Arthur Cole spoke with Charles King, principal analyst, Pund-IT, Inc.

Cole: The Shanghai processors came out earlier than expected and seem to be in good working order. Do you think the company learned anything from the Barcelona problems?
King: Sure: Avid hunters should avoid shooting themselves whenever possible. More seriously, the processor market and industry are moving so quickly that the punishment for serious mistakes like Barcelona's TLB bugs is swift, severe and long lasting. To its credit, AMD hardened testing procedures for Shanghai and brought server vendors into the process early. That allowed the company and its OEMs to ship solid products more quickly than anticipated. All in all, I think AMD's performance on Shanghai should put the problems with Barcelona to rest and leave the company in a stronger position.

Cole: The 45nm process is impressive, but Intel is already there with the Penryns and has since moved ahead with the Nehalems. Does it hurt AMD to be viewed as constantly playing catch-up?
King: Playing catch-up, at least in the commercial sense, seems to be a constant for AMD, so I'm not sure how badly it actually hurts the company. A bigger issue may be the two to three years of glory the company enjoyed after the introduction of Opteron. AMD basically cleaned Intel's clock in the server market during that period, and helped push Craig Barrett into retirement. Dramatic stuff, but also problematic in that they didn't seem to have a second act in mind after Intel inevitably caught up. But people remember Opteron's blaze of glory and wonder why AMD can't pull something similar out of its hat.

Cole: The Socket F design of the Shanghai allows it to easily replace the Barcelonas. Do you think it's wise to start cannibalizing them already? It seems that after the production delays and TLB bug fixes, they never had a chance.
King: Microprocessor technology is evolving so rapidly that it's usually safer and cheaper to leave the injured behind than it is to nurse them back to health. The Socket F design also allows customers affected by Barcelona's shortcomings, and who may be shy of re-engaging with AMD, to easily upgrade to next-generation chips. Overall, I'd say that it's more important to appease those clients than it is to resurrect hopes for Barcelona.

Cole: The company touts the virtualization prowess of the Shanghai as well, pointing to things like a faster world-switch time and improved indexing. Virtualization is hot right now, so this should count in AMD's favor, right?
King: It should, indeed. In fact, major server OEMs are delivering or developing virtualization-specific server platforms that should be able to take full advantage of Shanghai's capabilities, which look pretty impressive. The only problem is that this is an area Intel is targeting, as well. What it will come down to for many customers is which platform delivers the best price/performance, an issue that won't be entirely clear until the server version of Nehalem becomes available. Until then, Shanghai has a clear field.

Cole: In the end, AMD seems to be positioning itself as the lower-cost, less-complex competitor to Intel. Does that seem like a good position to take, considering the economic pressures that are starting to bear down on the enterprise?
King: That positioning could work to AMD's advantage, especially if the recession is as long and deep as many expect. Some companies will always need to upgrade systems hardware, and bargain hunting will be the order of the day, or year. But an economic downturn could also lead to increasingly aggressive component pricing, a scenario where larger companies typically have more leeway to cut prices than smaller vendors. If it comes to pass, that issue plays to Intel's strengths and could be extremely painful for AMD.