Thank you for this interesting article. You are taking into account the IC design / IC manufacturing / graphic parameters to analyse the evolution of this battle.
However, you are missing one very important point which is more and more critical nowadays which is IC packaging & 3D integration (by the mean of the capability to partition SoCs in terms of IP, manufacturing node and process)... and in this space, TSMC & Samsung are much ahead of Intel in terms of technology advancement. 3D IC integration by the mean of 2.5D interposer substrates (such as Xilinx Virtex 7.0 FPGA 3D-SOC) or Wide IO memory to logic 3D stacking with TSV will show up as soon as 28nm node generation and will provide additional flexibility versus cost versus performance value than 22nm single dies MSOCs. Maybe Intel can compete to 3DIC chips with single die MSOC on 22nm... but what about the assembly & packaging of such chip? What about the flip-chip laminate substrate price associated to this chip (ask Ibiden, Kinsus, Nanya or unimicron)... you will get to know with further detailed analysis that the packaging & 3DIC partitioning parameters are becoming much more critical starting on 28nm and even more on later nodes (22nm and below) to differentiate your MSOC product value on the market.
And in this space, Samsung and TSMC have much more advance than Intel to date... bringing up the competition older IC manufacturing designs (like gate last, 28-45nm) compared to leading edge 22nm and sub-20nm processes. And as TSMC / Samsung are serving the Apple, TI, Marvell, STEricsson, nVidia and Qualcomm of this world, the battle of 3DICs versus single die SOCs is just at its beginning... will be interesting to follow as Intel will push the 2D-SOC trend to its ultimate limits... while the other player have already anticipated the move to next generation 3DIC/3D-SOC partioning. Again, the wireless IC players will take more risk than either to keep ground in this booming mobile / tablet space... but are likely to be much better prepared for the next two decades of evolution which will leverage this unprecedented flexibility in the design and manufacturing of future high performance, low power, low cost SOC chips. Because it is all about finding the next machine that will drive for at least two more decades what the Moore's law has achieved the past 40 years.
by YOLE Developpement
Facebook
Twitter
Instagram