Typically computer parts follow a predictable performance curve. Just-introduced or highest-end products cost significantly more than parts that are even one step removed from them in performance; in order to gain 3-5% performance a buyer may pay a 40-50% cost premium. From the second-highest product on down, performance usually scales close to price; a 10% performance gain may cost between 5% and 15% more, but the pattern is definite. At the lowest end of the market the opposite pattern emerges. A Duron 700 probably doesn’t sell for much more than K6-2 500, for example, but the performance difference between the two is tremendous.

The highest-end product a company releases, therefore, is typically seen as a marketing tool and perception driver rather than as an actual revenue generator. While the highest-end products built the hype, it was the lower-end that drove OEM revenue and the midrange that most people bought—or, at least, that was the old theory. Now there’s something else going on. More and more, it seems, companies are eschewing the midrange markets to focus on high-end (and higher-margin) enthusiast parts.