Despite clear technological competence and the ability to generate compelling and attractive products, questionable marketing decisions, corporate conduct, and platform instability continue to hinder AMDís ability to succeed against chip giant Intel.
AMD currently markets its high-end Athlon 64 FX and Opteron processors on the Socket-940 platform. Socket-754 and Athlon 64 fill the upper mainstream. Mid-range and economy systems are addressed by the aging Athlon XP and Socket-A platform. Socket-939 is scheduled to arrive shortly and coexist somewhere between Socket-754 and Socket-940 for a period, until, at some point, AMD decides Socket-939 should go mainsteam, thus bumping Socket-754 down into the value ranks. And, oh, now we have murmurs of a different socket for DDR2 Hammer systems. Confused? Thatís the problem.
Contrast that with Intel. Intel currently markets Socket-604 for server-class Xeon processors, and Socket-478 for desktops. Thatís it (save for an insignificant number of remaining Socket-370 sales at the very bottom). Shortly, weíll begin the migration to Socket-775, starting at the high-end and eventually taking over from Socket-478 completely over the span of a year. Pretty simple in comparison, isnít it? Itís safe, and safety is of great importance to large volume buyers. If you bought a Socket-478 system a year or two ago, you knew you had an available upgrade path and no surprises for at least a few years. The same holds true for those that buy a Socket-775 system within the next few months. In comparison, buy a Socket-754 or Socket-940 system, and who knows what options will exist in a year, let alone two or three.
This is not to say Intel is perfect. There have been a great many bumps along the road, such as 90nm processors that donít work in older boards, and the dead-end Socket-423 platform. But in general, Intel is much more careful about its platform establishment. More importantly, even when something doesnít quite turn out as planned (such as Socket-423), Intel lets the channel know about it well in advance.
Which brings us to our second point of contention: The other critical ingredient missing from AMDís recipe is its ability or willingness to keep the channel informed. The industry has known full-well of Intelís planned migration to Socket-775 for at least a year, because Intel told the industry what to expect. AMD, in contrast, rarely informs the channel of decisions until theyíre imminent. For example, nobody really knew what was going to happen with the K8 architecture until very close to the launch. Similarly, now we have Socket-939 and murmurs of Socket-900, but no one is really too sure where theyíll both fit in, and what upgrade path will remain for older platforms.
Such actions give industry professionals the impression that AMD is flying by the seat of its pants, whether or not it actually is. If Iím an IT manager for a large company, I donít like flying by the seat of my pants. I like knowing what to expect within the next six, twelve, or more months, because my job may depend on it. Five different sockets within a year are bad enough, even if itís known well in advance; five different sockets with no one telling the channel what to expect amounts to little more than a crap shoot.
Back in the days of the K5 and K6, Intel suppressed AMD based on technological merit. Since the introduction of the K7, though, that can no longer be said. The formal technological lead has see-sawed back and forth at times from generation to generation, but the preposition that either firm has held any real technological lead is a difficult one to argue. No, AMDís problem isnít technology, and its greatest foe may no longer be Intel, but rather its own disorganization and strange insistence on corporate secrecy about future products and platforms.