参数资料
型号: PI7C9X20404GPBNBE
厂商: Pericom
文件页数: 3/79页
文件大小: 0K
描述: IC PCIE PACKET SWITCH 148LFBGA
产品变化通告: Product Discontinuation Notice 22/Jan/2010
标准包装: 189
系列: GreenPacket™
应用: 封装开关,4 端口/4 线道
接口: PCI Express
封装/外壳: 148-LFBGA
供应商设备封装: 148-LFBGA(12x12)
包装: 托盘
安装类型: 表面贴装
PI7C9X20404GP
4Port-4Lane PCI Express Switch
GreenPacket
TM Family
Datasheet
Page 11 of 79
June 2009 – Revision 1.6
Pericom Semiconductor
2 GENERAL DESCRIPTION
Similar to the role of PCI/PCIX Bridge in PCI/PCIX bus architecture, the function of PCI Express (PCIE) Switch is
to expand the connectivity to allow more end devices to be reached by host controllers in PCIE serial interconnect
architecture. The 4-lane PCIE Switch can be configured as 4-port type combinations. It provides users the flexibility
to expand or fan-out the PCI Express lanes based on their application needs. For some systems that do not need all
the 4 lanes, the unused lanes can be disabled to reduce power consumption.
In the PCI Express Architecture, the PCIE Switch forwards posted and non-posted requests and completion packets
in either downstream or upstream direction concurrently as if a virtual PCI Bridge is in operation on each port. By
visualizing the port as a virtual Bridge, the Switch can be logically viewed as two-level cascaded multiple virtual
PCI-to-PCI Bridges, where one upstream-port Bridge sits on all downstream-port Bridges. Similar to a PCI Bridge
during enumeration, each port is given a unique bus number, device number, and function number by the initiating
software. The bus number, device number, and function number are combined to form a destination ID for each
specific port. In addition to that, the memory-map and IO address ranges are exclusively allocated to each port as
well. After the software enumeration is finished, the packets are routed to the dedicated port based on the embedded
address or destination ID. To ensure the packet integrity during forwarding, the Switch is not allowed to split the
packets to multiple small packets or merge the received packets into a large transmit packet. Also, the IDs of the
requesters and completers are kept unchanged along the path between ingress and egress port.
The Switch employs the architecture of Combined Input and Output Queue (CIOQ) in implementation. The main
reason for choosing CIOQ is that the required memory bandwidth of input queue equals to the bandwidth of ingress
port rather than increasing proportionally with port numbers as an output queue Switch does. The CIOQ at each
ingress port contains separate dedicated queues to store packets. The packets are arbitrated to the egress port based
on the PCIe transaction-ordering rule. For the packets without ordering information, they are permitted to pass over
each other in case that the addressed egress port is available to accept them. As to the packets required to follow the
ordering rule, the Head-Of-Line (HOL) issue becomes unavoidable for packets destined to different egress ports
since the operation of producer-consumer model has to be retained; otherwise the system might occur hang-up
problem. On the other hand, the Switch places replay buffer at each egress port to defer the packets before sending it
out. This can assure the maximum throughput being achieved and therefore the Switch works efficiently. Another
advantage of implementing CIOQ in PCIe Switch is that the credit announcement to the counterpart is simplified
and streamlined because of the credit-based flow control protocol. The protocol requires that each ingress port
maintains the credits independently without checking other ports' credit availability, which is otherwise required by
pure output queue architecture.
The Switch supports two virtual channels (VC0, VC1) and eight traffic classes (TC0 ~ TC7) at each port. The
ingress port independently assigns packets into the preferred virtual channel while the egress port outputs the packet
based on the predefined port and VC arbitration algorithm. For instance, the isochronous packet is given a special
traffic class number other than TC0 and mapped into VC1 accordingly. By employing the strict time based credit
policy for port arbitration and assigning higher priority to VC1 than VC0, the Switch can therefore guarantee the
time-sensitive packet is not blocked by regular traffic to assure the quality of service. In addition, some data-centric
applications only carry TC0/VC0 traffic. As a result, there are no packets that would consume VC1 bandwidth. In
order to improve the efficiency of buffer usage, the unused VC1 queues can be reassigned to VC0 and enable each
of the ingress ports to handle more data traffic bursts. This virtual channel resource relocation feature enhances the
performance of the PCIe Switch further.
相关PDF资料
PDF描述
PI7C9X20404SLCFDE IC PCIE PACKET SWITCH 128LQFP
PI7C9X20505GPBNDE IC PCIE PACKET SWITCH 256BGA
PI7C9X20508GPBNDE IC PCIE PACKET SWITCH 256BGA
PI7C9X440SLBFDE IC PCIE-TO-USB 2.0 CTRLR 128LQFP
PI7C9X442SLBFDE IC PCIE-TO-USB2.0 SWIDGE 128LQFP
相关代理商/技术参数
参数描述
PI7C9X20404SLCEVB 制造商:Pericom Semiconductor Corporation 功能描述:PCIE 4 PORT SWITCH EVAL BOARD - Boxed Product (Development Kits)
PI7C9X20404SLCFDE 功能描述:外围驱动器与原件 - PCI 4port 4lane PCIe Packet Switch RoHS:否 制造商:PLX Technology 工作电源电压: 最大工作温度: 安装风格:SMD/SMT 封装 / 箱体:FCBGA-1156 封装:Tray
PI7C9X20404SLCFDEX 功能描述:外围驱动器与原件 - PCI 4port 4lane PCIe Packet Switch RoHS:否 制造商:PLX Technology 工作电源电压: 最大工作温度: 安装风格:SMD/SMT 封装 / 箱体:FCBGA-1156 封装:Tray
PI7C9X20505GPBNDE 功能描述:外围驱动器与原件 - PCI 5port 5lane PCIe PacketSwitch RoHS:否 制造商:PLX Technology 工作电源电压: 最大工作温度: 安装风格:SMD/SMT 封装 / 箱体:FCBGA-1156 封装:Tray
PI7C9X20508GPANDE 制造商:Pericom Semiconductor Corporation 功能描述:5-PORT, 8-LANE, GREENPACKET- PCIE PACKET SWITCH - Rail/Tube