Dual-Port Server Network Adapters

What a dual-port card is used for

A dual-port card gives a server two network links from one slot, and buyers use the pair in three ways: as a primary and standby link for failover, as a team that spreads traffic across both ports, or as two separate networks, for example storage on one port and virtual machine traffic on the other. The ports are independent, so each can run its own VLANs and addresses.

Dual-port cards by speed

A second port lets one slot carry two independent links, and the same two-port layout runs from gigabit to 100GbE. Broadcom documents its dual-port controllers this way:

Host interface matters most at the top of the range: a 100GbE card wants a Gen4 x16 slot, where a dual gigabit card needs a single lane. Other chipmakers build dual-port controllers too, so check the bracket height and host interface on each listing.

Where dual-port cards fit

A PCIe card needs only a free slot of matching height, while an OCP 3.0 card needs the OCP bay of Dell's 15th-generation servers such as the PowerEdge R750. Two-port rNDCs and blade cards are on network daughter cards and mezzanine adapters, and FCoE or iSCSI storage-offload cards on converged network adapters.

Shop one speed across all port counts on 10GbE adapters, 25GbE adapters and gigabit adapters, or four ports per card on quad-port adapters. Every card comes with an Enterasource warranty for the term on its listing.

Dual-port network card questions

What is a dual-port network card used for?

Redundancy, more bandwidth across many connections, or network separation. With two ports a server can survive a cable, port or switch failure if each port goes to a different switch, carry more total traffic when the ports are teamed, or keep storage, backup and management traffic off the production network. Both ports share one PCIe slot, so the slot must carry their combined rate.

Can I team the two ports for more bandwidth?

Yes, for many flows at once, not for one. Teaming spreads connections across the ports, but Microsoft's NIC teaming algorithms assign each flow to one team member, so a single transfer runs at one port's speed. Switch-dependent modes such as LACP need matching settings on the switch; switch-independent teaming does not, and either mode also provides failover.