Dual-Port 10GBASE-T Adapter with Full Offload
The BCM957712-T represents the industry's first dual-port 10GBASE-T adapter with full offload capabilities. Packaged in a low-profile PCI Express® form factor, the solution supports both networking and storage traffic over a single 10 Gigabit Ethernet (GbE) connection. Using cost-effective RJ-45 connections for distances up to 100 meters, the new Broadcom adapter delivers significant cost and power savings to Enterprise server farms and cloud data centers.
Based on Broadcom's proven BCM57712 MAC and BCM84823 PHY components, the new 10GBASE-T adapter is the first technology to bring full offload capabilities to 10GbE and the first to integrate TCP/IP and iSCSI offload on a 10GBASE-T medium. Tight integration between the MAC and PHY layers and end-to-end interaction between the server and switch point provides optimized performance and flexibility.
- High performance adapter delivers bi-directional line rate throughput across both ports, and over 1 million iSCSI IOPS
- The BCM957712-T adapter is based on Broadcom's complete suite of 10GbE solutions featuring the company's leading dual-port MAC and dual-port 10GBASE-T PHY. It utilizes field proven drivers for mainstream operating systems and a comprehensive unified NIC management system using the Broadcom Advanced Control Suite (BACS) management application, providing a single platform for network and storage I/O management
- The turnkey adapter operates at 100 meters over CAT6a unshielded twisted pair (UTP) cables. The adapter's 10GBASE-T PHY uses a highly advanced digital signal processor (DSP) based architecture that allows robust operation over UTP wiring
- The innovative offload architecture minimizes CPU overhead eliminating competition for host resources while freeing valuable CPU cycles for processing user applications
- The iSCSI operation is offloaded and accelerated for block storage resulting in high input/output instructions per second, as well as higher bandwidth
- The adapter uses a remote switching system (RSS) to avoid bottlenecks by distributing network processing across multiple CPUs