On Wed, 10 May 2006 23:39:19 -0300, "Carlos CZ"
Hi all, I have sent the page to some of you at your email accounts,
Yes, I got it, thanks, and I've replied by email.
I think they are saying AGP "replaces" PCI in the context of graphic
cards, because it's more effective - but you can use a PCI graphics
card instead of (or in some cases, as well as) an AGP graphics card.
I believed this scheme was correct, because if you are able to speed up
video signal, why not to try to do the same for other components ?
Part of the reason why they could get AGP to go faster than PCI in the
Pentium II era, was because it was dedicated to a single device.
It's far more challenging to support multiple devices on a true bus,
because devices can steal each other's bandwidth, and the flow from
each device has to be arbitrated so they don't get mixed up. There
may be raw electrical considerations, such as certain combined states
either shorting out something, or drawing too much current.
Another factor is that since the days of Windows 3.1, it has been the
need for faster graphics throughput that has driven faster busses.
In the DOS days. marginal improvements above ISA (for all cards) were
by and large ignored; when Windows 3.1 came out, the need was such
that mutually-incompatible proprietary "local bus" solutions gained
traction in the marketplace. VESA quickly brought out the VL-bus
standard that supported multiple cards of any purpose, but mainly
folks used VL-bus graphics and ISA for everything else.
PCI was a far more complete and robust development - it wasn't clocked
faster than VL-bus, but added a depth of intelligent device management
that later formed part of the basis for Win95's Plug-n-Play.
History is now repeating itself!
First, we have the single-device AGP, this time developed purely for
graphics, and now we are adopting the more robust PCI Express.
BTW, "legacy" PCI was indeed developed to provide more speed and bus
width (66MHz, 64-bit) for all cards, but this hasn't taken off on the
PC. I think some of the Dec Alpha systems used it.
But, now I can do only naive speculations. Perhaps, what I see for
"AGP for other components" is being accomplished by the ISA card.
ISA = Industry Standard Architecture, and is the oldest of the PC card
busses; it dates from the original PC (ISA-8) and 286 (ISA-16).
ISA was either 8- or 16-bits wide, and was clocked at the same speed
as the processor (starting with 4.77MHz for the first PCs). As
processor speeds increased, the ISA cards couldn't keep up, and the
speed was un-linked via divisors so as to clock the ISA at 8MHz or
10MHz. Most ISA cards are not rated above 10MHz.
EISA was 32-bit and clocked slighty faster, and (like VL-bus) it
allowed ISA cards to be used in the same slots. It was a response to
IBM's failed attept to regain lock-in of the PC's design via thier MCA
(Micro Channel Architecture) bus. The speed boost was similar, but
unlike EISA, you could not use ISA cards in MCA slots.
VL-bus was also liked to the processor speed, and at that time the
standard base processor speed was 33MHz. Faster speeds would stress
VL-bus cards, so you either could use only one VL-bus card, or there
would (once again) be a clock divisor. It was 32 bits wide.
PCI was 32-bit and supposed to always be 33MHz, irrespective of
processor base speed. It's a fundamental difference to VL-bus.
AGP is always 66MHz, though with data transferred at 1x, 2x, 4x or 8x
this clock speed. The same tactic may be behind PCI Express's speed
boost, as implied by the "16x" name for the long graphics slot.
There's no "AGP for other components". My Spanish (?) isn't up to
really understanding the page you sent me, but as far as I can tell,
it was referring to AGP replacing PCI purely in the context of
graphics, not for other cards.
There are no motherboards with AGP that also do not have PCI too.
------------ ----- ---- --- -- - - - -
Our senses are our UI to reality
