Dodgy said:
It's just removing a future bottle neck before we hit it.
You wouldn't like it if your house was only wired to support 10 x
60watt lighbulbs and you needed a complete rewire and new fuse board
when you wanted to a brighten up a few rooms with 100watt bulbs!
However the current situation is the opposite, right now we're not running
on the ragged edge of what AGP can do, as test after test has shown, we're
using at most half its capacity.
The real push behind it is that Intel wants to sell chipsets and force Via,
SiS, etc to play catch-up. Don't kid yourself that anybody is going to see
a real performance improvement from it any time soon. Maybe 5 years down
the road, but by that time Intel will be trying to kill it off in favor of
some new marketing standard.
Remember all the wonderful things that Slot 1 was supposed to do that
couldn't possibly be done with a socket? The only wonderful thing it did
was allow Intel to prevent their interface from being cloned long enough to
lock AMD into a different interface. When Intel is pushing some new
standard put one hand on your wallet and the other on your pistol because
you're about to be robbed.
Now, there are some real reasons to want a faster PCI--the current version
can't keep up with gigabit Ethernet for example. But the simple fact is
that the rest of the machine can't keep up with it either unless you've got
it specified to the standards of at least a midrange server, so the PCI bus
is not the bottleneck there in any but a few specialized applications. The
claim is that a PCI-Express machine will give the same performance as a
PCI-X machine for less money. However the PCI Express slots being put in
the first generation of machines do not match the performance of PCI-X and
I would be very surprised if the first generation of PCI Express boards
sold for less money than the current generation of PCI-X boards.
