Apple is going to Samsung, with whom they are having patent issues, for manufacturing of their product instead of TSMC because TSMC does not have a "stabilized" process and will only have low volume for them due to everyone else bugging TSMC for manufacturing.
Apple ships ~25m iphones / quarter ... that's a lot of demand.
I understand TSMC may be having issues with meeting that demand, but it may also be that the supply is limited and TSMC makes more money off of AMD/Nvida etc rather than a low-balling Apple which would suck up capacity.
The problem is that you're on the belief that the opposing argument is that 28nm is impossible. It's not impossible.
Indeed.
Now it's just a matter of breaking down what is deemed "too expensive" and how much production is enough?
360 chips had very awful yields going by the Takahashi book. The situation surrounding the 360's design process and launch was a helluva rushed schedule, and look where it got them with RROD. Of course, MS was much more desperate back then than they are now. Do you really think they want to repeat that?
I agree.
That's why I'm not expecting them to repeat that mistake by trying to launch q1/2012 when the 28nm process is just getting warmed up.
Q4 launch should be enough time to field a reasonable launch.
As yields improve, costs will go down, and availability at retail will go up.
Do you believe they even see Nintendo as a threat to their associated early-adopter audience, whom you also identify as the hardcore. Launching early doesn't matter to the mass market who jump on board later.
I believe Nintendo is a bit of a threat, but not significant. I do think some early adopters will buy into Wuu just to see what it's about and Nintendo may try and lure them in with compelling content.
Nintendo didn't always embrace this kiddy image. At the outset of the Ultra64 days they were headed in a very different direction with the likes of Killer Instinct and Cruisin USA (with roadkills) both are very much outside of the typical Nintendo image of the time so it wouldn't be unheard of for them to try and steal this market.
Sony is an unknown.
But Sony still has an edge in a few areas mostly centered on branding and image (see overall sales trend ps3 outselling xb360). In order to compete, MS needs every edge it can get.
Launching first v Sony and preferably near Nintendo allows the early-adopter/hardcore gamer to be won, and with it, the trends of the generation.
I've already described the inherent difficulties with process nodes as you go smaller and smaller. The equipment used is much more complex and expensive. These translate to higher manufacturing costs early on so that TSMC can recoup R&D (just how long have they delayed already) for starters. Limited supply also drives up that price further, and you've already shown us the article about TSMC asking for more money.
25% increase.
The prices outlined are still within the realm of what was estimated to be paid for RSX and Cell.
Both companies are trying to push the envelope of GPUs...
Yes, 8x trans is expected for a nextgen offering. In order to meet this, 32/28nm is necessary if the die size budget is to remain roughly the same thus keeping initial BOM investment in silicon roughly the same.
MS isn't in the same boat, not even close. The question for them is why they should jump onto a next gen whilst utilizing a new node (to push a design they were originally planning for 2013-2014) that is invariably going to cause similar horror flashbacks of 360 launch units for 2012?
The reasons to jump in 2012 are for competitive advantage in capturing the highest consumer spend.
Where are you getting this notion of 2013-2014 was the "plan" all along?
RRoD issues were more to do with design & engineering than yield as we saw the same RRoD in later versions of the console.
Neither Xenos nor Waternoose were anywhere near 300mm^2, let alone 200mm^2 and yet they had crazy yield issues.
Can you define "crazy"? 20% 30%?
How quickly were they resolved?
A couple things: Apple seems to be starting to stock up now (at Samsung). Second they've gone to Samsung out of necessity (despite other legal issues to boot) because TSMC quite clearly has been deemed unworthy of the risk. They simply don't believe they can get the capacity they need.
And with as much volume as Apple needs, I'm not surprised at the outcome. It isn't as though Samsung is luring AMD/Nvidia to use their 28nm process and since they are under-bidding TSMC, they obviously have over-capacity.
