Hi,
below is the artical which I am reading now. I did not understnad this
part
"The chipset speaks to the PCI bus at exactly HALF the
speed/frequency/tact, so that the CPU talks to all PCI devices through
the chipset with HALF the external tact or bus speed."
My questions are these
1) Which chipset is this??
There are two main chips on the motherboard that effectively tie all the
different parts together, ie. memory, PCI slots, ISA slots, AGP slot, etc.
These two chips are the 'chipset', known as the North Bridge and South
Bridge. They can be made by various people but they are designed to work
together. Popular makers are Intel, VIA, SIS, Nforce etc. For instance an
Aopen board could have an Intel chipset or a VIA chipset, both boards are
made by Aopen but the main two 'control' chips - the chipset - are made by
different people. Its a bit like a Ford car having Michelin or Dunlop tyres.
nice diagram of how the chipset ties the motherboard together here:
http://www.tomshardware.com/motherboard/20020311/sis645dx-01.html
2) Why the CPU need to go first to that chipset in order to talk to the
PCI devices??
The CPU is in charge and has lots to do. So the CPU asks the chipset to get
the information for it from the PCI cards, while the chipset is running off
with its orders the CPU can be doing something more important.
Looking back at the diagram from the above link you can see that the PCI
slots are connected to the CPU through the chipset. there is no way to
bypass the chipset.
On a related note; looking at the diagram you can see that if direct memory
access (DMA) is given to the drives they can get to the memory without
bothering the CPU, this is generally a good thing as the CPU can get on with
important tasks while information is transferred, instead of going
Drive - chipset - CPU - chipset - memory
it goes
Drive - chipset - memory
Hence DMA enabled = speed
3) and why the speed should be 1/2 if the CPU want to talk to the PCI.
Not sure on this one, but it could be many factors, CPUs are quick and
getting quicker, PCI is slow and not speeding up anytime soon, so it talks
slowly to the PCI slots so that every bit of information is understood. Also
PCs work in binary, (signal on or off), so dividing by two is easy to do to
a speed. It would be much harder to cut the speed by a third. Why should the
designers make it hard for themselves?