We have browser wars (Netscape vs. Internet Explorer), platform wars (Windows95 vs. Everything Else) and "Let's build a pretty case" wars by the likes of Packard Bell and Sony. We even have wildly popular war games like Quake! and MechWarriors in case we get tired of the industry wars going on in front of us. But like the man said, "Baby, you ain't seen NOTHING yet."
Enter the CHIP WARS. (Play your appropriate John Williams soundtrack here.) The forthcoming explosion of commercially available CPUs (Central Processing Units) is going to rock the average user's world the way Sitting Bull rocked General Custer's world at the Little Big Horn.
The players in this war game are: Intel Pentium with MMX, Pentium Pro, Pentium II with MMX, the AMD K6 with MMX, and the Cyrix MediaGX with MMX. Each of these contenders is wearing a different horseshoe in its boxing glove and will make for a lively (though commercially treacherous) next year.
For those of you who aren't hip to the acronym, MMX stands for Multimedia Extensions. The story goes like this: back in the dark ages, Intel laid down a set of instruction codes for its 8086 series of processors. Several years later, they added a few more instructions when they introduced the 80386. Since then Intel has stuck with increasing the number of transistors on the chips and shrinking the dies. But the set of instructions the chip responds to has remained unchanged.
The recent explosion of Multimedia requirements in the PC industry forced Intel to rethink its strategy. This influx meant more of the Multimedia intensive processing was being farmed out to accelerated video cards and specialized sound devices that helped take these fat data loads off of "the poor processor." Eventually, the industry reached a practical stand still in its capacity to process this information. No matter how fast an outside company made their video card, the card still got its instructions from the CPU for WHAT (though not HOW) to paint on the screen. The CPU was quick to send, and the video card got right to work as soon as it got the message, but there was a bottleneck in between: The motherboard.
So Intel took care of the problem by moving the HOW onto the motherboard by expanding the instructions that its Pentium Processor recognizes. These 57 new instructions are called the MMX.
While I could heave reams of data at you for which is the best chip, it is just easier to give you the breakdown of which, and why, and how and when.
The Pentium has been around since 1994 and is starting to show its age. MMX extends the life span of the Pentium a little longer, until Intel can move us to the next level. (More on that in a minute.) Because MMX machines also have a larger internal cache (where the CPU queues up instructions to be executed) it will run ALL programs faster than its non-MMX forerunner. But to really make it take off, you must have software written to invoke those new MMX instructions, and as of this writing, such software is as hard to come by as morals in Congress. Unfortunately, Autodesk chose not to get in on the ground floor of MMX, and the newly released AutoCAD R14 does not support MMX extensions.
Since the Pentium has reached its maximum practical speed of 200Mhz, Intel is hard at work shoving their next product out the door. Last year debuted the Pentium Pro, a significantly redesigned Pentium optimized for running 32 bit processes. Windows NT 4.0 fairly flies on this Processor but, as with all things in life, there are drawbacks: It runs 16 bit programs and processes SLOWER than a regular Pentium. That's right: the Pro isn't designed to run anything that doesn't have 32 bit in its blood. So your old programs that you wrote and compiled under Windows 3.11 will run like a dog on the Pro. Also, Pentium Pro is not extended with the MMX, though Intel has plans (codename: Klamath) to do so.
To fill this gap, and to build on the brand name recognition of the Pentium, Intel recently debuted the Pentium II. (Marketing genius in that name: I voted for the more logical Hexium, but no one listens to me.) This beast is the one to watch coming from Intel. It is also the ace-up-the-sleeve in Intel's bid to drive rivals Cyrix and AMD off the nearest cliff.
Pentium II has been debuted at 233Mhz with MMX, but runs like a classic Pentium 200Mhz. So why buy this nutty monstrosity? Because Intel has demonstrated that this pocket rocket is capable of being accelerated to 450Mhz. And pocket rocket it is: Pentium II is a completely new physical design that permits a radical reduction in physical size. Also, it's pins are gone, replaced by leads run across one edge of the chip which plug into a socket called a SLOT1, much the same way as a PC's internal card plugs into a slot. This complete redesign and change in manufacturing techniques means that Pentium II uses less power, produces less heat and is capable of being clocked to speeds that would melt a Pentium Pro and outright vaporize a classic Pentium.
This slot design also crowds out competitors like Cyrix and AMD, both of which presently manufacture CPUs to be used on the classic Pentium's Socket7 pinout. Intel is driving this change for good technological reasons, but the practice of crowding out competition that couldn't possibly do them any serious harm demonstrates why Microsoft and Intel work so well together.
AMD is fighting back, however. After a disappointing showing with the K5 processor last year (which turned out to be just a Pentium clone) AMD has come back into the ring swinging with its K6 processor, and even the industry insiders who had written AMD off as "bankrupt any day now" are taking another look.
K6 is based upon much of the manufacturing and architecture AMD inherited when it merged with NextGen who had been working far ahead of the curve on the Nx686. Now AMD is finally competing with Intel instead of just being an "also ran." The K6 has been benchmarked at the same performance speeds as the Pentium Pro, but is cheaper, smaller, runs MMX, runs cooler, and plays internal tricks with data (like out of order execution and branch prediction) that aren't the norm for x86 based systems. This makes K6 perfect for laptop solutions where size and heat control are major factors in product choices.
Finally, Cyrix is chasing the pack with its MediaGX machines. Forgoing the beaten path, Cyrix has long since struck out on its own in chip design and designed processors from the ground up that could rival Intel. MediaGX does a good job of pursuing the MMX Pentium but has veered off onto what I see as a dead end street. In Cyrix's pursuit of making systems of increasingly lower cost, they have stepped off the deep end into that purgatory of system development, the proprietary design. MediaGX offers no speed performance over Pentium and is available only with a "companion chip" to manage the MMX systems on a proprietary motherboard design. Not surprisingly, Compaq (the grand daddy of proprietary designs) is putting these systems in to their Presario line of computers. There is nothing wrong with Cyrix's tack, as long as the short term solution is all that you are concerned with. When a faster Pentium II comes out, you'll just remove the old CPU from the SLOT1 socket and plug in the new CPU; As easy as changing a lightbulb. Even with the AMD design, Socket7 has been around for several years and will continue for a time yet. But the Cyrix MediaGX is stuck: impossible to upgrade with anything but a Cyrix.
Naturally, the question of "Who will win?" arises. You don't have to be a rocket scientist to see that Intel keeps stretching for the future while AMD and Cyrix long to fix things in the here and now. But the answer isn't quite as obvious as you may think: Moore's Law states that technology increases exponentially in capacity and speed every 18 months. Intel is looking to hold motherboard design changes at bay for a few years with the implementation of SLOT1. (The idea being that they are Intel, and immune to such oracles. As I recall, that's what Henry Ford thought, too.) Cyrix seems to have decided not to fight Moore's Law and RELY on it instead to drive a constant need for new (even proprietary) hardware. Meanwhile AMD keeps trying to make the better mousetrap by blending both approaches.
Who will win this war? I'm not sure, but the beneficiaries of the carnage are you and I: faster, more reliable systems and radically lower prices bring top end computing within striking distance of most middle income budgets. I intend to sit back and munch my Doritos while this plays out. Heck, its cheaper to watch than baseball and I don't have to stand in line to go to the bathroom.
Peace,
Webwalker