Elmer, now you're just being plain silly. No, I don't want to include AMD's flash output. No, I don't want to include AMD's chipset output. No, I don't want to include Intel's StrongArm output.
Like? What's your problem here? This is not a matter of what I like or dislike. *sigh* Why do you assume everyone is h*ll-bent on making AMD look better than it is? I'm certainly not.
OK, I guess I am used to people trying to distort things.
I believe Intel says they have 5 fabs up and running their .13u process now but I think there are only 2 running .13u exclusively, F20 & F22. The trouble is that there are only about 3 fabs dedicated exclusively to desktop processors. F18 & F20 & F22. The other fabs have a mixture of .13u & .18u and some also run .25u and higher. They are making Tualatin desktop processors, huge die Xeons, Northwoods, 10s of millions of chipset components, StrongArm products, embedded controllers, telecommunications devices, network controllers and a variety of other communications devices plus Flash. They are constantly running new steppings of a multitude of products. Some extremely high volume, some low. Some products are on autopilot and require little nursing.
AMD on the other hand is basically a one trick pony. Athlons. The ratio of experimental runs to production can't be as high as Intel's. How many mainstream products can they be running on their .18u Cu Process in Dresden? 1 main product and a few experiments, almost certainly restricted to metal layers at this point. They aren't running their chipsets on their .18u Cu process and AMD is much more dependent on chipset support from outside than Intel is.
What am I getting at? The point is that it is much easier to estimate yields when there isn't much product mix, as in the case of AMD. Intel's product mix is far to diverse to try and estimate yields and what would the number mean anyway?. I'm sure they have products that yield only 20-30 DPW and others that get 6-700+. 50% could be great or it could be terrible depending on die size. Additionally they must have an order of magnitude more experimental runs going with so many different products. Plus Intel is making Flash in some of the same factories. AMD isn't making Flash in Dresden.
So trying to divine Intel's yields is hopeless while AMD's are possible to estimate fairly well.
EP |
