AMD and ATI head for uncharted territory with 'Spider'

In an effort to make its investment in ATI pay off, AMD is rolling out its first comprehensive PC platform today, aimed carefully at the lower half of the high-dollar enthusiast market.

The Spider platform aims to accomplish for the enthusiast market what Intel's Centrino accomplished for the mobile connectivity market: to specify components that are certain to work together without fuss. In this case, it's a three-way combo consisting of its long-awaited RV670 chipset family, now called its "7-series;" its ATI Radeon 3800 series graphics cards, unveiled just last Friday; and its Phenom series quad-core (not tri-core) processors, announced last May.

But for whom is this platform being marketed? It's midsized OEMs such as Falcon Northwest that are looking for platforms to serve as buildout guarantees, not just for the sake of reliability, but also in hopes that buying platform components in quantity can help reduce their cost. Without an official logo for Spider - indeed, as late as this morning, the company was using a poorly cut-and-pasted photograph of a spider as illustrations for its own literature - only enthusiast system builders may even be made aware of the trademark.

The initial verdicts for the trio of parts is somewhat less than stellar. As Tom's Hardware's Bert Kopelt discovered, a quad-core Phenom 9600 clocked at 2.3 GHz performs on average 4.3% slower, in a battery of 20 tests, than last year's dual-core AMD Athlon 64 X2 6400+ at 3.2 GHz. While the quad-core Phenom did indeed outperform the dual-core in rendering applications, sometimes by slightly higher than 40%, gaming performance actually suffered.

With the 2.2 GHz Phenom 9500 performing over 8% slower, you might expect the top-of-the-line Phenom 9700, expected to be clocked at 2.4 GHz, to match last year's AMD performance. But that processor won't be available until December at the earliest.

Even worse news, though, is Kopelt's results against Intel: The 9600 performed 13.5% slower on average than Intel's Core 2 Quad Q6600, which presently sells for about $300 on the street. AMD's initial price for the 9600 in 1,000-unit quantities is $283, which means its street price will be right in line. That fact might give Intel the early lead in price/performance for desktop platform quad-core.

AMD may very well be aware of this little problem, which could be why its rollout literature this morning is playing up the angle of HD video performance - about the only category where Phenom holds a candle to Intel Core 2 Quad right now. Spotlighting something AMD calls "The Enthusiast Dilemma," it tells a story of enthusiast customers who are looking for some longevity in their product purchases, and who need an alternative to nVidia's "one-dimensional" approach, which merely touts gaming performance while ignoring HD video performance.

PC World's testers this morning complained about their lack of access to the platform parts, and theorized that it had something to do with their general lack of availability to the rest of the market. It then cited benchmark results AMD had published claiming 32% faster performance than an Athlon 64 X2 processor at the same clock speed. That would be the model 4400+, which dates back to 2005.

Today's rollout, while perhaps not as historically significant as its marketing hype would lead you to believe, is still important for AMD's future plans. For it to make the ATI investment work, it needs to find ways for one division's products to give leverage to another's. Up to now, AMD's concept of a platform has been flexible, if not altogether fuzzy: It pre-tests equipment from other manufacturers - which to this day include competitor nVidia - and makes recommendations. That looks nice on paper, but doesn't translate into revenue.

But AMD may have a big problem ahead: Its sole target market for the Spider platform appears to be the very system builders who will have already absorbed those early under-performance figures, and may have already made their decisions based on those findings.