Replies to post #62255 on Intel Corp (INTC)

[tecate]

Perhaps. But I've not seen anyone really proven wrong here. I cannot comment or give technical insight per a request but I do think you are too negative on Intel :) I wish I could say more but I can't. I am not an authority on Intel's fabrication plants although I have a very good friend who works for IBM in Fishkill so I can run things by her :)

Add: I've not heard of ONE person stating yields are really bad, not one, I would have thought that. Maybe chipguy or elmer will come by with better insight, as they are more educated than I on this.

http://www.xbitlabs.com/news/cpu/display/20080415233144_Intel_Indicates_Speedy_45nm_Production_Ramp_Shipments_Crossover_in_Q3.html

[mas]

Why doesn't anybody mention the fact that 45nm yields are bad , really bad ?

To be fair Intel raised the possibility if that's any good ?

http://biz.yahoo.com/bw/080415/20080415006500.html?.v=1

Intel is in the process of transitioning to its next generation of products on 45nm process technology, and there could be execution issues associated with these changes, including product defects and errata along with lower than anticipated manufacturing yields.

[Wouter Tinus]

Yields must be horrible if 1 1/2 FABs ( D1D + FAB 32 ) managed to produce only 10m 45nm CPUs in 6 months.

What makes you believe that both factories are running at their maximum capacity (for 45nm) right now? They don't just flip a switch and go from 0 to 10000 wafers/week, that can take up to a year. Last time I saw one of those defect density graphs 45nm was approaching 65nm level (i.e. lowest ever!) fast.

[wbmw]

Re: Why doesn't anybody mention the fact that 45nm yields are bad , really bad ?

Because it isn't true.

Intel has been ramping servers and mobile at the expense of desktop, because servers and mobile are higher margin businesses. 45nm products were only just released at the end of Q4, and in one quarter, Intel already crossed over on server shipments, and their output has grown to 0.5M processors per week.

Furthermore, they still expect to cross over capacity in Q3, which is at or ahead of their usual schedules.

Re: Intel's double patterning requires the wafer to be process twice , the second time being mechanically repositioned.That is a recipe for disaster.

Yet at the same time, they've reduced wafer throughput from 90 days to 45 days.

Re: Yields must be horrible if 1 1/2 FABs ( D1D + FAB 32 ) managed to produce only 10m 45nm CPUs in 6 months.

That's what happens at the beginning of a ramp. Even if you could do a hard toggle of wafer starts (which isn't possible due to needing to qualify all the new machines at each new node), you still wouldn't want to toggle if you can help it. The defects on a brand new process are indeed going to be higher, and taking 6-9 months to optimize the process increases margins. This is congruent with Intel's strategy to ramp server and mobile first, since these segments have higher ASPs to offset the process ramp costs.

AMD has the same issues, but they will toggle hard anyway, just to close their competitive gaps. Their hard 65nm toggle affected margins in H1 2007 pretty badly, affecting their stock price. Intel has the luxury of a competitive 65nm product line, so they can ramp at a pace that will maximize margins.

Re: To achieve crossover with 65nm in Q3 , they need to ramp 45nm to about 2m parts per week .How exactly will they manage that considering that FAB 28 is supposed to come online June 29th ( at least a Q delay ) ?

They don't need another fab to hit this target. They can get 2M parts out of one fab, if they wanted, and depending on the mix of products. Yield is not the limiter, and if it were, you'd see a pretty big impact on margins, or a big slowdown in other segments. As it is, the shortages are in desktop only, or in segments where demand has significantly outstripped supply (in the case of Atom, for example). But then, high demand is a nice problem for Intel to have right now.