Fast Forward to 1997. Intel changes the CPU socket standards and makes the new slot a trade secret. Everybody sing along: "Everything Old is New Again!"

In a bid to prevent even fractional erosion of its CPU dominance in the consumer hardware market, Intel is making sure that they have all of the marbles. Let’s follow how they won them:

Two types of sockets have dominated the CPU scene: pin sockets and snap-ins. Early in the 8088 days, the CPU was soldered to the board. Who could ever imagine upgrading? But with the advent of the 80286, upgrades and hardware speed "tweaks" were becoming a viable alternative to shelling out more bread. (Just ask PAUG contributor Matt Hill, who used to entertain himself by overclocking CPUs to wring higher performance out of them. Matt, I’m glad your hair is growing back.) This was the day of the side pins on a CPU. Little wires stuck out of the four edges of a chip, and you pressed the whole thing into a socket with your thumb, hoping it didn’t break.

The 80386 introduced the standard push-in socket that was a series of tiny holes with conductors in them. You lined up the chip (which had an array of pins sticking out of its bottom side) and pushed it into place, again, with your thumb. The alternative was an over priced mechanical thumb that didn’t misalign the CPU pins as often as manual insertion, but if you bent a pin either way, it was a very expensive mistake.

Since CPU and motherboard connectors were not a very hot or disputed topic in those Halcyon days, the issues of who set the standards were pretty much left for standards bodies comprised of manufacturing interests. Intel put their two cents into these panels and even contributed to the proliferation of the ZIF (Zero Insertion Force) Socket during the reign of the 80486. ZIF let you drop the bottom-pinned design into a series of very loose holes which then tightened and made contact when you moved a small lever. An excellent contribution, to be sure.

It was about this time that semiconductor maker AMD stepped into the fray and made 386 and 486 CPUs that used Intel designs. Intel promptly sued AMD, only to discover that the wee company had Intel by the throat: Intel had licensed off some of its lesser technologies to AMD and that company applied the sloppy wording of the contract to the x86 CPUs as well. Intel took a fall in court. Embarrassing.

Not that anyone much cared. AMD’s facilities were so small that the competing chip was never produced in much quantity, and Intel was driving ahead with its OverDrive versions of the 486. AMD looked like a hiccup in the tight control Intel had on the market. That was, until the introduction of the Pentium.

The Pentium was a very different chip design. It expanded by 30% the number of pins on the chip, so a new socket (called Socket 7) became available to support the larger CPU. Pentium also eschewed the 80x86 architecture, leaving behind its vulnerability to AMD & Cyrix.

Yet AMD and Cyrix had gotten smart while Intel had been in its Uncle Scrooge vault counting its dollars. Knowing that mighty Intel would not let another licensing issue slip by, Heckle and Jeckle developed their own plan: beat Intel with its own stick. After Intel had abandoned the 486 market for the greener fields of Pentium technology, AMD & Cyrix kept making faster and faster CPUs for that old system and socket. These 486 based CPUs performed as well their rival, Pentium, were produced. Fresh from their success at making a better mouse trap, both AMD & Cyrix built their own CPUs on architecture that was not Intel based, but still ran in a Pentium socket. The performance gap began to close.

Then Intel stumbled. The famous bug in the Pentium Math Co-processor let AMD and Cyrix pull right up to Intel’s back bumper in performance, and still maintain their distance in price. With the current popularity of AMD’s K6 MMX processors, Intel suddenly had more to worry about.

With most of the world already running on Intel CPUs, the behemoth has had few places to turn to make more money. So, reasoning that if they could not influence the number of computers sold, Intel could increase the number of their components sold per computer. As of 1994, the company started rolling out motherboard chipsets that supported CPU functions. Since the functions that the motherboard chipsets were supporting are based on the original Pentium specification (which AMD & Cyrix could mimic, but not duplicate) there wasn’t a way to use this new market leverage to squeeze the competitors off the road.

Until Now.

Intel, claiming that Socket 7 had reached its performance ceiling and there was no point in making faster CPUs that would be bottlenecked by a slow bus design on the motherboard, began designing its new systems around a radically different processor and interface design called Slot 1.

To Socket 7, Intel has said, "Adios Tonto, and the horse you rode in on!" The pin cushion on the bottom of the CPU is gone. The ZIF socket is gone. Instead Slot 1 is a thin slot on the motherboard (not unlike a video card slot) that accepts a CPU which has all of its connectors on one edge. This strange piece of work has a push in connection that makes changing CPUs "as easy as changing a light bulb." Remember Bell & Howell? Indeed, Intel is saying more than they know.

Yet there are things about Intel’s claims that don’t stack up:

1. Socket 7 has not reached its performance limitation. A 300 Mhz AMD K6 MMX CPU was recently demonstrated running on a plain vanilla Socket 7 motherboard.
2. Socket 7 physical architecture may limit the bus speed, but the logical and electrical architecture doesn’t. Again, a recent demonstration showed a Non-Intel CPU running on a motherboard that had its bus speed stroked up to 100Mhz.
3. Intel just released their MMX Pentiums to the market, but has no intention to take the speed any higher than 233Mhz for these Socket 7 chips. To go faster, you must have a Pentium II on a Slot 1 system. That’s right: only on the proprietary Intel system can you go faster. Intel could chop off their own new product to force the market to migrate to their proprietary interface because rolling the legal dice with Janet Reno is safer than competing on level ground with AMD or Cyrix. Just ask Bill Clinton. (Coffee, anyone?)
4. Intel had, for a time, pitched the Pentium Pro (which runs on a modified Socket 7) as the "next big thing." But Pentium Pro turned out to be sort of a dud, and Pentium II has the same problems. (For the low down on these and other CPUs, see PAUG May ’97.) If Pentium Pro is a dead end for upgrades, why is Intel still pushing them?

Research & Development money must be recouped somehow, and if it is possible to sell more by simultaneously slowing down innovation and encouraging migration to a new proprietary system to boost profits, why not?

Instead of just stopping at the CPU level, though, Intel is branching out. Remember I mentioned the motherboard chipsets that provide support functions to the CPU? Well, Intel’s next move is to buy long-time video card maker Chips & Technologies to enhance their move into the Advanced Graphics Port (AGP) arena. This is not a CPU function. Not surprisingly, the Federal Trade Comission is staring down Intel’s neck during this maneuver.

Fortunately for the PC community world wide, AMD and Cyrix are not taking this blatantly monopolistic behavior lying down. AMD has recently partnered with Digital Equipment Corporation to build a Slot 1 system that will be physically identical, but electrically different. Credit AMD for walking right up to Intel’s front door and flicking a booger on the glass.

Cyrix’s response has been more along the lines of an attack by paper tigers. National Semiconductor is in the process of merging with Cyrix, and NS has an old licensing agreement with Intel that could be legal justification for access to the Slot 1 patent, much the same way AMD and Cyrix were able to use the 386 & 486 designs based on old or poorly written contracts. Here’s hoping.

Finally, let me assure you that you can buy AMD or Cyrix hardware with confidence. Remember Moore’s Law? "Speed and capacity double every eighteen months." If your computer buying habits revolve around this axiom, you can’t go wrong. The AMD K6 MMX CPU is a Socket 7 chip, but blows the door off of an equivalent speed Intel Socket 7 CPU. It is safe to buy now even if Intel pulls the plug on Socket 7 Pentiums next year, because AMD is pushing out a 300Mhz system in the near future, and Pentium II CPUs cost about $900 for just the chip.

With everyone from Microsoft and Intel, right down to your corner grocer under investigation by the government, perhaps the people who use their machines should be more directly involved. Will you let Intel chart the future? Be assured, left to themselves, they won’t "leave the light on for you."

Peace,

Webwalker