I'm starting to get worried about Intel (Nasdaq: INTC). Advanced Micro Devices (NYSE: AMD) is really is eating its lunch.
The whole ia64 vs x86-64 thing got me started looking into it, but the problems run pretty darn deep. Racy logic has crippled Intel's ability to move its designs to new manufacturing processes (like copper), and each manufacturing change calls for a processor bug hunt and often a redesign. Their code libraries literally aren't portable from fab to fab since they optimized them TOO much to the specific conditions they were being manufactured under at the time without enough of an eye towards the future.
AMD on the other hand got its clock cleaned for years doing basically the same thing, and two things happened. One, their engineers got really good at getting decent yields out of difficult-to-manufacture designs. Two, back when Compaq (NYSE: CPQ) bought Digital Equipment Corp., AMD hired away DEC's Alpha processor design team and set them to designing a new x86 chip with an eye towards not just performance but also portability between manufacturing processes.
Put the two together (with the Athlon release in August of last year) and AMD leapfrogged Intel into the high end with chips that yield over 90% just as Intel was having yield trouble. Athlon outclocks Pentium III on the same manufacturing process, AND is more easily ported to new manufacturing processes. AMD started shipping gigahertz Athlons in volume several months before Intel could, and is already at 1.1 GHz. (Intel sent out samples 1.13 GHz processors to prove they could match AMD, but ended up recalling them because none of them worked reliably. It was a PR disaster.)
And those 50% margins are coming back to haunt Intel too, since now that AMD's volume is up they have about the same cost structure as Intel but are used to being profitable off of much lower margins. AMD sees no problem scaling Athlon up to at least 2 GHz, whereas Intel has to bring out a new design (Pentium 4) to go much higher. But Pentium 4 is also looking like a real dog. It looks good on paper (high clock speed) but the real world performance doesn't look that impressive. And Intel's proposed specification for P4 requires a one-pound heat sink and enormous power supply. It sounds to me like they're trying to substitute brute force for new technology.
Intel has ALWAYS been the better manufacturer, not the better designer. AMD's out-yielding of Intel by a large margin is a BIG red flag, and a complete role reversal.
Duron, AMD's low-end chip, provides better price/performance than Celeron at the low end (in large part because Intel keeps crippling Celerons so it won't compete with Intel's own high end), and at the high end Intel can't produce gigahertz-plus chips in volume. Meanwhile, AMD is not just running its fabs at capacity with 90%-plus yields but paying down its debt and leaving enough cash left over to build more fabs. AMD beat Intel on the transition to 0.18 microns AND to copper, and x86-64 really looks like a better choice than iTanium.
Intel is positioning both iTanium (due to the incompatible software) and P4 (due to the heat and power requirements) as high-end niche products, for mondo servers only. These products are NOT expected to graduate to desktops. AMD's Athlons (and newer variants like Thunderbird, Duron, and the 64-bit Hammer) graduate all the way down the scale. So in a strange way, Intel's old game of "use the high end to pay for R&D and then cash in on volume in the low end six months later" is now only available to AMD. Intel has segmented its market and can't do that anymore.
Related Links:
