Communicating with Devices: Buses, Bridges, etc.

1
Communicating with Devices Buses, Bridges, etc.
2
Look to the Future

• From the beginning PCs have been made with
  expandability and adaptability in mind.
• While certain I/O devices came standard, there
  was room to add others to expand the computers
  capabilities and to adapt the PC to the
  particular needs.
• The circuitry required by the new device is
  placed on the expansion card or adapter card.
• We also need to connect to and communicate with
  the rest of the PC, this is the job of the
  expansion bus.

3
Standards

• IBM developed the first PCs, the first PC bus and
  the first PC expansion slots. It was an 8-bit
  bus. They allowed others to use that technology
  without having to pay royalties. (However, they
  never released the full specifications.)
• This step allowed competitors to make IBM
  clones which brought the price of PCs down. But
  the low price helped to establish a standard.
• The standard made peripheral makers jobs easier.
  They only had to design for one interface.

4
Bottlenecks

• The standard devices and expansion devices
  operate at a wide variety of speeds. We need a
  design that allows to whatever extent possible
  each device to operate at its own speed.
• Actually we will have a hierarchy of standard
  speeds different buses carrying data at
  different speeds.
• The interface between devices of different speeds
  can be a source of slow downs (bottlenecks).
• The chipset handles much of this interfacing.

5
Chipset

• The main components of the computer (processor,
  memory, cache, etc.) plug into the motherboard
  and communicate via the chipset.
• The chipset determines which type of processor
  can be used, how fast a processor, how much
  memory, what type of memory, and so on.
• The chipsets function are typically broken into
  two sets, one handled by the Northbridge and the
  other handled by the Southbridge.

6
Northbridge

• The Northbridge handles the high-speed,
  high-volume data communication between the CPU,
  cache, memory and AGP (accelerated graphics
  port).
• Since the Northbridge manages the
  processor-memory interaction (the front-side bus
  bottleneck), its performance is critical.
• The Northbridge determines the FSB speed.
  Typically the processor operates at a
  multiplicative factor (the CPU multiplier) of
  the FSB speed.
• It uses synchronous memory, recall SDRAM.

7
Southbridge

• The Southbridge allows communication with a
  greater variety of slower devices. It connects
  to the secondary buses USB, IDE, PS/2,
  Ethernet.
• Since it handles accessing the hard disk, the
  Southbridge manages a bottleneck of its own,
  especially for memory-intensive applications.

8
(No Transcript)
9
North
South
10
Bridge Interface

• Older chipsets used the PCI (Peripheral Component
  Interconnect) Bus to connect the Northbridge and
  Southbridge. But with more devices using the PCI
  bus, modern chipsets have introduced a dedicated
  interface between the chips.
• Another multiplicative speed factor is between
  the FSB speed and the PCI speed.
• The standard PCI bus speed is 33.3 MHz, while
  standard FSB speeds are 100 MHz (3X) and 133 MHz
  (4X).
• Motherboards that use this multiplicative factor
  are said to be synchronous.

11
Bus History

• When Intel moved to a 16-bit data bus (with the
  256), the expansion buses needed to be changed as
  well.
• IBM developed a 16-bit bus that was backward
  compatible with the PC 8-bit bus. They debuted it
  in their Advanced Technology (AT) computer and it
  became known as the AT bus.
• Based on the AT bus, some IBM competitors got
  together to lay out some specifications known as
  Industry Standard Architecture (ISA).

12
No more ISA

• ISA slots were kept in PCs for reasons of
  backward compatibility for many years.
• They have been officially dropped from what
  Microsoft and Intel see as a PC in their PC 2001
  System Design Guide.
• This move is to reduce legacy dependence.
• If needed one can get PCI cards that can simulate
  an ISA card to interface with legacy devices.

13
Link to PC 2001 System Design Guide
14
PC 2001 System Design Guide
15
Communication

• Expansion cards, as well as other peripheral
  devices, can communicate with the processor and
  memory through one or more of the following
• I/O Addresses
• Interrupt Requests Lines (IRQs)
• Direct Memory Access (DMA) channels
• Memory Addresses
• Known as system resources.

16
I/O Address

• The processor has an external data bus and an
  address bus.
• One use of the address bus is to indicate the
  word in memory that the processor wishes to
  access.
• A set of address bus combinations are reserved to
  communicate with peripherals. A given peripheral
  may have several addresses assigned to it, each
  corresponding to a different action that the
  processor is requesting.

17
Uniqueness

• As is normal with addresses, each I/O address
  should be unique. If two devices shared an I/O
  address, then it would be unclear which device
  the processor intended to communicate with.
• IBM set the early standards assigning certain
  standard devices set I/O addresses.

18
Start/Settings/Control Panel/System/
Hardware/Device Manager
19
Click to expand items under a particular device
then double click on the item
20
Resources
21
(No Transcript)
22
IRQs

• While I/O addressing allows the processor to
  initiate communication with a peripheral, the IRQ
  allows the peripheral to initiate communication
  with the processor. The processor must be
  interrupted.
• The processor has one pin to indicate that it has
  been interrupted. When it reaches a point at
  which it can process the interruption, it needs
  to trace back and determine which device
  interrupted it. This is the purpose of the IRQs.

23
Interrupt Controller

• After the processor has been interrupted, it uses
  the I/O Address approach to initiate
  communication with the interrupt controller which
  will help the processor determine which device
  made the request.

24
Table of IRQs
25
Another type of addressing

• IRQs are another form of addressing.
• There are far fewer of IRQs than there are I/O
  addresses and so it is more likely that IRQ
  conflicts will occur.
• Under certain circumstances it is possible for
  two devices to share an IRQ line provided the two
  devices are never active simultaneously (e.g. a
  fax and a modem that use the same phone line).
• Early on IBM assigned to IRQs to some standard
  devices.

26
Table of IRQs (Cont.)
27
NIC using IRQ 9
28
Ports

• To accommodate that many devices would need I/O
  addresses, an IRQ and a set of rules for
  communicating, two standards were established
  one for serial communication (the COM ports) and
  another for parallel communication (the LPT
  ports).

29
COM and LPT Assignments
Port I/O Base Address IRQ
COM1 03F8 4
COM2 02F8 3
COM3 03E8 4
COM4 02E8 3
LPT1 0378 7
LPT2 0278 5
30
DMA

• Direct Memory Access (DMA) allows devices to
  interact with the memory without the processor
  having to get involved.
• The processor delegates some of its traffic
  control duties.
• Now we have a situation in which a number of
  devices may want to use the buses to access
  memory, so we need another addressing/ requesting
  system.
• Note with our cache hierarchy, we hope that the
  processor accesses the memory infrequently (less
  than 5 of the time).

31
DMA Channel for Floppy
32
Bus Mastering

• Choosing to show the floppys DMA channel was not
  just a random example. The floppy uses classic
  (third-party) DMA.
• Third-party DMA was slow.
• In first-party DMA, a.k.a. bus mastering, the
  device (e.g. hard disk drive) controls the
  interaction with memory.
• The devices sense if some other device is
  accesses memory.

33
Changing View of the Device Manager
34
DMA
35
I/O Addresses
36
IRQs
37
Memory-Mapped I/O
38
Buses that have come and gone

• Micro-Channel Architecture (MCA) extends to
  32-bit width, increased speed, was
  self-configuring BUT was not backward compatible
  and was proprietary
• Extended ISA (EISA) was 32-bit, self-configuring,
  backward compatible, open standard BUT slow
• Video Electronics Standards Association Local Bus
  (VL-Bus or VESA Bus) was fast because it worked
  with the North bridge instead of the South Bridge
  BUT was NOT self-configuring.

39
And the Winner is PCI

• Peripheral Component Interconnect (PCI)
• Intel introduced the PCI bus structure pretty
  much at the same time it introduced the Pentium.
  
• Like the VESA bus, it interfaces with the
  Northbridge so it is fast.
• It is also self-configuring.
• It allows for bus mastering.
• It was effectively free (provided automatically
  by Intel).
• It allowed data bursting.
• It was scalable.
• While introduced by Intel, it works with other
  processors including Machintosh.

40
Bus Comparison Table
41
PCI Card Connectors
A PCI card has 47 pins with no bus mastering and
has 49 pins with bus mastering.
42
PCI Slots
43
Putting a card in

• Keep the card in its anti-static bag until you
  are ready to insert it.
• Turn off and unplug the PC.
• If you dont have a wrist strap to protect
  against electrostatic discharge (ESD) then after
  removing the card from the bag touch the PCs
  power supply. The smallest shock could ruin the
  card or the motherboard.
• Insert the card straight in or at most a slight
  angle.
• Screw the card to the box with a connection screw
  it keeps the card in place and can help ground
  it.

44
Putting a card in (Cont.)

• Close the computer, plug it in and start.
• You should get a message from the operating
  system that it has detected new hardware.
• It will search for a driver and/or request that
  you insert a CD with the driver.
• You may need to reboot depending on the device.
• You should look for the new device in the device
  manager to make sure there are no conflicts.

45
Hardware Compatibility List
46
Look for Compatibility between device and
operating system
47
Plug and Play

• Plug and Play (PnP) is a feature in which the
  system configures a device automatically rather
  than the user having to set jumpers and choose
  resources, etc.
• For PnP one needs
• A PnP BIOS
• A PnP operating system
• A PnP device
• All PCI devices are PnP
• A device which is not PnP is called a legacy
  device.

48
How Plug and Play Works

• Initially the PnP devices are put on hold and
  resources are assigned to legacy devices.
• Then the system starts querying the PnP devices
  about the resources each wants.
• A PnP device will provide a list of resources it
  could use.
• For example, it would not specify IRQ 3 but
  rather specify that it could use IRQ 3 or IRQ 5
  or IRQ 7.
• A file with PnP information, called the Extended
  System Configuration Data (ESCD), is kept.

49
AGP

• Specification developed by Intel, even faster
  than PCI and more directly connected to the
  Northbridge is the AGP (Accelerated Graphics
  Port).
• It is dedicated to graphics. It will have its
  own slot to be used only for this purpose.

50
References

• All-in-One A Certification, M. Meyers and S.
  Jernigan
• http//www.pcguide.com
• PC Hardware in a Nutshell, R. Thompson and B.
  Thompson