Professional Documents
Culture Documents
MECA587 Advanced Digital Design For Mechatronics: The Origin of Moore's Law ... and Its Consequences
MECA587 Advanced Digital Design For Mechatronics: The Origin of Moore's Law ... and Its Consequences
MECA587 Advanced Digital Design For Mechatronics: The Origin of Moore's Law ... and Its Consequences
The Babbage
Difference Engine
(1832)
25,000 parts
cost: 17,470
4
First transistor
Bell Labs, 1948
I c I B I Se
Courtesy, D. Foty
10
11
PFET
Moores Law
In 1965, Gordon Moore noted that
the number of transistors on a chip
doubled every 18 to 24 months.
He made a prediction that
semiconductor technology will
double its effectiveness every 18
months
12
L O G2 O F T H E N U M B E R O F
C O M P O N E N T S P E R IN T E G R A T E D F U N C TIO N
13
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
195 9
1960
1961
1962
1963
1964
1965
1966
196 7
1968
196 9
1970
1971
1972
1973
1974
1975
Moores Law
Road-Map
Year 2004 90nm
2007 65 nm
2010 45nm
2013 32 nm
2016 22 nm
http://www.intel.com/technology/silicon/mooreslaw/
15
2000
42 million transistors
2 GHz operation
16
Evolution in Complexity
17
A self-fulfilling prophecy?
Gordon Moore was the co-founder of
Intel, so his predictions are more
likely to be turned into reality than
ours.
The whole company tried to keep-up
with this expectation.
Competitors also had to keep-up with
Intel.
18
Moores law in
Microprocessors
Transistors (MT)
1000
100
10
486
1
386
286
0.1
0.01
P6
Pentium proc
8086
8080
8008
4004
8085
0.001
1970
1980
1990
Year
2000
2010
Transistors
Transistors on Lead Microprocessors double every 2 years
19
Courtesy, Intel
100
10
8080
8008
4004
1
1970
8086
8085
1980
286
386
P6
Pentium
proc
486
1990
Year
2000
2010
Die
Die size
size grows
grows by
by 14%
14% to satisfy Moores Law
20
Courtesy, Intel
Frequency
Frequency (Mhz)
10000
Doubles every
2 years
1000
100
10
8085
1
0.1
1970
8086 286
386
486
P6
Pentium proc
8080
8008
4004
1980
1990
Year
2000
2010
Lead
Lead Microprocessors
Microprocessors frequency
frequency doubles
doubles every
every 22 years
years
21
Courtesy, Intel
Power Dissipation
Power (Watts)
100
P6
Pentium proc
10
8086 286
1
8008
4004
486
386
8085
8080
0.1
1971
1974
1978
1985
1992
2000
Year
Lead
Lead Microprocessors
Microprocessors power continues to increase
22
Courtesy, Intel
18KW
5KW
1.5KW
500W
Power (Watts)
10000
1000
100
Pentium proc
286 486
8086
10
386
8085
8080
8008
1 4004
0.1
1971 1974 1978 1985 1992 2000 2004 2008
Year
Power
Power delivery and dissipation
dissipation will
will be
be prohibitive
prohibitive
23
Courtesy, Intel
Power density
Power Density (W/cm2)
10000
Rocket
Nozzle
1000
Sun
Surface
Nuclear
Reactor
100
8086
10 4004
Hot Plate
P6
8008 8085
Pentium proc
386
286
486
8080
1
1970
1980
1990
Year
2000
2010
Power
Power density
density too
too high to keep junctions at low temp!
temp!
24
Courtesy, Intel
AlCu
SiO2
Tungsten
poly
p-well
n+
p-epi
p+
p+
25
SiO2
n-well
photoresist
removal (ashing)
photoresist coating
stepper exposure
26
Patterning of SiO2
Chemical or plasma
etch
Si-substrate
Hardened resist
SiO
2
Si-substrate
Photoresist
SiO
2
Si-substrate
Si-substrate
(e) After etching
Exposed resist
Si-substrate
(c) Stepper exposure
27
SiO
2
Si-substrate
(f) Final result after removal of resist
Design Metrics
How to evaluate performance of a
digital circuit (gate, block, )?
Cost
Reliability
Scalability
Speed (delay, operating frequency)
Power dissipation
Energy to perform a function
28
Recurrent costs
silicon processing, packaging, testing
proportional to volume
proportional to chip area
29
30
Die Cost
Single die
Wafer
Going up to 12 (30cm)
31
From http://www.amd.com
-per-transistor
1
0.1
0.01
0.001
0.0001
0.00001
0.000001
0.0000001
1982
1985
1988
1991
1994
1997
2000
2003
2006
2009
2012
Courtesy, D. Foty
1/DSM
Macroscopic Issues
Time-to-Market
Millions of Gates
High-Level Abstractions
Reuse & IP: Portability
Predictability
etc.
Microscopic Problems
Ultra-high speed design
Interconnect
Noise, Crosstalk
Reliability, Manufacturability
Power Dissipation
Clock distribution.
Everything Looks a Little Different
34
Design crisis
Process technology provides a 59% per year increase of
complexity (Moores law)
Design efficiency increases by only 25% per year
Log #
transistors
Technology
59% / year
Design
gap
Design
25% / year
Time
35
Why Scaling?
Technology shrinks by 0.7/generation
With every generation can integrate 2x
more functions per chip; chip cost does not
increase significantly
Cost of a function decreases by 2x
But
How to design chips with more and more
functions?
Design engineering population does not double
every two years
Log #
transistors
New
process
New reuse
technology
method
Design
25% / year &
paradigm
shifts
Time
37
38
Intellectual Property
Adapted from (Jan. Rabaey, Digital Integrated Circuits). Copyright 2003 Prentice Hall/Pearson
MODULE
+
GATE
CIRCUIT
DEVICE
G
S
n+
41
D
n+
VDD
M2
Vout
Vin
M1
42
M4
Vout2
M3
42
Gajkskis Y chart
43
Gajkskis Y chart
44
45
Partitioning
Floorplanning
&
Placement
Routing
Compaction
46
Fabrication
48
Full custom
design
This approach is
extremely slow,
expensive
Only used to
design very high
performance
systems
Standard cell
based design
This approach is
reasonably fast,
less expensive
Most ASICs are
currently designed
using this method
Gate array
based design
FPGA based
design
This approach is
fast and less
expensive
Relatively slow
performance
Slow performance
Choice depends on area, speed, power dissipation, design time and cost.
49
Fail
Function
And Timing
verification
Post-Layout
simulation
Pass
Go to fabrication
Pass
ASIC Chips
It is a time consuming all manual process, pre-developed libraries not needed.
50
51
High-level verification
Pass
Logic gate library
Logic synthesis
Fail
Gate-level verification
Pass
Cell layout library
52
Post-Layout verification
Pass
Go to fabrication
Adapted from (Jan. Rabaey, Digital Integrated Circuits). Copyright 2003 Prentice Hall/Pearson
53
A
D
Post-Layout verification
55
This approach is faster and cheaper than using standard-cell, because part of
the fabrication process has been complete.
Gate Array
Sea-of-Gates
Schematic
Capture
netlist
FPGA cell
library
Implementation
Fail
Verification
Pass
Technology mapping
Placement & routing
Timing verification
Download
Standard cell
based design
Gate array
based design
FPGA-based
design
+++
++
+++
++
--
++
++
---
--
+++
All
All
Some
None
Fabrication time
---
--
+++
Time to Market
---
--
++
+++
Risk reduction
---
--
+++
Future design
modification
---
--
+++
Speed
Integration
density
High-volume
device cost
low-volume
device cost
Custom mask
layer
58