meeting with the forthcoming ic design - asp-dac 1995 · pitfalls in power-gating 1) clock buffers...
TRANSCRIPT
T.Sakurai
Meeting with the forthcoming IC design– The era of power, variability and NRE explosion
and a bit of the future –
- Solving issues of IC’s by 3D-stacking –Prof. Takayasu Sakurai
Center for Collaborative Research,and Institute of Industrial Science,
University of TokyoE-mail:[email protected]
ASPDAC’07 Keynote Address II 9am – 10am Jan.25, 2007
T.Sakurai
Moore’s law continues as a backbone but…
Drain Source
Gate
0.2micron
Drain SourceGate
0.2micronSize 1/2
Size x1/2Voltage x1/2Electric Field x1Speed x3Cost x1/4
Power density x1.6RC delay/Tr. delay x3.2Current density x1.6Voltage noise x3.2Design complexity x4
Unfavorable effectsFavorable effects
T.Sakurai
Three explosions threat Moore’s law
Explosion of power
Explosion of integrity attackers
Explosion of complexity
Explosion of NRE*
*) Non-Recurring Engineering Cost
T.Sakurai
3D Stacking for solving issues
Silicon chips stacking
for lower power & lower NRE
Organic sheets stacking
for newer applications
With design tool implications
T.Sakurai
Power is a stumbling block to Moore’s law Po
wer
per
chi
p [W
]
1980 1985 1990 1995 20000.01
0.1
1
10
100
1000
Year
MPU
x4 / 3
years
DSP
x1.4 / 3 years
Processors published in ISSCC
2005 2010 2015
x1.1 / 3 years
ITRS requirement
10000
Dynamic
Leakage1/100
T.Sakurai
Trade-off between power and delay
012345x 10-7
0
10
20
0.30.50.7-0.1
0.10.3VDD [V] V TH [V
]
Pow
er [W
/gat
e]
Del
ay [p
s]
DDs
V
DD VIVCfaPowerTH
⋅⋅+⋅⋅⋅=−
1002
0.30.50.7-0.1
0.10.3VDD [V] V TH [V
]
( ) 31 .THDD
DD
VVVCDelay
−⋅
∝
Equi-delay
65nm node, FO3 INV
(+ other leakage)
Change power depending on required performance.
T.Sakurai
Example of power-aware design
G. Uvieghara, et al., “A Highly-Integrated 3G CDMA2000 1X Cellular-Baseband Chip with GSM/AMPS/GPS/Bluetooth/Multimedia Capabilities and ZIF RF Support,” ISSCC paper#23.3, Feb. 2004.
Multi-VDD: Seven VDD’sDual-VTH: Two VTH’s for PMOS and NMOS eachPower gating: Selective MTCMOS reduced standby to 1/3~1/4.
T.Sakurai
Power-gating to cut-off VDD of inactive blocks is getting a major leakage-
aware design style and seems easy but…
CMOS logic
Low-VTH
st’by
VDD
VSS
Virtual VSS Power switch
Z
T.Sakurai
Pitfalls in power-gating1) Clock buffers
2) Repeaters
3) Gates receiving power-gated signals
4) Registers will lose stored info
5) Other common blocks like analog and sleep signal generator
(1)(2)
(3)
‘0’
Power-gated block
D
(4)
Big and rather slow (µs-order) noise on power supply line when switched off and on.
T.Sakurai
Tool implications of power-gating1) High-level description of power-gating
2) Power estimation tool
3) Sizing of power switch
4) Delay estimation under the effect of the power switch
5) Automatic generation of sleep signal
6) Layout level considerations
7) Power line integrity at sleep and wake-up
New tools for each new design approach
T.Sakurai
Adaptive VDD & VTH control for lower power with variability
T.Sakurai
Adaptive voltage for variation control
V TH
varia
tion
VTH
Del
ayLe
akag
e
CenterLow High
Improved
Improved
VSS
VDD
VBS=VDD=>VDD±v
VBS=0 => 0 ± v
pMOS
nMOS
T.Sakurai
Transmeta Efficeon 1.5 GHz Fixed VDD
VID = 1.0 volt
Power Spec 9.5 watts
Courtesy: Transmeta Corp.
T.Sakurai
Transmeta Efficeon 1.5 GHz with Voltage ID
VID = 0.8, 0.85, 0.9, 0.95, 1.0 volts (5)
Power Spec6.5 watts
Courtesy: Transmeta Corp.
T.Sakurai
1.5 GHz with VTH Control and Fixed VDD
VID = 1.0 voltsPower Spec
4 watts
Courtesy: Transmeta Corp.
T.Sakurai
1.5 GHz with VTH and VDD Control
VID = .85, .90, .95, 1.0, 1.05 (5)
Power Spec 3.6 watts
Courtesy: Transmeta Corp.Statistical simulation tools with VDD and VTH adjustment
T.Sakurai
History and perspective of low-power VLSI systems designOnce upon a time on a peaceful chip of VLSI
before the notorious power war …
We were using single VDD and single VTHfor an entire chip.
Then, the power war began andwe started using multiple VTH and VDDdepending on the location on a chip.
Still, the power war gets severer andwe started to vary VTH and VDD in time
adaptively.
In finer granularity in time and space …with a help from circuit and software.
T.Sakurai
Future power-aware VLSI
Voltage converters & power blocksneed L & C.
• High-voltage VDD distribution to lower noise issues• Distributed voltage conv. for fine-grain adaptive VDD & VTH
1V 50W → 50A current5% noise → 0.05V noise → 1mΩ sheet R → 15µm thick CuThick interconnects on interposer, package or something else
Power-grid
T.Sakurai
Further power-awareness by 3D stacking
Specialized blocksSensors and information collecting circuits
Proc. unit
Stacked analog/RF
Power unit
Stacked memories
Power grid
Power control proc.
T.Sakurai
3-D is good for low-power and high-perf.
System-on-a-Chip
3-D assembly
More devicesin closer vicinity
↓Less R and C
↓Lower power
Higher performance↓
Supporting design tools are the enabling technology.
Substrate< 20µm thick
System simulation & synthesis with 3D, simulation with electro-magnetic interaction
T.Sakurai
Three explosions threat Moore’s law
Explosion of power
Explosion of integrity attackers
Explosion of complexity
Explosion of NRE*
*) Non-Recurring Engineering Cost
T.Sakurai
Exploding NRE (Non-Recurring Eng. Cost)
By Rick Merritt, EE Times Oct 25, 2002
◆ NRE for a SoC is getting $5M~10M and without over $20M sales it is difficult to make a SoC. by Bryan Lewis, Gartner Dataquest
◆ Number of design starts is declining from 1997.
ASIC
ASSP
02000400060008000
100001200014000
2000 2001 2002Year
Des
ign
star
tshttp://www.eetimes.com/printableArticle?doc_id=OEG20021025S0052
Generation [μm]
Mas
k Se
t Cos
t (M
$)
Source: Rahul Goyal, Intel –Nov 2001 SEMI Mtg
0
0.5
1
1.5
2
2.5
0.4 0.25 0.2 0.180.150.1390n65n
T.Sakurai
Development is also risky
Dr. Kurt Keutzer, Ahmed JerrayaSpecial Session: “How do you design a 10M gate ASIC?” 2002 Design Automation ConferenceDr. Fumiyasu Hirose, Cadence Japan
Fix 1
Development (months)
Design, Fab, Test
Rev
enue
12
Time-to-Market (Samples)Time-to-Volume (Production)
ShipmentsWindow
181260
Fix 3
Shipments(months)
621 / 0
Fix 2
2nd Si OK
20%1st Si
OK
60%
3rd Si OK
15%
3rd Si not OK 5%
T.Sakurai
CANDE meeting 2005The final 5-year predictions (to be reviewed in 2010) were:
1) IC-Package CAD will be a part of standard design flow 2) India and China will have more EDA startups than U.S.3) Analog Designers will still resist high-level models and
languages 4) A complete Open Source RTL-GDS tool flow will exist5) Nearly all EDA tools will take advantage of multi-
processors 6) Fewer than 200 commercial chips released in 45 nm
technology7) No practical nanotech computing products 8) SPICE-type simulators will still be the workhorse of
analog designs 9) Moore's law will be dead10) System in Package will boomhttp://www.ics.uci.edu/~rgupta/cande/
T.Sakurai
System-on-a-Chip
3D stacking of chips/packages will be an assembly style suitable for low-NRE systems
MPU Core
CacheROM
Logic
AnalogUSB
DRAM
3D SiP or stacked packages
MPU Core
Logic
AnalogCache
USB ROM
DRAM
System-on-a-BoardLower NRE, QTATHigher-perf., lower power
T.Sakurai
ProcessorFPGAAnalog TCP/IPUSB
Amortizing NRE by volume production
RF
Interface VideoDVI DC-DC Graphics
MemoriesDRAM SRAMNVA/D D/ABluetooth
GSM
“Mix & Match” increases number of chips per development.
Systems
Components
T.Sakurai
Mixed business model
Tota
l cos
t / s
ales
[M$]
Number of chips sold [million units]
SoC total cost with $10M NRE & $10 per unit
0 1 2 3 4 50
50
100
SiP total cost with $0.1M NRE & $15 per unit
Total sales ($20 per chip)
Design tools from high level to physical level are needed.
T.Sakurai
Wired WirelessFa
ce to
face
Mor
e th
an 3
chip
s
[1] T.Ezaki [2] J.Burns [3] K.Kanda [4] D.Mizoguchi, (ISSCC’04, ’01, ’03, ’04)
Interconnections for new 3D SiP’s
Micro-bumps [1]
Substrate vias [2]
Capacitive coupling [3]
Inductive coupling [4]
T.Sakurai
Through substrate via (TSV)
http://www.zy-cube.com
Substrate via
Micro-bump
Adhesive layer
Adhesive layer
Micro-bump
By ZyCube
T.Sakurai
3D wafer stacking for high performance
Courtesy of Kerry Bernstein, IBM T.J. Watson Research Center
Via size is 0.2 microns diameter at the top, and 0.14 microns at the bottom.
SOI wafer is thinned down to several microns.
T.Sakurai
Inductive wireless superconnect
D. Mizoguchi, Y. Yusof, N. Miura, T. Sakurai, T. Kuroda, "A 1.2Gb/s/pin Wireless Superconnectbased on Inductive Inter-chip Signaling (IIS) ,”ISSCC’04, pp. 142-143, Feb. 2004. N. Miura, D. Mizoguchi, Y. B. Yusof, T. Sakurai, and T. Kuroda, “Analysis and Design of Transceiver Circuit and Inductor Layout for Inductive Inter-chip Wireless Superconnect,”Symp. on VLSI Circuits, pp. 246-249 June 2004.
for 3D-stacked VLSI’sMetal Inductor
Chip#1(Logic)
Chip#4
Chip#2(Memory)
Tx
RxChip#3
Tx
RxChip#1
Tx
RxChip#2
VR
VR
ITIT
Chip#3(Memory)
Clock & Power
Analog
T.Sakurai
1Tbps inter-chip wireless by 30µm coils
Tx chip (upper)1024ch data tranceiver
Rx chip (lower)
Coils for data Tx
30μm
Chips photo
15μm
Coils for clock Tx
200μm
Tx chip
Rx chip
Clock tranceiver
BIS
TC
lock
con
trol
0.18μm CMOS, 3fJ/bit energyN.Miura, D.Mizoguchi, M.Inoue, K.Niitsu, Y.Nakagawa, M.Tago, M.Fukaishi, T. Sakurai, T. Kuroda, "A 1Tb/s 3W Inductive-Coupling Transceiver for Inter-Chip Clock and Data Link," ISSCC'06, Paper#23.4, Feb. 2006.
Low-loss interconnects are preferable for micro transformers.
T.Sakurai
Inductively-coupled inter-chip communicationB
and
wid
th [b
/s]
’961G
10G
100G
1T
Year’98 ’00 ’02 ’04 ’06
Hitachi
NEC
NEC
NEC
TINTT
This Work
Our Work
Intel
Toshiba
Rambus
Hotrail
Sony, TeraChipOur Work
RambusFlexIO
1
10
100
1000
Ener
gy p
er b
it [m
W/G
b/s=
pJ/b
]
’96 ’98 ’00 ’02 ’04 ’06Year
Toshiba
NTT
Hitachi
NEC
NECIntel
TI
NEC
Our Work
Our Work
This Work
TeraChip
RambusFlexIO
N.Miura, H.Ishikuro, T.Sakurai, T.Kuroda, "A 0.14pJ/b Inductive-Coupling Inter-Chip Data Transceiver with Digitally-Controlled Precise Pulse Shaping," Paper#20.2, ISSCC, Feb.2007.
’07
Currentproducts
Circuits and electro-magnetism hybrid simulation.
T.Sakurai
Even wireless power transmission
700x700µm2
K.Onizuka, H. Kawaguchi, M. Takamiya, T. Kuroda and T. Sakurai, “Chip-to-Chip Inductive Wireless Power Transmission System for SiP Applications,”Paper#15.1, CICC, Sept. 2006.
1
2
3
0 0.2 0.4 0.6 0.8 1 1.2 1.4Output Voltage (V)
Tran
smitt
ed P
ower
( mW
)MeasuredSimulated
RL=100Ω
T.Sakurai
VDD, High Voltage
LF
CF
RS
Load
Controller
Output filter
VDD,INTERNAL
On-chip distributed DC-DC converter
Need good low-resistive L & high-capacitive C.They are also needed for low-power RF circuits.
Si with L and C
Chopping frequency > 300MHz
T.Sakurai
Distributed DC-DC converter by 3D stacking
K.Onizuka, H. Kawaguchi, M. Takamiya, and T. Sakurai, “Stacked-chip Implementation of On-Chip Buck Converter for Power-Aware Distributed Power Supply Systems,” A-SSCC, Nov. 2006.
01020304050607080
0 20 40 60 80Output Current (mA)
Effic
ienc
y (%
)
MeasuredSimulated(HSPICE)
T = 5nsf = 200MHz
Ripple<10%2mm
Cap1nF
PAD
Output filter chipOn-chippowersupplies SoC/SiP
Package
L&C cells
Pads
Upper Chip(Output Filter)
Lower Chip(Active & Load)
T.Sakurai
5ns transition on-chip power supply
@ 0.18µm CMOS
Quick dropper
Linear regulator
Load 80µm
K.Onizuka and T.Sakurai, "VDD-Hopping Accelerator for On-Chip Power Supplies Achieving Nano-Second Order Transient Time," A-SSCC'05, Paper#6.1, Nov. 2005.
VDDHVDDL
< 5ns
Transition time smaller than 5ns
Measurement
For 25k-gatesequivalent load
T.Sakurai
3D Stacking for solving issues
Silicon chips stacking
for lower power & lower NRE
Organic sheets stacking
for newer applications
With design tool implications
T.Sakurai
Emergency control
On-demand logistics
Wearable
Individual adaptive marketing
Accidentavoidance
Nano sensor-net on bodyPersonal dataSmart shelfSmart cashier
Pedestrian assistance
Productiondata tracking
Anti-theft
Life supporting robots
Remote medicalassistance
Health monitoring
Environment monitoring
Rescue support Anormaly monitoring& alarming
Physical space
Virtual space
New electronics targets physical space
Disaster prevention
Human bodyHuman body
Transportation systemTransportation system
HomeHome
EnvironmentEnvironment
Town & streetTown & street
Original drawing: Courtesy of Prof. Morikawa, Univ. of Tokyo
T.Sakurai
Increasing VLSI penetration into people’s lifePr
oces
sor c
ount
per
per
son
1950 1960 1970 1980 1990 2000 2020Year
Large scale
Office / middle
WS
PC
historicalVac. tube
Transistor
IC
LSI
VLSI
0.001
0.01
0.1
1
10
100
Embedded
UbiquitousComputing
System LSI
1000SiP
Electronics is part of environments enhancing convenience and security of daily life.
People use electronics consciously.
UbiquitousElectronics
Ubiquitous devices
2010
T.Sakurai
Food 25.74%
Housing 7.82%
Electricity/gas 6.63%
White appliances 0.52%
Air conditioner 0.34%
Housing (others) 2.56%
Clothing 5.08%
Medical 3.67%
Transportation 2.46%
Automobile 1.58%
Transportation 4.61%
Communication 1.57%
Phone 1.52%Other comm.0.36%
Education 3.46%
TV 0.24%
Audio0.02%
Tape recorder0.01%
Others 20.92%Edutainment 5.90%
Books1.66%
Education appliance 2.32%
TV game machine 0.09%
Entertainment 2.32%
Video camera0.04%Camera 0.08%
VTR 0.03%
PC 0.46%
Emerging App.: Ubiquitous Electronics
Expenditure distribution of Japanese household
Digital consumer
Ubiquitous electronics is required to provide physical applications.
T.Sakurai
Required innovations forubiquitous electronics
Huge number of devicesExtremely low power (3D SiP)Low cost (3D SiP)
EverywhereVery short distance communication (L & C stacked)Ubiquitous energy source (Large-area electronics)
Interfaces to real-worldSensors & actuators (Large-area electronics)Heterogeneous systems (Heterogeneous ICs stack)
Design tool needs for 3D hybrid systems synthesis and simulation
T.Sakurai
Features of organic IC’s
Advantages
Low-cost manufacturingMechanical flexibility
Disadvantages
Low speed (<10-5 of Si VLSI)Low density (<10-4 of Si VLSI)
T.Sakurai
Organic electronics
RFID tags
Kennedy group
University of TokyoPioneer
Studio Del Sole
Infineon
Polymer Vision
Samsung
Sensors WearableElectronics
Solar Cells
Smart Cards
Message Boards
FlexibleDisplays
Flat PanelDisplay Gyricon
T.Sakurai
Printable electronics
T.Sakurai
Organic transistors
Source Drain
Insulator+ -
Gate
OFF ONVoltage
+ -Current
Pentacene
Organic semiconductors: main elements --- C & HSlow but flexible and low-cost manufacturing
Organic semiconductor
T.Sakurai
SPICE still works•Match level 1 SPICE MOS model with 200kΩ
Mobility 0.5 ~ 1.4cm2/VsIon/Ioff >106
L=100µmW=2mm
-40 -30 -20 -10 00
10
20
30
VDS [V]
I DS
[µA
]
VGS=-40V
-30V
-20V
-10V
200k
200k
MeasurementSimulation
S
DG
Level 1 SPICE MOS model
Bending proof
Preferably new model like μ ∝ (VGS-VTH)γ (0<γ<1)
T.Sakurai
Large-area electronics by stacked sheets
E-skin
IEDM’03ISSCC’04
Sheet scanner
IEDM’04ISSCC’05
Slow but low-cost for large-area (Si can’t cover)
Braille display
IEDM’05ISSCC’06
IEDM’06ISSCC’07
Wireless power
transmission sheet
Pressure sensors + OFETsPhotodetectors + OFETs
Actuators + OFETsCoils + MEMS + OFETs
T.Sakurai
Braille display by organic FETs
Y.Kato, S.Iba, T.Sekitani, Y.Noguchi, K.Hizu, X.Wang, K.Takenoshita, Y.Takamatsu, S.Nakano, K.Fukuda, K.Nakamura, T.Yamaue, M.Doi, K.Asaka, H.Kawaguchi, M.Takamiya, T.Sakurai, and T.Someya, "A Flexible, Lightweight Braille Sheet Display with Plastic Actuators Driven by An Organic Field-Effect Transistor Active Matrix," IEDM'05, Paper #5.1, Dec.2005.
M.Takamiya, T.Sekitani, Y.Kato, H.Kawaguchi, T.Someya, and T.Sakurai, “An Organic FET SRAM for Braille sheet display with back gate to increase the static noise margin,”ISSCC’06, Paper #15.4, Feb. 2005.
T.Sakurai
Plastic actuatorsUp
The displacement of the actuators to read Braille is 0.2 mm.
Actuator
Braille dot
Down
100μF
Equivalent circuit
T.Sakurai
Speed up by higher-voltage drive
Time (s)0 5 10 15 20 25 30 35 40
V AC
T(V
)
0
-1
-2
-3
-4D
ispl
acem
ent
of a
ctua
tor (
mm
)
00.20.40.60.8
VX = –3V
–5V–10V–20V
VX = –3V
–5V
–10V–20V
V PL
(V) 0
VX
17x speeding up
49000/20
Braille dot
DATAb(-20V)
VACT
Actuator
VPL
VPH = floating
Reliability limit
T.Sakurai
Overdrive for speed and reliability
Time (s)0 5 10 15 20 25
Dis
plac
emen
tof
act
uato
r (m
m)
0
0.2
0.4
-0.2
-0.4
10
5
0
-5
-10
Volta
ge (V
)DATA
DATAb
VPL
VPHVACT
1.6 s 0.74 s
Braille dot isUp
Down
~30s 1.6s
Hybrid simulation will be required (Mechanical – Electrical)
T.Sakurai
M.Takamiya, T.Sekitani, Y.Kato, H.Kawaguchi, T.Someya, and T. Sakurai, "An Organic FET SRAM for Braille Sheet Display with Back Gate to Increase Static Noise Margin," ISSCC'06, Paper#15.4, Feb. 2006.
FrameActuator
OFETsdriver
OFETsSRAM
Braille sheet display stacking four sheets
T.Sakurai
Braille reading test
All of 4 blind subjects can read our Braille correctly.
T.Sakurai
Wireless power transmission sheet with plastic MEMS switches and OFETs
T.Sekitani, M.Takamiya, Y.Noguchi, S.Nakano, Y.Kato, K.Hizu, H.Kawaguchi, T.Sakurai, and T.Someya, "A large-area flexible wireless power transmission sheet using printed plastic MEMS switches and organic field-effect transistors," Paper#11.1, IEDM 2006, Dec. 2006.
M.Takamiya, T.Sekitani, Y.Miyamoto, Y.Noguchi, H.Kawaguchi, T.Someya and T.Sakurai, "Design Solutions for a Multi-Object Wireless Power Transmission Sheet Based on Plastic Switches," Paper#20.4, ISSCC, Feb. 2007.
T.Sakurai
Efficiency ~ 0.1%
Efficiency > 60%
Position-sensing and selective activation Large coil
30x30 cm2 X 1 coil
Receiver coil
Receiver coil
1 inch2 X 64 coils
Many coils& one selected 1 inch2
Selective activation is the key.
1 inch2
Electro- magnetic induction works
T.Sakurai
Stacking sheets for MEMS and Organic FETs
Wireless power transmission system
Power transmission coils
Plastic MEMSswitches
Low lossSlow ~ 0.1s# of switching limited
Organic FETsPosition-sensing coils Contactless
position-sensing system
21 x 21 cm2 (8 x 8 cells)
ResistiveFaster < ms# of switching unlimited
Hybrid simulation of IC, mechanics and electro-magnetism
T.Sakurai
High power
Contactless position sensingLarge-area & Low cost
Lightweight & Printable
Wireless power transmission sheet (3D-stacked)
Size : 21 x 21 cm2Thickness : 1 mmWeight : 50 gEfficiency : 62.3%Max received power : 29.3 W
T.Sakurai
X’mas tree w/o a battery wirelessly powered
21 LEDs
13.56 MHz
Received power : 2 W
T.Sakurai
No electrical shock
I touched it by my hand. No problem ☺
T.Sakurai
Providing infrastructure ubiquitous electronics
Home-care robot Vacuum cleanerIn the floor
In the wall
TV on a wall
In the table
Mobile phone & PC & e-accessories(data can be wireless but USB’s wire delivers power)
Ambient illumination
Wirelessly powered room in the future
T.Sakurai
Summary
3-D stacking, new frontier of integrated circuits, is opening up new electronics.
Solving issues for higher performance, lower power, lower cost and newer functions
New design tools for hybrid simulation and synthesis to fully enjoy 3-D stacking