
click to enlarge
If you have opened a computer since the mid-nineties, you’ve seen a Peripheral Component Interconnect slot (pictured at right). What USB did for things outside your computer, PCI did for things inside it. Instead of having to make sure your motherboard could accept one of several, often bizarre, standards to plug in what it was you were buying, just about every motherboard has PCI slots and almost any expansion card would fit in it.
That was even more important back then, because motherboards didn’t come with all kinds of stuff built in, we currently take for granted. Network connections, and before that, even hard disk controllers, were not built into motherboards. Forget about built in sound. You had to drop in a card, and it always using the same kind of slot was great.
Sorry if I sound like an old man who has taken leave of his senses: “We walked 3 miles to school; in the snow; uphill; both ways. One year we didn’t have enough coal to run the modem…”
Anyways, at its zenith, the PCI bus could handle 133 MB/s. That is enough for even today’s sound and network cards… if you only have one. The problem is that all the PCI traffic goes back to the same bus and shares that 133 limit. A single Gig-Ethernet card needs almost all of it just to itself.
A larger slotted 64-bit version of PCI never got that popular for a pair of reasons. Things that really needed the bandwidth, like hard drive controllers and network connections, were by now being included on the motherboard (or the connected device), eliminating them from the PCI bus ridership. Also, the only thing that needed the bandwidth that wasn’t being moved to the motherboard (at least not universally, today’s integrated graphics still can’t handle many applications) was a video card. And those were using AGP slots.