CMPEN 411VLSI Digital Circuits
Spring 2011
Lecture 01: Introduction
Sp11 CMPEN 411 L01 S1
Lecture 01: Introduction
� Kyusun Choi
� Course Website: http://www.cse.psu.edu/~kyusun/class/cmpen411/11s/index.html
� [Adapted from Rabaey’s Digital Integrated Circuits, Second Edition, ©2003 J. Rabaey, A. Chandrakasan, B. Nikolic]
How Do the Pieces Fit Together?
I/O systemInstr. Set Proc.
Compiler
OperatingSystem
Application
Instruction SetArchitecture
Firmware
Memory
system
Sp11 CMPEN 411 L01 S2
Digital Design
Circuit Design
� Coordination of many levels of abstraction
� Under a rapidly changing set of forces
� Design, measurement, and evaluation
Datapath & Control
Course Contents
� Introduction to digital integrated circuits
� CMOS devices and manufacturing technology. CMOS logic gates and their layout. Propagation delay, noise margins, and power dissipation. Combinational (e.g., arithmetic) and sequential circuit design. Memory circuit design.
� Course goals
Ability to design and implement CMOS digital circuits and
Sp11 CMPEN 411 L01 S3
� Ability to design and implement CMOS digital circuits and optimize them with respect to different constraints: size (cost), speed, power dissipation, and reliability
� Course prerequisites
� EE 310. Electronic Circuit Design
� CMPEN 471. Logic Design of Digital Systems
Background from CMPEN 471 and EE 310
� Basic circuit theory
� resistance, capacitance, inductance
� MOS gate characteristics
� Hardware description language
� VHDL or verilog
Sp11 CMPEN 411 L01 S4
� Use of modern EDA tools
� simulation, synthesis, validation (e.g., Synopsys)
� schematic capture tools (e.g., LogicWorks)
� Logic design
� logical minimization, FSMs, component design
Course Structure
� Design and tool intensive class
� Industrial Standard toolset for layout
- Online documentation and tutorials
� HSPICE for circuit simulation
� unix (Sun/Solaris) operating system environment
� Lectures:
� 2 weeks on the CMOS inverter
Sp11 CMPEN 411 L01 S5
� 2 weeks on the CMOS inverter
� 3 weeks on static and dynamic CMOS gates
� 2 weeks on C, R, and L effects
� 2 week on sequential CMOS circuits
� 2 weeks on design of datapath structures
� 2 weeks on memory design
� 1 week on design for technology scaling, trends
Schedule is on-line syllabus
What is the most important invention for the last 50 years?
Sp11 CMPEN 411 L01 S6
The evolution of IC
� When was the first transistor invented?
A. 1945 B. 1947 C. 1951 D. 1958
� The inventors were in which company?
Sp11 CMPEN 411 L01 S7
� The inventors were in which company?
A. IBM B. Bell Lab C. TI D. Motorola
The evolution of IC
� When was the first transistor invented?
� Modern-day electronics began with the invention in 1947 of the transfer resistor, also known as the bi-polar transistor by Bardeen et.al at Bell Laboratories
Sp11 CMPEN 411 L01 S8
The evolution of IC
� When was the first IC invented?
A. 1956 B. 1958 C. 1959 D. 1961
Sp11 CMPEN 411 L01 S9
� The inventor was with which company?
A. IBM B. Bell Labs C. TI D. Motorola
The evolution of IC
� When was the first IC (integrated circuit) invented?
� In 1958 the integrated circuit was born when Jack Kilby at Texas Instruments successfully interconnected, by hand, several transistors, resistors and capacitors on a single substrate
Sp11 CMPEN 411 L01 S10
Transistor Revolution
� Transistor –Bardeen et.al. (Bell Labs) in 1947
� Bipolar transistor – Schockley in 1949
� First bipolar digital logic gate – Harris in 1956
� First monolithic IC – Jack Kilby in 1958
� First commercial IC logic gates – Fairchild 1960
Sp11 CMPEN 411 L01 S11
� First commercial IC logic gates – Fairchild 1960
MOSFET Technology
� MOSFET transistor - Lilienfeld (Canada) in 1925 and Heil (England) in 1935
� CMOS – 1960’s, but plagued with manufacturing problems (used in watches due to their power limitations)
� PMOS in 1960’s (calculators)
Sp11 CMPEN 411 L01 S12
� PMOS in 1960’s (calculators)
� NMOS in 1970’s (4004, 8080) – for speed
� CMOS in 1980’s – preferred MOSFET technology because of power benefits
� BiCMOS, Gallium-Arsenide, Silicon-Germanium
� SOI, Copper-Low K, strained silicon, High-k gate oxide...
Worldwide Semiconductor RevenueSource: ISSCC 2003 G. Moore “No exponential is forever, but ‘forever’ can be delayed”
Sp11 CMPEN 411 L01 S13
Transistors shipped per year
Sp11 CMPEN 411 L01 S14
How many transistors you can buy with 1$?
Average Transistor Price by Year
Sp11 CMPEN 411 L01 S15
1’’ Wafer in 1964 vs. 300 mm (12 ”) Wafer in 2003
Sp11 CMPEN 411 L01 S16
The IC in 1961 vs. Intel Pentium 4 in 2004
Sp11 CMPEN 411 L01 S17
Moore’s Law
� In 1965, Gordon Moore predicted that the number of transistors that can be integrated on a die would double every 18 months (i.e., grow exponentially with time).
� Amazingly visionary – million transistor/chip barrier was crossed in the 1980’s.
Sp11 CMPEN 411 L01 S18
crossed in the 1980’s.
� 2300 transistors, 108 KHz clock (Intel 4004) - 1971
� 16 Million transistors (Ultra Sparc III)- 1998
� 42 Million, 2 GHz clock (Intel P4) - 2001
� 125 Million, 3.4Ghz (Intel P4 Prescott)- 2004 Feb 02
� 234 Million, IBM Cell processor, 2005
� 1.7 Billion, 1.6Ghz (Intel Itanium-2)-2006, Sept.
Moore’s Law plot (from his original paper)is
tors
106
107
108
integrated
circuit
invented
109
Sp11 CMPEN 411 L01 S19
year
# t
ran
sis
t
100
101
100
102
100
103
104
105
10
memoryCPU
19701960 1980 1990
invented
2000 2010
Moore’s Law in Microprocessors
Pentium® procP6
10
100
1000
Transistors (MT)
2X growth in 1.96 years!
# transistors on lead microprocessors double every 2 years
Sp11 CMPEN 411 L01 S20
400480088080
8085 8086
286386
486Pentium® proc
0.001
0.01
0.1
1
1970 1980 1990 2000 2010
Year
Transistors (MT)
Courtesy, Intel
# of Transistors per Die
Source: ISSCC 2003 G. Moore “No exponential is forever, but ‘forever’ can be delayed”
Sp11 CMPEN 411 L01 S21
Intel 4004 Microprocessor (10000 nm) 1971
2300 transistors13.5 mm2108k Hz
Sp11 CMPEN 411 L01 S22
Intel Pentium 4 –Prescott (2004)
90 nmArea: 112 mm2125 M transistorsL1-Instruction: 16KL1-Data: 16KL2: 1MB
Sp11 CMPEN 411 L01 S23
Two chips you are seeing today
Sp11 CMPEN 411 L01 S24
Microprocessor ASIC (Application Specific IC)
366MHz 40mm2 3.65M 40Mhz 10 mm2 500K
IBM Cell Overview
PPU
SPU
SPU
SPU
SPU
MIC
RRAC
BIC
MIB
Cell Prototype Die (Pham et al, ISSCC 2005)
Sp11 CMPEN 411 L01 S25
� IBM/Toshiba/Sony joint project - 4-5 years, 400 designers, 3/9/2001, $400M, 234 million transistors, 4+ Ghz, 256 Gflops (billions of floating pointer operations per second)
SPU
SPU
SPU
SPU
C
State-of-the Art: Lead Microprocessors
Freq
(HZ)
Transistors Die size
mm2
Power Date
Server IBM Power 4+ 1.7G 180M 267 N/A 2003
Itanium 2 1.5G 410M 374 130W 2003
IBM Power 5 2G 276M 389 N/A 2004/2
PC IBM Power PC970 1.8G 58M 118 42W 2003/6
Pentium 4 3.2G 55M 131 82W 2003/6
AMD Athlon 64 2.2G 105M 192 89W 2003/9
Sp11 CMPEN 411 L01 S26
Pentium 4 180 nm (2001) 1.7 G Hz 42 M transistors 217 mm2 Pentium 4 130 nm (2003) 3.2G Hz 55 M Transistors 131 mm2 Pentium 4 90 nm (2004) 3.4 Hz 125 M Transistors 112 mm2 Pentium on 65nm (2005/2006) 250 Million Pentium on 45nm (2007) 400 to 500 Million
AMD Athlon 64 2.2G 105M 192 89W 2003/9
Pentium 4 (Prescott)
3.4G 125M 112 103W 2004/2
(All use 0.13 um technology except Pentium 4 – Prescott, which uses 90 nm tech)
State-of-the Art: Lead Microprocessors (up to date)
Sp11 CMPEN 411 L01 S27
300mm wafer and Pentium 4 IC. Photos courtesy of Intel.
64,000
256,000
1,000,000
4,000,000
16,000,000
64,000,000
100000
1000000
10000000
100000000Kbit capacity/chip
Evolution in DRAM Chip Capacity
0.35-0.4 µm
0.18-0.25 µm
0.13 µm
0.1 µm
0.07 µm
human memoryhuman DNA
book
4X growth every 3 years!
Sp11 CMPEN 411 L01 S28
64
256
1,000
4,000
16,000
10
100
1000
10000
1980 1983 1986 1989 1992 1995 1998 2001 2004 2007 2010
Year
Kbit capacity/chip
1.6-2.4 µm
1.0-1.2 µm
0.7-0.8 µm
0.5-0.6 µm
0.35-0.4 µm
encyclopedia2 hrs CD audio30 sec HDTV
page
Die Size Growth
486Pentium ® proc
P6
100
Die size (mm)
Die size grows by 14% to satisfy Moore’s Law
Sp11 CMPEN 411 L01 S29
40048008
808080858086
286386
486Pentium ® proc
1
10
1970 1980 1990 2000 2010
Year
Die size (mm)
Courtesy, Intel
Clock Frequency
Lead microprocessors frequency doubles every 2 years
P6
Pentium ® proc100
1000
10000
Frequency (Mhz)
2X every 2 years
Sp11 CMPEN 411 L01 S30
Pentium ® proc486
3862868086
8085
8080
8008
40040.1
1
10
1970 1980 1990 2000 2010
Year
Frequency (Mhz)
Courtesy, Intel
Power Dissipation
P6Pentium ® proc
486
386
2868086
10
100
Power (Watts)
Lead Microprocessors power continues to increase
Sp11 CMPEN 411 L01 S31
3868086
80858080
80084004
0.1
1
1971 1974 1978 1985 1992 2000
Year
Power (Watts)
Courtesy, Intel
Power delivery and dissipation will be prohibitive
Power Density
100
1000
10000Power Density (W/cm2)
Nuclear
Reactor
Rocket
Nozzle
Sp11 CMPEN 411 L01 S32
4004
8008
8080
8085
8086
286386
486Pentium® procP6
1
10
1970 1980 1990 2000 2010
Year
Power Density (W/cm2)
Hot Plate
Power density too high to keep junctions at low temp
Courtesy, Intel
Power Density
100
1000
10000Power Density (W/cm2)
Nuclear
Reactor
Rocket
Nozzle
Sp11 CMPEN 411 L01 S33
4004
8008
8080
8085
8086
286386
486Pentium® procP6
1
10
1970 1980 1990 2000 2010
Year
Power Density (W/cm2)
Hot Plate
Power density too high to keep junctions at low temp
Courtesy, Intel
ITRS
� The “International Technology Roadmap for Semiconductors” (ITRS) is the industry’s prediction for the future of semiconductors.
� It is mostly an extrapolation of existing trends.
� The ITRS is often “slow”.
� http://public.itrs.net
Sp11 CMPEN 411 L01 S34
� http://public.itrs.net
Technology Directions: “Old” SIA Roadmap
Year 1999 2002 2005 2008 2011 2014
Feature size (nm) 180 130 100 70 50 35
Mtrans/cm2 7 14-26 47 115 284 701
Chip size (mm2) 170 170-214 235 269 308 354
Signal pins/chip 768 1024 1024 1280 1408 1472
Clock rate (MHz) 600 800 1100 1400 1800 2200
Wiring levels 6-7 7-8 8-9 9 9-10 10
Sp11 CMPEN 411 L01 S35
Wiring levels 6-7 7-8 8-9 9 9-10 10
Power supply (V) 1.8 1.5 1.2 0.9 0.6 0.6
High-perf power (W) 90 130 160 170 174 183
Battery power (W) 1.4 2.0 2.4 2.0 2.2 2.4
http://public.itrs.net
Technology Scaling
� Technology shrinks by ~0.7 per generation
� With every generation can integrate 2x more functions on a chip; chip cost does not increase significantly
� Cost of a function decreases by 2x
� But P
Sp11 CMPEN 411 L01 S36
� But P
� How to design chips with more and more functions?
� Design engineering population does not double every two yearsP
� Hence, a need for more efficient design methods
� Exploit different levels of abstraction
Design Abstraction Levels
SYSTEM
GATE
MODULE
+
Sp11 CMPEN 411 L01 S37
GATE
CIRCUIT
VoutVin
CIRCUIT
VoutVin
DEVICE
n+
S D
n+
G
Design Productivity Trends
Logic Tr./Chip
Tr./Staff Month.
x x
58%/Yr. compoundedComplexity growth rate
10,000
1,000
100
10
1
Logic Transistor per Chip(M)
10
100
1,000
10,000
100,000
Productivity
(K) Trans./Staff -Mo.
Complexity
Sp11 CMPEN 411 L01 S38
2003
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2005
2007
2009
xxx
xxx
x
21%/Yr. compoundProductivity growth rate
x0.1
0.01
0.001
Logic Transistor per Chip
0.01
0.1
1
(K) Trans./Staff
Courtesy, ITRS Roadmap
Complexity outpaces design productivity
Design Productivity Crisis
Year Tech. (nm)
Complexity Frequency 3 Yr. Design Staff Size
Staff Costs
1997 350 13 M Tr. 400 MHz 210 $90 M
1998 250 20 M Tr. 500 MHz 270 $120 M
1999 180 32 M Tr. 600 MHz 360 $160 M
2002 130 130 M Tr. 800 MHz 800 $360 M
Sp11 CMPEN 411 L01 S39
��We need to improve the productivity via design automationWe need to improve the productivity via design automation
� 1996: 100 person in P6 team
� 2007: 1600 person in P10 team
� Question: ?? Person in P38 team ?
� Answer: Every inhabitant of our planet
Major Design Challenges
� Microscopic issues
� ultra-high speeds
� power dissipation and supply rail drop
� growing importance of interconnect
noise, crosstalk
� Macroscopic issues
� time-to-market
� design complexity (millions of gates)
� high levels of abstractions
reuse and IP, portability
Sp11 CMPEN 411 L01 S40
� noise, crosstalk
� reliability, manufacturability
� clock distribution
� reuse and IP, portability
� systems on a chip (SoC)
� tool interoperability
Next Lecture and Reminders
� Next lecture
� Design metrics
- Reading assignment – 1.3
� Reminders
Sp11 CMPEN 411 L01 S41