|
PCI Bus
The PCI local bus
(Peripheral Component Interconnect bus) is now the standard local I/O bus,
not only with Pentium CPUs but also with RISC CPUs. Standard PCI has a
32-bit data path and runs at 33 MHz when the systemboard runs at 66 MHz.
However, the PCI
specifications can also use a 64-bit data path and can run at a speed of
66 MHz when the system bus runs at 133 MHz. Also, an addendum to the PCI
specifications, called PCI-X, enables the PCI to run at 133 MHz.
One advantage of the
PCI local bus is that devices connected to it can run at one speed while
the CPU runs at a different speed. Devices connected to the VESA bus must
run at the same clock speed as the CPU, which forces the CPU to endure
frequent wait states. The PCI bus expansion slots are shorter than ISA
slots as shown in the figure below, and set a little farther away from the
edge of the systemboard.
Bus Master
The PCI bus also
supports bus mastering. A bus master
is an intelligent device (i.e., it has a microprocessor installed that
manages the device) that, when attached to the PCI bus, can gain access to
memory and other devices on the bus without interrupting the action of the
CPU. The CPU and the bus mastering devices can run concurrently and
independently of each other.
Throughput
Performance
Because of the
effective design of the PCI bus, the throughput performance or the data
transfer rate per second is 132 MB. Throughput
performance or data
throughput
is a measure of the
actual data transmitted by the bus, not including error-checking bits or
redundant data.
Bus Comparisons And
Bus Speeds
The tables below
compare the buses discussed in this section. Most often when comparing
buses users focus on the width of the data path and the overall bus speed.
However, you also should consider the type of expansion slot the bus
allows. The number of fingers on the edge connector of the expansion card
and the length of the edge connector are determined by the bus that
controls that expansion slot. Various bus connections are shown in the
figure below.
Wait States
A wait state isn’t a
systemboard component, but it is an idea that you need to understand in
the overall performance of a computer system. A
wait state occurs when the CPU must wait for some other
component to catch up, for example when slower dynamic RAM reads or writes
data. To allow time for the slow operation, the CPU is told to maintain a
wait state by CMOS setup information.
If normally the CPU can
do something in two clock beats, it is told to take an extra clock beat
for a total of three clock beats. It works for two beats and then waits
one beat, which makes for a 50-percent slowdown in speed.
<Previous>
<Home>
<Next>
|