PCI

Peripheral Component Interconnect (PCI)

The PCI bus was developed and introduced by Intel at the same time the second-generation Pentium processor was released in 1992. Being a local bus, PCI is very similar to the VL-Bus it replaced except for one important difference.

Rather than tap directly into the delicate processor bus the way VL-Bus did, with PCI a special set of control circuitry was integrated into the Northbridge of the system chipset. This arrangement maintains the delicate processor bus while also maintaining a stable and high-speed local interaction between high-speed devices and the CPU (it also maintains the direct memory access (DMA) from PCI devices that VL-Bus had).

This allowed PCI devices to directly access the system memory without the direction of the CPU (DMA transfers are done under the direction of the system logic controller, AKA the Northbridge). First generation Pentium system bus ran on a 66Mhz bus, which was divided down to 33Mhz for the PCI bus using a clock divider circuit.

The PCI bus in practice

The PCI bus was a fantastic solution for video adapters since it maintained the high bandwidth of the VL-Bus, while providing better control of the expansion cards and their resources, all without loading down the system processor. PCI also did not have the electrical limitations present on the VL-Bus.

Extensions upon the PCI bus specification would expand the bus to 64-bit, and increase the bus speed to 66Mhz. On the other hand, in consumer PCs the PCI bus is only 32-bits, operating at 33MHz (providing a total of 132MB/s of data throughput).

The PCI bus was very fast when it was released, in reality it was actually too fast for most expansion card devices. Only video accelerators and storage cards truly took advantage of the PCI bus's speed. The ISA bus was still a more cost effective solution for sound cards and modems because it was cheaper to build, while offering more than enough bus speed for the cards to use.