BCM5708C
®
10/100/1000BASE-T TCP Offload Engine, RDMA, iSCSI/iSER and
Ethernet Controller with PCI-Express™
FEATURES
•
Single-chip solution for LAN on Motherboard (LOM) and
Network Interface Card (NIC) applications
•
Integrated 10BASE-T/100BASE-TX/1000BASE-T transceivers
•
Host interfaces
- PCI-Express™ x4 Host interface
TCP offload engine
•
Full “fast Path” TCP offload
•
Compliant with Microsoft’s TOE Chimney Architecture
iSCSI controller
•
iSCSI initiator
•
iSER (iSCSI over RDMA)
RDMA controller (RNIC)
•
RDMA over TCP (iWARP)—RDMAC 1.0 compliant
•
Hardware-based data placement in application buffers without
CPU intervention (User and Kernel modes)
Other performance features
•
Receive Side Scaling (RSS)
•
TCP, IP checksum
•
TCP segmentation
•
Adaptive interrupts
•
Message Signal Interrupt (MSI) support
Robust manageability
•
Universal Management Port (UMP)
•
PXE 2.0 remote boot
•
Alert Standard Format (ASF v1.0) support
•
Wake-On LAN
•
IPMI 'pass-through' feature
•
Statistics gathering (SNMP MIB II, Ethernet-like MIB,
Ethernet MIB (802.3x, clause 30))
•
Comprehensive diagnostic and configuration software suite
•
ACPI 1.1a compliant power management
Advanced network features
•
Virtual LANs—802.1q VLAN tagging
•
Jumbo frames (9 KB)
•
802.3x flow control
Low-power CMOS design
On-chip power circuit controller
400-ball 21x21 mm FBGA package
3.3V I/Os
JTAG
SUMMARY OF BENEFITS
•
Industry’s first 10/100/1000 TOE solution—power and space
•
•
optimized for server blade and low-profile NIC applications.
Extremely low CPU utilization for TCP/IP applications
•
Host CPU is free to run application code
•
Easy integration with Microsoft’s TOE Chimney Architecture
Accelerated IP-based storage
•
Lower CPU utilization for file-level storage protocols such as
CIFS and NFS
•
iSCSI functionality with low CPU utilization
RDMA support for data placement in application buffers
reduces CPU utilization and lowers data transit latencies.
Future-proof
•
Flexible implementation for TCP, iWARP and iSCSI can
accommodate specification changes and interoperability
issues.
Performance-focused – optimized for throughput and CPU
utilization
•
Adaptive interrupts
•
RSS reduces CPU utilization on multi-CPU systems.
•
MSI allows interrupt distribution in a multi-CPU host system.
•
PCI-Express host interface allows a low-latency access to CPU
and Memory resources.
Robust and highly manageable
•
UMP enables high bandwidth "out-of-band" system
management functionality over shared infrastructure.
•
PXE 2.0, ACPI 1.1, Wake-On LAN, ASF 1.0.
•
IPMI 'pass-through' capability allows on-board management
controllers access to the network in OS-present and OS-absent
states.
Server class reliability, availability, and performance features
•
Link aggregation and load balancing
- Switch-dependent
•
802.3ad (LACP), generic trunking (GEC/FEC)
- Switch and NIC independent
Low power for zero airflow implementations
•
Advanced power management
Minimal real estate—ideal for LOM
•
On-chip power circuit controller
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•