Fudzilla reports that we won’t see 45nm dual-cores until well into 2009.
This is not good. Why?
Quad-cores are not mainstream now, and will not be in 2008 (they’ll account for about 10% of Intel’s product mix). They will not be the norm in 2009, and probably not even in 2010. Certainly for at least the next two years, the vast majority of CPUs sold will be dual cores.
Dual-cores are easier to make, cheaper to make, and handle most tasks just as well as quads. That doesn’t mean you shouldn’t buy a quad, it does mean that most computer buyers won’t buy quads anytime soon.
Intel is well-aware of this, which is why they will manufacture mostly dual-cores, even at 45nm, and 45nm dual-cores are being introduced in the same timeframe as the quads.
So why did AMD put dual-K10s last on its rollout list? Why is it going to put dual-K10.5 last on its 45nm rollout list?
There’s only one rational explanation for both, and it’s AMD’s biggest “secret.”
AMD can’t make these chips in volume, and because they can’t, they’re stunting with low-volume quads to make themselves look more advanced than they really are, and hide the fact that they can’t ramp these new generations to serious levels of production.
Obviously AMD has had more than its share of technical problems with the Barcelona and Phenom quads, but I don’t think technical issues are the core problem here. I think this is mostly a capacity issue.
The fact is AMD can’t introduce a dual-core K10 or K10.5 until AMD is ready to make a lot of them, millions of them. No big OEM is going to accept just getting a handful of them.
I think it’s very safe to say that for at least the next two years, the dude getting the standard Dell or HP or Acer or whatever box is going to get a dual-core CPU.
Over the next two years, Intel will have dual-core Celerons, dual-core C2Ds, dual-core Penryns and last but not least dual-core Nehalems. Intel should have little difficulty beating AMD on the high-end, while the dual-core Celerons will cause AMD some pricing problems on the low end (especially if X2s stick around a while).
What’s AMD going to have? For the next three-four months, all they have are X2s. After that, some dual-core K10s are supposed to start dribbling in, but even if they do, they’ll be beaten by the Penryns and the C2Ds. Even assuming these 45nm chips show up on time, their prime-time competition by then will be dual-core Nehalems.
Not being able to do better than second-rate chips from new generations of CPUs is bad enough, but when the time schedule of the second-rate chips boils down to “too little, too late,” it’s hard to see how AMD can successfully follow a “pile ’em high and sell ’em cheap” strategy.