Thanks for the contribution.
Your response would appear to be that planar HKMG on bulk CMOS from Samsung is an ST insurance policy against FDSOI not working at 32/28 and 22/20.
You also appear to assert that Soitec is not (yet) delivering thin enough top layer SOI to enable 32/28, 22/20 FDSOI.
Is ST going to be a fabless company? what about the Crolles and Catania facilities that will be stopped in Q4? Is ST management paving the way to loose their in house Si technology know-how? Europe is becoming more and more the periphery of technology arena.
The SOI technology was developed by IBM, and PD (partially depleted) SOI was a big success up to 45-nm. But since then no product based on FD SOI has been developed yet although enormous resources have been spent by IBM and its fab club members including ST. The primary reason why ST has selected Samsung bulk 32/28-nm is that FD SOI technology at 32/28-nm and below has not demonstrated its manufacturability yet by IBM and its Alliances. This is because for 32/28-nm FD SOI to be fully depleted to suppress short channel effects the transistor channel or SOI body thickness of 8-nm or less and for 22/20-nm 6-nm or less are required. However, such ultrathin 8-nm Si channel layer for 32/28-nm and 6-nm for 22/20-nm are not manufacturable by Soitech, the largest SOI wafer supplier. What Soitech can deliver is 12-nm Si layer and 25nm Box layer for 22/20-nm nodes as announced at 2012 SOI conference. Soitech is also a member of IBM Alliances. The second reason why Samsung is selected is that Samsung published its 20-nm bulk technology at 2011 IEDM. Therefore, ST can extend to 20-nm bulk technology. IBM Alliances also published 20-nm bulk Si in 2012 VLSI, not significantly different from Samsung’s paper. However, when the planer bulk 20-nm will be manufactured is not announced yet by IBM or by Samsung. See my post for more information on this topic in “Ecosystem emerges around mobile chips tech,” EE Times 7/11/2012. SKim
I don't think it means that. They have only recently been very gung ho about FDSOI.
But it does beg the question if nobody else goes for FDSOI will they have to drop it?
Is it in any case only a different route to FinFET-on-SOI, which is where bulk FinFET players will go anyway.
Multi foundry strategy is not an easy way to manage and gets maximum value out of 28nm and beyond technologies. Besides, Samsung is still suffering the role-conflict of IDM and foundry. I guess the reason Samsung still earns A6 order is due to a. cost/tech advantages in memory chips and leading 2.5D packaging technology b. TSMC or other foundries can not guarantee enough 32/28nm capacity.
The situation would be changed in 2013 due to capacity boomed and yield improved in TSMC. That would be the reason Samsung allures ST in foundry deal to fill the gap. I am just wondering that Samsung already got tapeout from Qualcomm and others but why Qualcomm still sticks on tsmc's 28nm supply. Same as nVidia. It is an interesting topic to be followed up. Besides, it also implies Global Foundries would be having some troubles with ST.
Blog Make a Frequency Plan Tom Burke 17 comments When designing a printed circuit board, you should develop a frequency plan, something that can be easily overlooked. A frequency plan should be one of your first steps ...