{ quantityOfStock }} 件
The NVIDIA ConnectX®-6 Dx EN (MPN: MCX623106AN-CDAT) is an enterprise-grade Smart Network Interface Card (SmartNIC) engineered to accelerate mission-critical cloud, virtualization, storage, and artificial intelligence workloads. Delivering up to 200Gb/s aggregate bandwidth across dual 100GbE QSFP56 ports, this adapter leverages PCIe 4.0 x16 connectivity to unlock maximum server throughput with minimal CPU overhead.
This specific model (MCX623106AN-CDAT) is a non-cryptographic variant configured with a standard tall bracket. It is optimized for environments where hardware IPsec/TLS encryption is not required, making it a streamlined, power-efficient choice for hyperscale data centers, NVMe-oF storage arrays, and high-density compute nodes.
Dual 100GbE QSFP56 Ports: Provides flexible network topologies with up to 100Gb/s per port using 50G-PAM4 or 25G-NRZ SerDes technology.
PCIe Gen 4.0 x16 Host Interface: Full backward compatibility with PCIe Gen 3.0 slots, offering 16 lanes of 16.0 GT/s bus speed to eliminate host bottlenecks.
Zero-Touch RoCE (RDMA over Converged Ethernet): Enables ultra-low latency (sub-0.8µs) and high-throughput data transfers without requiring complex PFC (Priority Flow Control) or ECN network tuning.
ASAP² Accelerated Switching & Packet Processing: Hardware offload for SDN software-defined networking, SR-IOV, VirtIO, and overlay tunnels (VXLAN, GENEVE, NVGRE) to free up CPU cores.
Advanced Storage Offloads: Direct hardware acceleration for NVMe over Fabrics (NVMe-oF), iSER, NFSoRDMA, SMB Direct, and T10-DIF signature checks.
Streamlined Non-Crypto Profile: Designed without inline cryptographic engines for reduced thermal output and simplified compliance in non-regulated environments.
| Parameter | Specification |
| Manufacturer | NVIDIA / Mellanox |
| Part Number (MPN) | MCX623106AN-CDAT |
| Network Protocol | Ethernet (1/10/25/40/50/100 GbE) |
| Max Data Rate | Dual-Port 100GbE (200Gb/s aggregate) |
| Connector Type | Dual QSFP56 |
| Host Interface | PCIe 4.0 x16 (16.0 GT/s per lane) |
| Form Factor | Low-Profile PCIe Standalone (HHHL) |
| Bracket Type | Full-Height / Tall Bracket Attached (Short Bracket optional) |
| Crypto Support | Disabled / No Crypto Engine |
| Secure Boot | Disabled |
| Latency | < 0.8 µs |
| Message Rate | Up to 215 Million Packets Per Second (Mpps) |
| Typical Power | ~19.5W (Passive Copper Cables) |
| Operating System Support | RHEL/CentOS, Windows Server, VMware ESXi, FreeBSD, Ubuntu |
High-Performance Computing & AI: Enables GPUDirect® RDMA for direct, low-latency communication between GPUs and network storage nodes.
Enterprise Cloud & Virtualization: Accelerates hypervisor switching performance and guest VM routing via SR-IOV and VirtIO offloads.
NVMe-oF Storage Target/Initiator: Replaces dedicated Fibre Channel SANs with high-speed, scalable Ethernet-based storage fabrics.