couple days have been dominated by Intel Corp.'s (INTC)
key competitors 
. Today at the hardware convention
Computex 2011 in Taiwan, Intel fired back reaffirming its commitment to the
I. Die Shrinks Every Year
Intel's Atom is behind. It's behind Advanced Micro Devices, Inc.'s (AMD)
similarly priced Fusion processor in performance. It's
behind ARM Holdings plc's (ARMH)
designs in power consumption.
Despite that, it's ahead in the one thing that really matters
-- market share.
Intel is hoping to maintain that lead by getting more aggressive with Atom.
On Monday, Intel executives reaffirmed their commitment to accelerating
the roadmap of Atom.
Atom's codenamed Cedar Trail (part of the Saltwell family)
-- a 32 nm Atom die shrink -- is supposed to launch in Q4 2011, three and a
quarter years after the launch of the original 45 nm Atom in Q3 2008. By
contrast Intel plans to release a 22 nm Atom, dubbed Silvermont only
two years later in 2013, and a 14 nm Atom dubbed Airmont in
That's the same plan that Intel revealed at its annual investors' meeting
two weeks ago.
Cutting its die shrink time from 3 years to two years, then to a single year
seems no easy task. So how is Intel accomplishing this feat?
Well it turns out it already has these die shrinks, on its
high-end personal computer lineup. It shrunk to 32 nm with Westmere,
the die-shrink of Nehalem, which launched in Q1 2010. So the
die shrink on the Atom side is almost two years behind.
Similarly, Intel plans to release its 22 nm Ivy Bridge CPU in
2012 (the die shrink of Sandy Bridge), followed by the 14 nm Broadwell in
2014 (the die shrink of Ivy Bridge successor Haswell).
So, in short, Intel should have the technology on hand to shrink atom to
the promised marks.
That's great news for Intel as it would put it at least a year or two ahead of
the competition such AMD spinoff Global Foundries, South
Korea's Samsung Electronics (SEO:005930), and Taiwan Semiconductor
Manufacturing Company, Limited (TSM).
Paired with Intel's recent announcement of tri-gate transistors at the 22 nm node (a 3D
design that "wraps" the channel in three dimensions to reduce
leakage), Intel may have a winner on its hands.
However, the plan still carries significant risks.
The biggest risk is price. Intel must keep prices low, even if its new
Atom chips begin to close the gap with its competitors in power consumption
and/or performance. That's no easy order, as Intel's aggressive
acceleration of the Atom roadmap mandates that it expand its manufacturing
capabilities to support full Atom production at a smaller node. All that
costs money, and the questions are where that money will come from and whether
that expense will impact the critical price per chip.
A second risk is in a potential letdown post-2014. Assuming Intel can hit
14 nm by 2014, it likely won't hit the next smallest node until 2016 (following
its progress on the desktop side). And from there it's anyone's guess --
the physical limits of optical die shrinks will likely stretch the time it
takes to implement them.
In short, the next several years could be very good for Atom -- or very bad.
II. Medfield, Tablets, and Ultrabooks, Oh My!
Intel also talked a bit about its upcoming smart phone processor "Medfield".
Mired in the departure of a key executive in charge of
its mobile unit and delays, Intel hopes to get Medfield out the door
sometime next year.
Little is known about the chip.
Intel promises a fanless Atom, which can deliver enough processing power for a
tablet or smart phone. The only real detail it shared is that it's
targeting "sub-9mm designs". That figure is in line with
current state-of-the-art Android smart phones, which are thinner than 8.5 mm,
or the somewhat outdated iPhone 4 from Apple, Inc. (AAPL),
which is 9.3 mm.
Intel was also plugging its tablet efforts. The company in April launched Oak Trail, a
high-end tablet platform, which is primarily targeted at Windows tablets.
Oak Trail is essentially the same as the Moorestown platform
(complete with Z6xx branded Lincroft), with a few minor additions,
such as the inclusion of a PCI link.
Sales of Atom-powered tablets so far haven't been stellar, but it's still
early and Intel has great expectations.
Rounding off Intel's ultra-mobility push is the so-called
"ultrabook". Intel defines an ultrabook as a small notebook
with a touch screen for tablet-like features. Intel says it plans to
transfer 40 percent of consumer laptop production to ultrabooks by 2012.
The ultrabook will come at a sub-$1,000 price point. Intel gave a teaser
of one design -- the ASUS UX21 ultrabook from ASUSTEK Computer Inc. (TPE:2357).
ASUS Chairman Jonney Shih plugged the format, stating, "At ASUS, we
are very much aligned with Intel’s vision of Ultrabook. Our customers are
demanding an uncompromised computing experience in a lightweight, highly
portable design that responds to their needs quickly. Transforming the PC into
an ultra thin, ultra responsive device will change the way people interact with
III. USB 3.0
Last, but not least, Intel dropped mention of USB 3.0. In his
keynote Executive Vice President Sean Maloney called USB 3.0 a
"complementary" technology to Intel's "Thunderbolt" -- the preliminary implementation of its "Light
Peak" connectivity technology.
Intel has long turned its back on USB 3.0, choosing to push Light Peak instead.
Thunderbolt -- the preliminary implementation of Light Peak --
disappointed, however, offering copper-based connections, rather than the
optical connections that Intel (and the tech's title) suggested.
USB 3.0 support is currently provided by third parties who make controller
chips for compatible motherboards. Thus far Intel continues to give the
tech a cold shoulder production-wise, even while rival AMD announced the
availability of new USB 3.0-ready chipsets in April.
But Intel is rumored to finally be preparing to get onboard, shipping
chips with USB 3.0 in 2012. The fact that Intel briefly
plugged the technology in its keynote lends some credence to this theory.