Chipsets

There are more parts to a computer than those described so far. Some, like mice and keyboards, are obvious to the average user. But these also don't have much to do with core computer technology and, in consequence, won't be discussed here. The Suggestions for Further Reading provides references that will be helpful for those who want to know how these minor peripherals work. The chipset, however, is a central element of computing.

Chipset (also chip set) is a somewhat loosely used term that refers to a critical part of a microcomputer (all computers have something analogous). The chipset performs what is known as glue logic—an interesting and descriptive term for chips that bind different parts of the computer together. For example:

• If the chipset doesn't support a particular kind of memory, it doesn't matter if you have it because you won't be able to take advantage of it.

• If the chipset doesn't know about new buses like the USB and FireWire, the computer can't use them.

• If you want to run your memory bus at 133 MHz instead of 100 MHz, the chipset has to accommodate this.

The chipset also manages connections to the CPU. It has to be designed to support CPU-specific features like clock doubling, L1 cache, and so on. Physically, the chipset always sits close to the CPU and memory. The normal functions of the chipset include the interrupt controller, the memory controller, and the I/O controller (providing, for example, an interface to the PCI bus—see Figure 2.5).

You can see that the chipset is important. Early microcomputers achieved the necessary functions by using lots of what are called discrete logic chips—general purpose and off-the-shelf chips that were linked together so that they could complete a series of actions. The IBM PC had 63 of these chips. Since they were standard and easily available chips, competitors didn't have much trouble producing clones. But someone had a better idea. A company called Chips and Technologies (C&T, still around—recently purchased by Intel), decided to collapse many discrete logic chips into a much smaller chipset specifically designed to support the needs of the PC. C&T needed only five chips to handle the IBM PC's glue logic. Their chipsets were then sold to clone manufacturers in volume and for considerably less than the standard parts. Not only were the parts cheaper, but it was also less expensive to manufacture a computer with fewer chips. Finally, fewer chips meant greater reliability. A win-win-win situation.

C&T's success created a huge change in the market. Clone makers surged. The sudden availability of the space vacated by the original chips made smaller computers, including laptops, feasible. The cloners were themselves cloned—many vendors now began to produce chipsets that competed with C&T's versions. A good deal all around for most everyone except IBM (which eventually began making chipsets of its own, of course). The chipset market is much less competitive today, but before considering its current status, a brief digression to consider motherboards is in order.

The term motherboard denotes the part of the computer that everything else is attached to. Normally, the motherboard is on the side (tower systems) or on the bottom. The CPU, the memory, the chipset, and various other parts are attached to it. The bus or buses and other connecting circuitry are actually embedded in it. Drives, the power supply, and so on are attached to the case, which is in turn connected to the motherboard.

There isn't a lot of technology in the motherboard itself—that's all in the parts that sit on or around the motherboard. Still, that doesn't mean that motherboards are easy to make. Manufacturing is complex, and tolerances are quite high. Metal parts, such as the edge connectors on the bus (where the adapter cards are attached), are particularly tricky to make. The good news is that the motherboard isn't exposed to much action, so doesn't get the kind of wear and tear that would cause it to fail in long-term use. The bad news is that if you don't take tremendous care in manufacturing a motherboard, it won't work the first time. Higher bus speeds (e.g., 100 MHz+), as well as double data rate and synchronous timing, have really challenged motherboard makers.

The motherboard was a natural for process-oriented Asian producers and, by the mid- to late-1980s, a raft of Taiwanese companies had almost dominated the market (at least for clones—companies like IBM and Compaq continued to make their own). While there is no standard pattern in the kinetic world of microcomputers, the typical motherboard manufacturer added value to the product by including not only the chipset (often of its own design and manufacture) but also most of the other chips (memory, cache, and BIOS). Clone makers then had only to snap in a CPU and a couple of adapter cards, add a case with power supply and drives, stick the company logo on the front, and voila—the [your name here] PC. Most often, this latter stage was performed in the U.S., where it could be done by unskilled labor. To say that this process worked well is an understatement; clone-makers like Dell and Gateway 2000 used it in a quick climb from business no-names to multibillion dollar corporations that most everyone has heard of. One of the Taiwanese motherboard and chipset companies, Via Technologies, is emerging as a major player in computer systems. The others are still unknown outside of their home island, but what do they care; they're rich too.

Beginning in the early 1990s, Intel began to interfere in this cozy and prosperous business. Remember that the chipset is two-sided glue—it has to adapt both to new kinds of peripherals and to changes in the CPU. When Intel introduced a new CPU, the chipset makers had to wait for Intel to publish the specifications needed to create a chipset. This, of course, depended on Intel's cooperation. However, Intel, which hadn't licensed others to make its CPUs since the 80386 appeared in 1985, decided that it could further strengthen its market position by taking a part of the system action for itself.

Intel didn't want to offer Intel-brand microcomputers (it never has, except for some very limited, mostly scientific, applications) because to do so would be to compete very directly with its customers. Still, Intel could put more inside the computer than just the CPU. It could add a chipset of its own, for example. When Intel started offering chipsets, it put all its competitors in this market on the defensive. Even if Intel published the specifications of a new CPU, it would obviously still have a big headstart in adapting its chipset to it. Moreover, people would naturally assume that a product from Intel would connect more reliably to the CPU than would one from a manufacturer that, by definition, knew less about it. Companies that decided to buy Intel chipsets, like Gateway 2000, were now the first to market with new versions of Intel CPUs. From this point, the next step was quick and obvious—Intel built the entire motherboard and gave the clone makers all of the basic electronics, including the CPU, in one package.