|
Devices such as keyboards and mice are so common that their controller
circuitry and connectors are built right into the motherboard.
Even audio, video and LAN capabilities - at one time considered
"specialized" devices - are now built into most motherboards. Such motherboards are designed for
their simplicity, and will usually decrease the cost of the overall
PC. However, it makes the system more expensive to repair,
since a faulty component will usually require the entire motherboard
to be replaced. However, integrated motherboards are not always
suited for specialized applications. For example, the data
transfer rate for integrated video (133 MB/s) may not be enough for
games or high-end graphics applications. In addition, you may
want to connect a specialized device that does not have a suitable
port or connector on the motherboard, for example, a Firewire
camcorder. For these devices, the controller circuitry and connectors are built into special circuit
boards called expansion cards (or controller cards, or adapter
cards), which are physically inserted into special “expansion slots”
on the motherboard.
The narrow channel in each expansion slot contains tiny copper
contacts, which are connected to expansion buses. When an
expansion card is inserted into the expansion slot, a connection is
made between the contacts in the slot and the copper “fingers” on
the expansion card. This allows the expansion card to use the
associated expansion bus to communicate with an appropriate support chip,
which ultimately allows the device to communicate with the
CPU.

Expansion Bus Evolution
| ● |
ISA (Industry Standard
Architecture) |
| |
The first (non-proprietary) expansion slots were developed by IBM for
the IBM PC in 1981. The expansion slot (and its supporting
bus) was given the name ISA (Industry Standard Architecture), and as
the name suggests, became the industry standard for the next 10
years.
The supporting ISA bus had a very low data transfer rate (16 bits @
8 MHz = 16 MB/s). However, at the time, expansion slots were only used to
connect low-speed peripherals such as mice, modems and sound cards,
and so the data transfer rate was sufficient.
|
| ● |
PCI (Peripheral Component Interconnect) |
| |
In the early 1990s, processor speeds were increasing, and applications
were switching from text-based to graphics-based. The ISA bus
became a major bottleneck to the overall system performance
(increasing the speed of the CPU accomplished little if it was
always waiting for a slow bus to transmit data). The solution
was to create a new type of expansion slot that would support a much
faster bus.
The result was the Peripheral Component Interconnect (PCI) bus,
developed by Intel in 1992. It had a much higher data rate than the
ISA bus (32 bits @ 33.3MHz = 133 MB/s), and could therefore
be used for high-speed devices such as video adapter cards and
network cards.
Over the years, the PCI standard was enhanced to include a 64 bit bus
and faster bus speeds; however, these were found primarily on
motherboards intended for server PCs that required high data
transfer rates for multiple storage devices (eg, SCSI / RAID). Modern motherboards
usually include one or two standard PCI slots (32 bits @
33.3MHz), due to the abundance
of PCI devices still in existence. However, as these
devices begin to disappear, the PCI slot will soon join the
ranks of ISA as obsolete technology.
|
| ● |
AGP (Accelerated
Graphics Port) |
| |
By the mid-1990s, most video adapter cards used a PCI slot.
However, with computer graphics becoming more and more
sophisticated, the standard PCI bus was no longer fast enough to
transmit video data to and from the CPU. In addition, video
adapter cards could not supply enough video RAM for graphics
applications involving 3D acceleration and full-motion video.
In response, Intel developed the Accelerated Graphics Port (AGP)
in 1997. The initial version of AGP used a 32-bit bus running
at 66.6MHz, giving a data transfer rate of 266 MB/s.
However, the main advantage of AGP was that its bus was attached to
the northbridge chip rather than the southbridge. This allowed
data to travel more efficiently between the video adapter card and
the CPU, resulting in better performance. AGP was also able to
utilize the PC's RAM for doing its calculations. As a result,
large amounts of expensive video RAM were no longer required on the
video adapter card.
AGP was later enhanced to allow signal rates of 2, 4
and 8, giving data transfer rates of 533MB/s,
1 GB/s and 2 GB/s respectively.
However, the emergence of PCI Express has rendered AGP
virtually obsolete, as new motherboards rarely include
an AGP slot.
|
| ● |
PCI Express (Peripheral Component Interconnect Express) |
| |
PCI Express (PCIe) slots
are the newest and most advanced expansion slots found
on motherboards today. PCI Express was developed
jointly by Intel, IBM, Dell and HP in 2004. PCI Express is a
serial technology, similar to USB. In
other words, a PCIe bus transfers a single bit at a time (unlike a
parallel bus, which transfers 32 or 64 bits at a time). Serial
data transfer eliminates many problems associated
with parallel transfers, such as skew (bits arriving at
different times) and jitter (voltages floating above or below
target levels). This allows serial bus speeds to be
increased significantly, and therefore maintain relatively high
data transfer rates.
In fact, PCI Express uses a super high-speed serial bus running
at 2.5GHz. Using a signaling rate of 2, and allowing for 20%
overhead processing, this gives a data transfer rate of 500 MB/s -
which is four times greater than standard PCI, and twice the rate of
standard AGP.
In addition, there are five categories of PCI Express, each with
an increasing data transfer rate as follows:
|
Category |
Data Transfer Rate |
PCIe x1 (read "PCIe by one")
PCIe x2
PCIe x4
PCIe x8
PCIe x16 |
500 MB/sec
1 GB/sec
2 GB/sec
4 GB/sec
8 GB/sec |
The numeric designator in the PCI Express category defines the number of
serial "lanes"
that are being used to carry data. Therefore, PCIe x16 has 16
serial lanes, each with a data transfer rate of 500 MB/s, giving an
overall data transfer rate of 8 GB/s!
Using multiple serial "lanes" to carry data is similar to a
point-to-point network topology. At the source, data is broken
up into smaller units called "packets". Packets are then
numbered, and sent to the destination via the serial "lanes".
When they arrive at the destination, they are re-assembled into the
original data (this accounts in part for the 20% overhead
processing). Therefore, the more lanes, the more data can be
moved per second - which is why PCIe x16 (at 8 GB/s) is primarily
used for high-volume graphics adapter card.
The number of serial lanes also determines the physical size of
the expansion slot, with PCIe x1 being the smallest (about 1" long),
and PCIe x16 the largest (about 3.5"). In addition, a PCIe
card will physically fit (and work correctly) in any slot that is at
least as large as it is. As a result, a PCIe x16 card can only
fit in a PCIe x16 slot, whereas a PCIe x1 card can fit in any size
PCIe slot.
The increased data transfer rates of PCI Express has
led to a gradual "phasing out" of all other types of
expansion slots. For example, most motherboards
today have one or two PCIe x1 slots, which are used for
moderate-volume devices (modems, sound cards, network
adapters etc.) that were previously connected using a
standard PCI slot. Most motherboards also have a
single PCIe x16 slot, which has virtually replaced the
AGP slot for connecting high-speed video adapter cards. |

A motherboard with two ISA slots (black), five PCI slots (white)
and one AGP slot (brown). |

A motherboard with two PCI slots (white), one
PCIe x1 slot (small green), and one PCIe x16 slot (large green) |
|