Pemalite said:
Double patterning was only supposed to be an interim solution, Intel has used it in the past, for parts of it's 65nm and 45nm chips, TSMC used it for parts of it's 32nm lithography too.
Intel also used it lightly at 22nm and 14nm.
Intel started investigating the use of EUV back in 2003, which could have driven things down to 10-7nm without the need for double or quad patterning, unfortunatly I haven't heard anymore on that front in years, I assume it's been abandoned due to a number of possible reasons.
There are also differen't patterning types, for example spacer-based double-patterning which is oft-used in NAND, Double, Triple, Quad-druple and beyond patterning, dual-tone, self-align, double expose, double etch and more all have different pro's and cons as well as costs.
But as things stand right now, double patterning is still relatively economically feasible, the foundries have done extensive research into it for years and consumer products have been using it.
However, it probably wouldn't be economically feasible to a point. (I.E. Tiny, low profit chip.)
I agree, that Apple will probably have to go back to Samsung, eventually, or take advantage of some spare capacity from Intel, who are looking to expand to producing chips for other companies.
However, TSMC still has a few tricks up it's sleeve and Apple can easily buy some I.P. from IBM to improve it's chips. (Resonant, Clock-mesh anyone? Or just buy IBM outright...)
To be honest, Apple probably has more important things on it's hands, like upgrade the pathetic amount of Ram in it's phones, which affects performance.
However, we probably won't see a shift during the next GPU cycle but the one after, to much effort has probably been done to get designs working with TSMC's fabs, or things could go as status quo, who knows, all speculation at this point. :P
|
