It's No Guarantee, But AMD's Socket A Strategy Beats Intel's Constant Changes
The difference in Intel's and AMD's architecture strategies couldn't be plainer. Intel changes processor form factors every chance it gets, introducing socket designs like seasonal fashions with each new or enhanced CPU -- and usually without throwing away the previous format. Even just looking at Intel's desktop and workstation lineup, you've got the Socket 370 Celeron/Pentium III, 423- and 478-pin versions of the Pentium 4, 603-pin Xeon, and the bizarre Slot M of the Itanium. Confused yet?
Meanwhile, AMD has stuck with the same format since the introduction of the second-generation Athlon (Thunderbird) in June 2000. Athlon, Duron, Athlon XP, Athlon MP? Socket A, Socket A, Socket A, Socket A.
This looks pretty good for AMD, at least on paper, and the company is clearly eager to press the advantage -- in recent roadmaps and sound bites, AMD has poked fun at the fractured Intel landscape, while reassuring the big industry players that it's committed to the Socket A architecture. Certainly, having a stable CPU form factor is great for PR. But is it really an important factor in buying decisions?
That depends on the PC buyer's point of view. Most businesses have historically given little thought to future-proofing new systems, putting stability, price, and performance well above upgradability on their shopping lists. Home and small-office buyers are much the same, regularly buying a Dell or Compaq with little inkling of what's actually inside.
But as money gets tighter and more businesses look to get the most use out of their PCs, this buying methodology is changing. Over the last year or so, I've noticed more companies -- firms that in past years merely signed fat checks for a new fleet -- opting to upgrade system memory and CPU speed instead. Home users are getting into the act, too, at least to the point of asking about obsolescence or trying to avoid integrated or orphaned architectures.
And of course this has long been common practice in the enthusiast market, where a wild array of socket designs can be the kiss of death. If there's one thing early adopters hate, it's buying expensive new PCs and then finding themselves on a dead-end street when it comes to future upgrades or improvements.
Same As It Ever Was?
The plot thickens when you bring all the members of the supply chain into the discussion, as one of the more obvious benefits to a unified socket architecture is the freedom it allows. This gives OEMs a greater degree of flexibility with initial system design, or at least saves them from scrambling to clear inventories each time Intel decides to add a few new pins.
Remember the hoopla surrounding the Pentium 4's transition to its current 478-pin design? The supply chain was in disarray for some time, worried about getting caught with Socket 423 doorstops while trying to ensure an adequate supply of 478-pin processors. To make matters worse, sales of the new CPUs were so high and initial supplies so low that orders had to be adjusted and readjusted to fill the unforeseen demand.
However, AMD is not immune to platform incompatibilities and the problems they entail. The presence of a Socket A interface is often mistaken for complete compatibility across the AMD processor line, but this could not be more of a fallacy. The form factor is not the only variable in the compatibility equation.
For instance, it was at best an error in judgment to allow VIA to introduce chipsets that supported only 200MHz front-side-bus Durons and Athlons, thereby confusing Socket A vendors and buyers even more. I have a contact at a system vendor where the "stable Socket A" mantra caused another headache -- switching to a new supplier for Athlon XP and Duron SDRAM motherboards, since the model and revision in stock proved simply not up to spec with newer core designs.
Although these problems haven't turned many totally off the idea of AMD platforms, it's inherently dangerous to promote Socket A as a stable format without addressing the wide array of potential CPU support problems dependent on chipsets and motherboard revisions.
Pin the Tail on the Pinout
If you do your research and shop smart, the potential for future incompatibilities will certainly be reduced, but it's arguable that AMD is in some ways adding a new level of mystery to CPU support, whereas Intel just lays out a new form factor to be done with it. Intel may change sockets frequently, but if a Pentium 4 fits, then it works. AMD lets you plug in any Socket A processor, but there's the off chance that it won't work.
We can debate which strategy is actually the best, but there's no denying that AMD holds the upper hand in terms of public perception. Whether or not it was technically true for every possible configuration, AMD has successfully promoted the newest Athlon XP and Duron processors as true to the stable Socket A format. That's a strong attraction for business and home buyers alike, conveying the message that AMD got Socket A right the first time and has a consistent plan for the future. If there's one thing that will get you kudos from IT managers, it's flaunting a strong plan for future growth.
By contrast, each time Intel announces a CPU that's totally incompatible with previous parts, I have to shake my head. The shift from 423 to 478 pins left the Pentium 4 with a bit of a black eye, or at least battling the perception that Intel erred with the original Socket 423 design and had to improve it for higher-frequency chips.
Whether the redesign was planned all along or solved an unforeseen problem, Intel's overall strategy seems influenced by profit-driven motivations. Planned obsolescence in a flat PC market is one business strategy Intel has been pursuing for some time now, both as a boost for component sales and stock prices and a method for keeping competitors in line. No doubt Intel believes it's a powerful enough company to call the shots, but in the months ahead we'll see if selling more chipsets is worth handing the platform and PR advantages to AMD.
Images
IE 7
Firefox 2.0
