July 99 = .25 micron 500MHz K7 debuts in very limited quantities and marketed mainly for high-end servers. Leveled against the Katmai/Xeon server competition, the .25 K7 will be very expensive. Speculative prices fall somewhere around $800 and higher depending upon L2 cache options.

Oct. 99 = K7 .18 die shrink occurs, allowing much higher margin rates for AMD, and lower prices for consumers. This allows AMD's desktop marketing div to finally call on the K7 for duty (the K6-3 will be the lone Katmai placed competitor from AMD until the die-shrink of the K7 according to their own literature) Full .18 production volume should be achieved by late Novemeber.

There are no plans in 1999 for a "3DNow! 2" or large redesign of the original 3DNow! instruction set. AMD feels confident that the original 3DNow! instructions can compete in performance with even the monstrous Intel KNI (MMX2) instructions of Katmai.

My personal two cents on the K7's future look as follows:

First, the problems AMD encountered in the move from .35mu to .25mu this year caused massive delays and limited production to plague both their mainstream K6-2 CPUs as well as their timeline. This is the only reason that the K7 will be forced to debut in a .25mu version instead of the profitable .18mu version that AMD had hoped would be possible. The move to .18 is difficult enough for Intel, much less a company like AMD that has less than half the total chip fabrication facilities and R&D to draw from. Nothing I've seen over the last ten years would lead me to believe that AMD or any other semiconductor manufacturer could suddenly find a way to ease the pains and low yields that always hinder die-shrinks. It doesn't matter if its Intel, AMD, Digital, MIPS, or anyone else, die-shrinks are a bitch.

AMD will live up to the dates we've posted above, but for the "pie in the sky'ers" that believe AMD's press releases like they're a religious prophecy we only ask that you let history be your guide to rationality.

The AMD K7 absolutely kicks Intel's ass all over the place until roughly mid-1999. After that point Intel's 600+MHz .18mu Katmais will hit the scene and the picture looks much grimmer for AMD. The bottom line is that if AMD (and their core logic providers ALi) can ship a 500MHz mainboard/K7 combo in high volume quantities by July 1st, then they'll be able to go toe to toe with the high-end Katmais at that time. If AMD hits bumps and delays along the way which force only limited and expensive production of the K7 by July 1st, then the game might be over before it's started. By November Intel will have long left AMD in their rear view mirror with 800+MHz desktop Katmai entries, as well as their initial "Willamette" class of CPUs.