MacqTech Express Semi: Intel‟s 3-D transistor ready at 22nm :: Macquarie Research

Please Share:: India Equity Research Reports, IPO and Stock News
Visit http://indiaer.blogspot.com/ for complete details �� ��


MacqTech Express
Semi: Intel‟s 3-D transistor ready at
22nm
Event
 Intel today showcased its revolutionary technology tri-gate transistor
(generally know as FinFET/3-D) with its next-generation microprocessor "Ivy
Bridge" at the 22nm node, which should start shipping in 4Q11/1Q12.
According to Intel, this technology provides up to 37% performance boost at
the same power consumption or consuming 50% less power when operating
at the same performance compared to its current 32nm technology.

Impact
 The key difference between 3-D and planar transistor lies within the
number of gate terminals (3-D has 3 and surround the channel between the
drain and source terminal, while planar has 1) that are used to control the
electric current flow in the channel between the source and drain during on
and off states, thereby improving the short channel effect with greater drive
current when the transistor is on (meaning greater performance) and lower
current leakage when the transistor is off (meaning more power efficiency). In
short, this is a major breakthrough technology, in addition to the HKMGtransistor
(which was first adopted by Intel in „07), to drive continue scalability
of the transistor beyond 22nm, effectively sustaining the Moore's Law.
 In contrast to HKMG, tri-gate technology completely changes the
structure/formation of a transistor. HKMG does not change the transistor
structure but simply uses different materials (ie replace polysilicon with metal
for the gate electrode and replace silicon dioxide with hafnium dioxid for the
gate dielectric) to make the gate terminal of the planar transistor. These two
technologies are supplementary and Intel is adopting both at 22nm.
 Foundries are behind. TSMC has similar finFET transistor in the works (So
has IBM/GlobalFoundries/Samsung). TSMC unveiled its finFET development
last year with its 22/20nm but indicated that it is unlikely to be used for
commercial production until it reaches the 15/14nm node on 450mm wafer in
2015-16, suggesting that Intel may be 3-4 years ahead of the competition.
 Implications: Up to these days, Intel has been almost non-existent in the
mobile space due to the power efficiency shortcomings of its x86 architecture
vs the ARM-platform. While Intel's next-generation Atom CPU “Medfield”,
based on its 32nm HKMG technology and set for 2H11 launch, may gain
some traction, we believe Intel has a greater chance of penetrating into the
mobile space when it attempts to overcome the power efficiency gap with its
22nm tri-gate process technology next year and gain share from ARM.
 Will Apple move to Intel? We believe this is highly unlikely, but not
impossible, as it would too create a conflict of interest with Intel‟s mobile biz.
Unless Apple decides to switch to x86, realizing that Intel has a superior
mobile product roadmap ahead, combined with its advanced process
technology (as it was the case when Apple switched from IBM RISC to Intel‟s
x86), TSMC should remain the most likely alternative foundry for Apple‟s AP
biz, which would potentially benefit the backend peers ASE and SPIL as well.