I don’t think anyone sane would argue that the year 2007 could be described as “Intel stayed a step or two ahead of AMD, which drove AMD into the bargain basement.”
Looking ahead to next year and a bit beyond, I just see history repeating itself, again and again and again.
AMD counters C2D with K10. From everything we’ve seen, K10 is a tweak, a good sizable tweak, but still a tweak, and a tweak handicapped by the limits of AMD’s 65nm SOI process technology.
Next month, Intel will introduce Penryn. From everything we’ve seen, Penryn (outside of SSE4) is a tweak, too. If anything, Penryn is a somewhat less sizable tweak than K10, but enough to keep a sufficient lead to keep AMD down, especially given the rather slow frequencies we’ll see from K10 for at least the next six months.
Presumably, hopefully, AMD will eventually get around to making 3GHz K10, but by the time they’re likely to do that and endanger Penryn’s lead, Penryn will get a clock boost.
OK, AMD’s answer to that is to go to 45nm. At best, if AMD 45nm production begins “in the first half of 2008” (which means in AMDese “towards the end of the first half”), the first 45nm chips will show up at earliest in September/October.
The ramping of 45nm processing will be slow; Digitimes recently reported that 45nm tricores wouldn’t show up until “the first half of 2009”, which sounds like the middle of the year.
In any event, even if you believe AMD can hit its dates, shortly after the first processors come out, Intel will put out the first Nehamems.
What are Nehalems likely to do? Well, integrating the memory controller into the CPU got AMD a 20% boost when they first did it. If Intel gets anything close to that, that will certainly cover any likely improvement from AMD’s 45nm process conversion.
Sometime in 2009, AMD is supposed to come out with Bulldozers and Fusion, but by then, Intel will have its own Fusion-like products, and Intel’s 32nm process technology won’t be far away, and . . . .
Do you see a pattern here?
It’s not that AMD hasn’t gone through periods where they were definitely second fiddle and made up for it by charging less.
The problem is during those times, AMD didn’t owe five billion dollars, or run up big loss after big loss, quarter after quarter.
Nor did Intel have the antitrust suits and actions currently pending as an incentive to kill/maim AMD, then say to the courts/agencies, “Look, they can’t compete even when we leave them alone.”
Back around the year 2000, neither I nor anyone else thought Intel ever really wanted to kill AMD, they just wanted to keep them in their place without hurting their profits too much.
Today, I think Intel really wants to kill AMD because the only way they’ll ever get back to high profits is by doing just that, and if they’re going to get antitrust hell anyway, better that they pay any award to a corpse, or let a corpse compete freely against it.
Unless AMD comes up with superior products, Intel will just keep them in the chokehold until they either say “Uncle” and scale back their ambitions, or just . . . die.