First, it looks like we’re not going to see Northwood until the beginning of 2002. Should this be the case, that means two less months Intel will only have the .13 micron advantage on the high end over AMD. If AMD gets on the stick, might be only a few months.
Second, it looks like Intel plans to take its sweet time ramping this chip upward throughout all of 2002. 3Q 2002 only promises 2.53GHz, and 4Q promises “more” than that.
This looks very odd given:
Intel’s not in the habit of sandbagging with its roadmaps. They just don’t pull things out of the blue way ahead of when the roadmap says they’ll come out. Two or three months earlier, yes, but not six or nine.
Finances and production over performance
When in doubt, figure Intel’s being cheap. 🙂
From a performance junkie’s standpoint, it seems insane to make .13 micron Celerons. However, if you have a limited capability of making .13 micron chips, it isn’t.
I suspect Intel will have its hands full taking care of Christmas orders and shifting production over from .18 to .13 micron. Not because Christmas is going to unleash tremendous demand, but the shift over to .13 micron technology is going to take some time.
New manufacturing equipment designed to handle .13 micron technology is scarce, workarounds with older equipment is expensive, and Intel probably doesn’t want a repeat of the Coppermine shortages from a while back.
So they’ll move relatively slowly into .13 micron for the PIV, partly due to manufacturing constraints, partly because AMD is in no position to push hard for a while.
How’s AMD Doing With .13 micron?
The latest word from AMD is “initial production in 4Q, mass production in the 1H ’02.”
If you take “1H ’02” to mean the end of “1H ’02,” which is usually what such forecasts end up meaning, this seems to be an awfully long transition period.
Perhaps AMD will have its own transition problems. After all, they can hardly shut down Dresden to retool like Intel can with its multiple fabs.
I suspect we’ll see a repeat of what happened with the Palomino core. We’ll see mobile and MP chips first at .13, desktop chips later on.
Stretching Out the Production Cycle
I think we tend to forget that retooling fabs cost a lot of money, billions of dollars, and it’s pretty rough expecting to recoup your cost in a year or so, especially with rockbottom CPU prices.
Right now, the financial trend for the CPU companies is that it’s costing them more and more to make their product, and they’re getting less and less for it. The easiest way to square this circle, especially when demand is down, is to lengthen the life cycle.
So Intel and AMD will get what they can out of .18 micron, and they may stay at .13 micron for quite some time to come.
What Does This Mean For Overclockers?
I see one plus and one negative.
The advantage is that it’s likely we’ll be able to overclock the initial chips to levels well above rated speeds, and they won’t become “official” until long after we’ve done it.
The big disadvantage to that (especially when it comes to Intel) is that .13 micron chips will cost an arm and a leg for a long time. You can’t overclock a chip you don’t own.
Historically, CPUs have to get below $250 in price before they attract any real attention (the pain point may be even lower now), and Northwoods may not reach that point until summer (by which time AMD should have Thoroughbreds out of the gate).
I suppose that’s good news for AMD in the high-performance arena; it will make a lot more financial sense for a lot of people to wait for AMD than shift to (or back to) Intel.