• No results found

• Bandwidth of 2.5 Gigabits per second

N/A
N/A
Protected

Academic year: 2022

Share "• Bandwidth of 2.5 Gigabits per second "

Copied!
20
0
0

Loading.... (view fulltext now)

Full text

(1)

P.C.I. Express

Sushil Chauhan 01005002

(2)

Introduction

• High speed, serial pathway.

• Dual – unidirectional paths.

• Bandwidth of 2.5 Gigabits per second

per direction.

(3)

Benefits of PCI Express

• Layered Architecture.

• Compare maximum theoritical bandwidths.

™ PCI : 32-bit bus at 33 MHz , 132 MB/sec

™ PCI-X : 64-bit bus at 66.66 MHz , 533 MB/sec

™ PCI Express : Bandwidth of 2.5 Gigabits per

second per direction and potential for growth to 10 Gigabits/sec/direction.

• Point-to-Point Interconnect.

(4)

PCI Express Architecture Overview

• Links and Lanes.

• Serial links : x1 , x2 , x4 , x8 , x12 , x16 and x32.

P O R T

O R T P

Link Lane

Device A Device B

(5)

• PCI Express Device Types

™ Root Complex

™ PCI Express to PCI Bridge

™ Endpoint

™ Switch

• PCI Express Transactions

™ Request and Completion

(6)

PCI Express System Architecture

CPU

Memory Root Complex

Switch PCI Express to

PCI Bridge

PCI Express Endpoint

PCI Express Endpoint

PCI Express Endpoint Legacy

Endpoint

Legacy Endpoint PCI

PCI Express

PCI Express PCI Express

PCI Express Express PCI

PCI Express

PCI Express

(7)

Transaction Types

™ Memory Transactions

™ I/O Transactions

™ Configuration Transactions

™ Message Transactions

(8)

Architectural Build Layers

™ Transaction Layer

™ Data Link Layer

™ Physical Layer

(9)

Transaction Buildup through Architectural Layers

Data

Data

ECRC

ECRC LCRC

Header

Header Sequence

Number

At Transaction Layer

At Data Link Layer

Optional

(10)
(11)

Transaction Layer Architecture

• Create PCI Express request and completion transactions.

• Both Transmit and Receive Functions.

• Receives request data from Device Core.

• Receives incoming transactions from Data Link Layer.

• Uses Transaction Layer Packets for

(12)

Transaction Layer Packet (TLP)

• Request and Completion Information is communicated.

• Transaction Layer

™ Generates outgoing TLPs.

™ Accepts incoming TLPs.

• Transaction Layer Packet consists of

™ Header

™ Data payload

™ TLP Digest or ECRC

(13)

Data Link Layer Architecture

• Serves as gatekeeper for each individual link.

• Ensures that each packet makes it across the link.

• Takes TLPs from the transmit side of the Transaction Layer.

• Adds a sequence number and an LCRC.

• Accepts the packets from Physical Layer.

(14)

Data Link Layer Packet (DLLP)

• Originate at the Data Link Layer.

• Intended for the Device on the other side of link.

• DLLPs are of four types :

™ Ack DLLP

™ Nak DLLP

™ FC DLLPs

™ PM DLLPs

(15)

Physical Layer Architecture

• Sub-blocks of Physical Layer :

™ Logical Sub-block

™ Electrical Sub-block

• Logical sub-block has separate Transmit Unit and Receive Unit.

• Electrical sub-block has separate

Transmit and Receive buffers.

(16)

Transaction Unit Receive Unit Transmit Unit Receive Unit

From/To Data Link Layer

Packet Packet

Scrambled De−Scrambled

Packet

Removed Framing Framed

8−Bit / 10−Bit 8−Bit / 10−Bit

Encoded Decoded

Electrical Sub−Block Logical Sub−Block

(17)

Flow Control

• To prevent receiver buffer overflow.

• Local to each link.

(18)

• Virtual Channels.

• Traffic Classes.

• Each VC has its own set of queues and buffers and control logic.

• Supports 8 different Traffic Classes.

• Flow Control Rules.

™ PCI works as Single Lane Bridge.

™ PCI Express Flow Control Model works as Highway with four lanes in both directions.

(19)

• Flow Control at the Transmitter.

™ Credits-Consumed

™ Credits-Limit

• Flow Control at the Receiver.

™ Credits-Allocated

™ Credits-Received

• An Example of Flow Control Credits.

™ B indicates 4 PH Credits and 40 PD Credits.

™ A sends 2 requests, 1 PH + 10 PD units each.

(20)

Thank you .

References

Related documents

• Packet generation at MAC layer to remove Packet generation at MAC layer to remove  additional delays in generation from upper 

According to client requirements and available link bandwidth, encoding rates are decided per feedback interval and data is sent to clients at this encoding rate.. Here we will

The network access layer corresponds to the physical and data link layers.. The Internet Layer corresponds to the OSI

• For a 6-pole dc armature with 16 slots having two coil sides per slot and single turn coils, calculate the relevant pitches for a wave winding and draw the. developed

Routing Transport Layer : provides a reliable flow of data between two hosts Application Layer : handles the details of the particular application..

This project has been done on the Link Manager Protocol layer of the bluetooth which basically manages the logical link that is established between the devices before the

It provides bi-directional, low- cost and high speed serial interface for data transfer.. Multiple devices can be attached through a hub to the

The number of transactions versus bandwidth utilization is shown in Fig. 10, we see that bandwidth utilization increases as transaction rate increases. The proposed