Transcript The Future of Integrated Circuits

Integrated Circuits and the Future of
Semiconductors
MSE-630
The Electronics Industry is
BIG Business!!
•In 2008, global
sales of
semiconductor
devices and
components are
projected to be
$309 billion
•When
incorporated into
end-use
components built
on semiconductor
devices, the
market is projected
to be 1.7 trillion
dollars
MSE-630
The Revenues of U.S.-based
chip companies account for nearly
half of global semiconductor sales
and more than three-quarters of
U.S.-owned chip manufacturing
Capacity is located in the United
States.
The U.S. chip
industry provides
more than $100
mission annually to
support research in
U.S. Universities,
invests $15 billion
in R&D, and
employs 226,000
people
Technology
exports account
for 23% of total
exports. 75% of
the chip
industry revenue
is from export
sales.
MSE-630
Historically, as the speed
of devices increases, their
size decreases
In a 1965 paper, Gordon Moore
stated that the number of
components on the most
complex integrated circuit
chip would double each year
for the next ten years
The transistors
manufactured today are
20 times faster and
occupy less than 1% of
the area of those built
20 years ago
In 1965, there were 50-60 components on the average chip
MSE-630
• 65nm Logic Technology
Intel recently disclosed details of its 65 nm
generation logic technology which includes
numerous features to improve performance and
reduce power. This technology is being
demonstrated on fully functional 70 Mbit SRAM
chips with over 1/2 billion transistors. Once
again proving that Moore’s Law is alive and well,
Intel’s 65 nm technology is on track for delivery
in 2005.
MSE-630
To continue to grow, the
semiconductor industry has to
overcome several technological
challenges:
•Lithography
•Transistor scaling
•Interconnections
•Circuit families
•Computer memory
•Circuit design
MSE-630
Moore pointed out in his original paper that the doubling
0f the number of components on an integrtated circuit was
due to three factors:
•Half the increase is derived from improvement in lithographic
resolution
•A quarter comes from larger chip sizes, made possible by
improved manufacturing techniques and getter lithography
•The remaining 25% is due to innovation, such as more creative
techniques for forming components on a chip
The industry will continue to grow so long as the rate
of increase of components and functions on a chip
exceeds the rate of increase of the cost per chip.
MSE-630
Lithography
Deep ultraviolet reduces dimensions
to 250-nm. Smaller wavelengths
present challenges due to diffraction
of light and distortion, as well as
photoresist materials
Electron-beam lithography
can, with l = 0.01-nm, has
high resolution. Key
problems include:
Historical and future trends of
lithographic resolution capability
-multiplicity of masks
required
Originally mercury lamps were
used to get resolution of ~350 -mask integrity
nm
-cost
MSE-630
Lithography
Proximity X-ray lithography has been used to fabricate ICs to 150-nm.
The primary problem is that lenses and mirrors are not available for
these wavelengths. Blocking masks must be used with features of the
same dimension as that on the wafer. The cost and difficulty of
fabricating thee masks without distortion are key challenges
The biggest risk of any new lithographic
technique is that the benefits derived from
increased component density are outweighed by
the increased cost.
MSE-630
Transistor Scaling and
Design
The key tradeoff in technology
adaptation is speed vs. cost.
Bipolar transistors are faster
than CMOS, but CMOS has higher
circuit density – thus, CMOS
wins out
Comparison of projected vs. actual
device performance
For any reduction a in linear dimensions
the voltage and doping levels can be
adjusted to increase performance by a
and decrease power density by a2
Trend of microprocessor clock
frequency
MSE-630
Transistor Scaling and
Design
Because devices operate at room
temperature, off current limits
designs to threshold voltages of 0.3V
or higher
Future improvements will
require significantly
lower operating
temperatures
Continued speed increases will require improved
software and I/O design
MSE-630
Shrinking components results in gate-oxide tunneling. The limit of a useful
device with an on/0ff current ratio of 1000 due to source-drain tunneling alone
appears to be about 5-nm separation between source and drain. Accounting
for dopant fluctuations, the lower limit is likely 10-nm.
Possible
advances in can
come from:
•Shorter channel
lengths
•Materials with
higher
performance
Plausible evolution in transistor structure toward a more symmetric structure
that results in better control of the fields in the gate region, regulating device
condition. The FETs pictured are: (a) bulk Si, (b) silicon-0n-insulator (SOI), (c)
ground plane, counter electrode (d) verticle double gate and (e) fully
symmetric double gate
MSE-630
Transistor Scaling and
Design
New device
designs will
move toward
threedimensional
arrays of
devices.
These devices
reduce space
while using the
same size
components as
current
devices
MSE-630
Low Temperature Operation
Operating at LN2
temperatures (77K) would
improve performance by a
factor of 2. Problems
include:
•Refrigerator cost
•Reliability
•Need to redesign
technology to optimize
low-temperature operation
Optimum
cost/performance
operation may occur
at -50oC using
thermoelectric
methods
The best candidates
for this are high-end
servers
MSE-630
Wiring and Connections
Traditional interconnects are
Al or Cu – both of which have
resistance and capacitance
Using materials with
low dielectric
constants in insulation
layers allows
continued decreases in
size
Another potential
solution is to use a
hierarchical wiring
scheme, which
combines high-density
wiring at the first few
levels with larger,
lower-resistance and
capacitance wires at
upper levels
MSE-630
Circuit Families
Bipolar
transistors
generate
more heat
than CMOS
Bipolar transistors win
on speed – CMOS wins
on device density
Bipolar vs CMOS performance trends
The large number of components on a chip
lead to a superior system performance and
lower cost per components. With more
devices, more functions can be designed into
a single chip.
MSE-630
No alternative logic technology is evolving within
cost/performance products on the market today to
threaten CMOS dominance. In light of the continuing
CMOS performance evolution, an even steeper evolution
and learning curve would be required to displace CMOS.
No radical shift in circuit type seems to be on the horizon.
Moore’s law will continue to hold for
approximately 10 more years
MSE-630
Memory Cells
DRAM is a cell consisting of a
transistor and capacitor.
For the past 20 years,
DRAM products have
followed a generational
evolution leading to a 4X
increase in bits per chip
every three years.
The current limit in size
is 4X the square of the
lithographic dimension
In 1990, 1Mb of
memory cost
~$175 retail.
Now, a 1Gb
DIMM costs $130
As lithographic improvements slow, so
will growth of Memory
MSE-630
Design
The increase in function
– the ways the devices
on a chip can be
arranged, will be the
driving force for new ICs,
not the sheer number of
devices on a chip
For a 10GHz processor, the clock
cycle time is 100 ps. Since light
travels at 300 mm/ps, in vacuum,
the space reachable by light in one
clock cycle is 30-mm. Assuming a
medium consisting of typical
dielectrics rather than vacuum, the
reachable space is of the order of
15-20-mm, roughly the size of
today’s chips. This places an
upper bound on clock speeds and
planar chip sizes
MSE-630
Cost
Cost reduction is a major tenet of Moore’s law. The primary factor
underlying the decreasing cost per circuit or memory bit is the increase in
density, or circuits per square millimeter. The cost of processing a silicon
wafer must increase much less rapidly than the density in order to achieve
cost reduction. The rate of cost increase in the silicon chip manufacturing
is approximately 15% per year. This si sdue to:
•Stabilization of clean room requirements
•Better equipment productivity and utilization
•Slower increase in the number of process steps
The rate of increase in costs must be matched by a greater rate of increase
of components per chip to continue to thrive
MSE-630
Future Directions
CMOS technology is likely to continue to evolve and dominate
the semiconductor industry for the next 10-15 years
•Optical lithography must be extended to unanticipated levels or
be replaced by non-optical techniques.
•Transistors must be replaced with a radical new structure
using new materials
•DRAM cells must be designed in as-yet-unknown structures to
achieve economically viable increase in memory chip integration
•Wires must be fabricated at tenth-of-a-micron dimensions in
hierarchical structure with low-dielectric constant materials
•Dynamic circuits and SRAM cells must be designed to
provide more function for a given set of transistors
•Cost reductions will continue to be driven by the ability to
integrate more functions on a chip
MSE-630
MSE-630
MSE-630
MSE-630
MSE-630