32nm is now fine for AMD
Six plus months ago AMD was working hard to launch Llano, and around September 2010 they finally admitted that they have to delay Llano but at least they managed to speed up the introduction of the 40nm Brazos platform.
The reason for delays, as you can imagine, were very poor yields. Most of chips didn’t work and AMD simply had to delay the launch. The reason was quite simple, the lack of Tick - Tock strategy at AMD. Llano is AMD's first 32nm chip with a new architecture, and if you look at Intel, you simply never see them making a brand new architecture on a new process. This is why Intel uses the so-called Tick – Tock approach where a new process is always used for a shrunk core rather than a new architecture, eg Sandy Bridge at 32nm, and the second chip on eg 22nm is Ivy Bridge which is more or less shrink of the Sandy bridge architecture. This is a lot safer than simply launching an entirely new 22nm chip with a new architecture.
The reason is quite simple, making a new architecture on a new process is a recipe for disaster as you double the number of factors that can go wrong. AMD has experienced this with Llano. This strategy didn’t work for Swift, the first Fusion that got canceled, and it resulted with a delay of second Fusion, Llano for at least three quarters. The good news is that Llano now has good yields. We are looking at good availability at launch, but again you won't see too many design wins at first, but they will get more of them with time.
Llano will be a nice CPU with a very good graphics performance, so good that it will be able to put a lot of pressure on the mainstream market in general.