Chip PC Technologies Innovative Desktop Solutions for Server ...
Diversified Technologies for System-on-Chip
Transcript of Diversified Technologies for System-on-Chip
1/19/2005 1®
Diversified Technologies for System-on-Chip
Polytechnique, 9 March, 2004
Gerard Mas, Programme Management, Director Central R&D, Crolles
1/19/2005 2®
AgendaThe Market trendsCMOS roadmap and added-value optionsTechnology support for Low-power• Optimized MOS, multi-Vt• SOI• eDRAM
Design / Manufacturing linksSystem On Chip Design ChallengesCompetition.Conclusions
1/19/2005 3®
23%North America 5%
Japan
33%Asia
Pacific
36%Europe
3%Region
5*
2000 Sales: US$ 7.8B43,000 employees17 main production sites9 advanced R&D sites35 design and application centers62 direct sales offices in 24 countries3th* largest semiconductor supplier
(*source Dataquest)
2000 Sales: US$ 7.8B43,000 employees17 main production sites9 advanced R&D sites35 design and application centers62 direct sales offices in 24 countries3th* largest semiconductor supplier
(*source Dataquest)* Emerging markets
SMicroelectronics - A Global Semiconductor Company
1/19/2005 4®
Les prochains defis de l’industriede la microelectronique
Des defis d’ordre technique• Technologie Nanometre• Systeme sur siliciumDes defis d’ordre economique• Alliance• co-competition
1/19/2005 5®
The market trends
1/19/2005 6®
Living-Room Appliances
Audio Out
Video OutSatellite
DVD
Cable
Dig Terrestrial
MPEG A/VDecoder
BroadbandInterface
ONE PLATFORM
1/19/2005 7®
DVD – Variety of Applications
Basic DVD Player
DVD Recorder
Personal DVD
DVD TV Combos
DVD Receivers
DVD Audio
In Car Navigation
1/19/2005 8®
Migration of Digital TV
Analog Signals+
Digital ImageProcessing
Analog TV +
Free to AirDigital Decoder
Terrestrial Tuner
Fully Digital –Digital Signal,
Digital Tuner/DecoderImage Processing
Home Network
Home NetworkEthernet – 802.11 - HPNA
IP Telephony
TV BroadBandMedia Streaming
Cable/Satellite/DTT
PC BroadBandWWW Connection
CableModem/ADSL
WirelessInterconnect
Home NetworkRouters
Home Entertainment SystemsMedia Servers
1/19/2005 9®
Convergence of Technologies…
RFRF SecuritySecurityLow Power
Low Power
WEBWEB Flat PanelFlat
PanelCPUCPUHDDHDD NetworkNetwork
xDSLxDSL VoIPVoIPSecure NetworkSecure NetworkISDNISDN Cable
ModemCable
Modem
VideoVideo AudioAudioSecure AccessSecure Access
ImageSensors
ImageSensors GamesGames Video
Analog
Digital
CM
OS
DIG
ITA
LC
ON
VER
GEN
CERFWIRELESS
Analog/MixedSignal
CONSUMER
COMPUTER
TELECOM
1/19/2005 10®
1/19/2005 11®
……Divergence of Divergence of Applications
AUTOMOTIVE
Secure NetworkSecure Network
AudioAudio
Flat PanelFlat
Panel
HDDHDD
ImageSensors
ImageSensors
VideoVideoCPUCPUVideoVideo
MULTIMEDIAPHONE
RFRF
SecuritySecurityLow Power
Low Power
WEBWEB
NetworkNetwork
OFFICECOMPUTER
Secure NetworkSecure Network
Flat PanelFlat
Panel
HDDHDD
VideoVideo
CPUCPU
xDSLxDSL
HOMECONSUMER
VoIPVoIP
HDDHDD
Flat PanelFlat
PanelVideoVideo
ImageSensors
ImageSensors
CPUCPU
xDSLxDSL
CableModemCable
ModemCM
OS
DIG
ITA
LC
ON
VER
GEN
CE
Secure NetworkSecure Network
1/19/2005 12®
More Options in shorter timeN
b of
opt
ions
0.18µm
0.12µm 90 nm
AnalogRFBiCMOSe-SRAMe-DRAMPassivesetc…
Years
1/19/2005 13®
High Performance Technologies for …
Very Low power consumption in SoCand also…Ultra Low noise RF performanceHigh Speed digital and analogue (SiGe)Refined Analogue precisionEnergy Management (high voltage)High Density Memories (eDRAM, MRAM, Flash)
1/19/2005 14®
CMOS roadmap and added-value options
1/19/2005 15®
CMOS roadmap
70 -
100 -
130 -
180 -
Generation(nm)
0.18 Shrink110 kgates/ mm²
0.18 µm6 ML/ 1.8 V
85 kgates/ mm²
m0.12 µ6-8 ML/ 1.2 V
200 kgates/ mm²
90nm7-9 ML/ 1.0 V
360 kgates/ mm²
1998 1999 2000 2001 2002 2003 2004
65nm9-10 ML/ 0.9 V
650 kgates/ mm²
200mm
Right edge of boxes is PG acceptance
1/19/2005 16®
CMOS Roadmap: from R&D to production
2001 2002 2003 2004 2005 2006 2007
1st silicon
Protos acceptance
Production start
CMOS 45 nm
Advanced R&D
CMOS 32 nm
Advanced R&D
CMOS 90 nm
CMOS 65 nm
Advanced R&D
CMOS 120 nm
4 advanced CMOS platforms developed in parallel
1/19/2005 17®
Evolution in the last 10 yearsContact in 0.5µm = 0.64µm2
6-transistor SRAM cell in 65nm
From HCMOS5 (1992) to CMOS065 (2002)
1/19/2005 18®
CMOS090 300mm 6LM integration
Metal 6
Metal 1
Metal 2
Metal 3Metal 4
Metal 5
1/19/2005 19®
Extending Core CMOS for SOC
CMOS coreprocess
BICMOSRF (cellular phone)Fiber Optic IC
HFCMOSPower Management (cellular phone)
SRAM/DRAMPC peripherals
AnalogTuner (Set-Top Box)
RFCMOSBluetooth
ImagerWeb Cam
1/19/2005 20®
ABOVE IC INTEGRATIONSystem on Chip (SOC) approachOpen the door for innovative architecture & RF front-end
CoCo--integration of ICs, passives and RFMEMS integration of ICs, passives and RFMEMS New RF functions through strong coNew RF functions through strong co--development development
between technology & RF designbetween technology & RF design
High Q inductor
FBAR filter
RF Micro-switch
RFIC circuit
1/19/2005 21®
SoC’s with CMOS-Imager process
For mobile phone applications:ZS450: CIF format (~100000 pixels)
For webcam applications:ZS422: QVGA format (~75000 pixels)
=> audio/video/video processing (SOC)
Area24.0 mm²
Area39.6 mm²
ZS550: VGA format (~300000 pixels)
Area24.1 mm²
1/19/2005 22®
Production 0.18um SoC’s with eDRAM
Low-End Printer High-End Printer(20mm²) (34 mm²)
4 Mbits eDRAM
6 Mbits eDRAMCamera for cell phone
(15 mm²)Includes ARM microIncludes ARM micro
3 Mbits eDRAM
Disk Controller DVD recorder(26 mm²) (34 mm²)
4 Mbits
eDRAM
Includes ST10 microIncludes Super10 micro
1/19/2005 23®
Low-Power
Process / Design Trends
1/19/2005 24®
2 opposite requirements for the future
Increase processing power (GOPS):• Video, Audio, Graphics, Communications:
High performance dedicated processorsA lot of embedded memories
• Replace Analog by Digital (e.g. digital Radio)
Reduce power consumption (Watt/Op):• Dynamic power = CV2f
C: Need SOI-like junctions and LowK dielectricsV: decrease VDD to the minimum to achieve Frequency fF: keep Frequency pretty low, use parallelism
• Static power = as low as possible
1/19/2005 25®
Power Density in Microprocessors
i386 i486
Pentium
Pentium IIPentium Pro
Pentium III Pentium IV
Nuclear Reactor
1
10
100
1000
i386 i486 Pentium Pentium Pro Pentium II Pentium III Pentium IV NuclearReactor
W/c
m2
i386 i486
Pentium
Pentium IIPentium Pro
Pentium III Pentium IV
Nuclear Reactor
1
10
100
1000
i386 i486 Pentium Pentium Pro Pentium II Pentium III Pentium IV NuclearReactor
W/c
m2
1.51.5µµ 11µµ .7.7µµ .5.5µµ .35.35µµ .25.25µµ .18.18µµ .035.035µµ
Hot PlateHot Plate
Fred Pollack, IntelFred Pollack, Intel
Required Performance for Multi-Media Processing(source ITRS Design ITWG July 2003)
GOPS0.01 0.1 1 10
VideoVideo
AudioAudioVoiceVoice
CommunicationCommunicationRecognitionRecognition
GraphicsGraphics
FAXModem
2D Graphics
3D Graphics
MPEGDolby-AC3
JPEG
MPEG1Extraction
MPEG2 ExtractionMP/ML MP/HLCompression
VoIP Modem
Word Recognition
Sentence Translation
100
Voice Auto Translation
10Mpps 100Mpps
MPEG4
Face RecognitionVoice Print Recognition
SW Defined Radio
Moving Picture Recognition
1/19/2005 26®
GOPS: Giga Operations Per Second
SoC at the heart of conflicting trends
ST20ST20
MPEG2 MPEG2 VideoVideo
Dolby Dolby AC3AC3
FEIFEI
LinkLink
2 x 3 DAC2 x 3 DAC
Den
cD
enc
Time-to-market:Time-to-market:Process roadmap
accelerationConsumerization
of electronic devices
Process roadmap acceleration
Consumerizationof electronic devices
Complex systems:Complex systems:uCs, DSPs HW/SWSW protocol stacksRTOS’sDigital/Analog IPsOn-Chip bussesProcess options
explosion (analog, RF, imagers, …)
uCs, DSPs HW/SWSW protocol stacksRTOS’sDigital/Analog IPsOn-Chip bussesProcess options
explosion (analog, RF, imagers, …)
1/19/2005 27®
Deep sub micron effects:Deep sub micron effects:crosstalk
electro migrationwire delays, on-chip-variation
mask costs (OPC, PSM)copper wires
crosstalkelectro migration
wire delays, on-chip-variationmask costs (OPC, PSM)
copper wires
1/19/2005 28®
Static versus Dynamic Power
Dynamic Power per Chip is rather stable due to system limitation: mechanical / batteryDynamic Power is under control: P=CV2f• VDD is decreasing (slowly)• Power management is in practice (clock gating, power shut
down, …)Static Power is not under control: natural leakage of transistor multiplied by 10 every 2 years:• Power shut down is the only way to cut leakage• Very low voltage retention or zero leakage Non Volatile
memories are potential solutions (Flash, MRAM)Static Power is becoming as high as Dynamic Power at high temperature (wasted power, no solution)
1/19/2005 29®
Gate and Interconnect Delay vs technology node:
Good for Dynamic Power as well
05
1015202530354045
0.65 0.5 0.35 0.25 0.18 0.13 0.1
technology (µM)
dela
y (p
S)
Interconnect Delay, Cu & Low KInterconnect Delay, Al & SiO2Gate DelayTotal AlTotal Cu
Al 3.0µOhm-cm
Cu 1.7µOhm-cm
SiO2 K = 4.0
Low K K = 2.0
Al & Cu 0.8µ thick
Al & Cu line 43µ long
1/19/2005 30®
1
10
100
1000
10000
100000
300 400 500 600 700 800
Ion (µA/µm)
Ioff
(LO
G pA
/µm
)
C350C250
C180
C090GP
C120GP
C120LP
Ion/Ioff evolution is a BIG concern
1/19/2005 31®
SOI ?Advantages:• Reduced S/D junction capacitance• Full dielectric isolation => latch up immunity• BOX & STI cap are nearly null
Benefits:• power/performance improvement
C
GATE
TR.OX
SOURCE DRAIN
P -WellC2 C2
SiSUBSTRATE
N+ N+P+
TR.OX
BULK
TR.OX
C
D
GATE
BURIED OXIDE
Si SUBSTRATE
BURIED OXIDE400nm
TR.OX
SOURCE DRAIN
C1 C1
Si 150nm TR.OXN+ N+P+
SOI
1/19/2005 32®
Results in 120nm SOI
improved performances (on SRAM)
The gain achieved on 0.12um SOI technology corresponds to the next generation on Bulk = 90nm
Power Supply
Speed Gain
Dyn Power Gain
LeakageGain
1.08V 0 % 20% x 0.1
1.2 V 10% 17% X 0.15I
1.32 V 20% 14% x 0.5
1/19/2005 33®
Flash NOR Cell – ST vs. ITRS Roadmap
0.2480.186
0.34
0.125
0.061
0.81
0.16
1.2
0.32
0.46
0.080.05
0.01
0.1
1
10
0.0 0.1 1.0Technology [um]
cell
size
[um
2 ]
ITRS 2001
ST
10 F2
1/19/2005 34®
Low Power Design ActivitiesSystem Level
RTOSMemory size and bandwidth optimisationArchitecture Power/Performance trade-off
RTL to LayoutActivity control
Gated clockPower based synthesis
Leakage controlSupply controlVT control thru back-biasMix High-Speed and Low-Leakage
Smart Power/Low-voltage operationMulti-voltage supportOn-chip voltage regulator(s)
Disk Drive Controller0.13um
0.25um SiGe BiCMOS
GSM Power Management
Research activities Labs / Universities
90nm ST20 / ST40
1/19/2005 35®
Low-Power techniques for
wireless Power-up control
VddSleep
Vdd
Core power pads with switches
Vddfillercut
Io pad supply
Vdd Sleep pad zone
Two supply zones: Vdd and VddSleep.Specific rules at RTL for boundary between Vdd and VddSleep areasGnd is common for all cells.Pads are always powered.3 power pads with embedded switches
1/19/2005 36®
Single Package Triple Stacking
Multichip chip wire bonding
Triple stacked die
1/19/2005 37®
Design / Manufacturing links
1/19/2005 38®
Design For Manufacturability Gains
Reliabilityimprovement
Acceleration for ramp-up process (~months)
Yield gain for mature process (3 to 5%)
Standard approach
DFM approach
Process Maturity
Prod
ucts
Yiel
d
Robustnessincrease(versus design marginalities)
1/19/2005 39®
> Library and IP maturity, knownresults
> Chip contents
> Engineering data
LibYielddatabase
Impact of Test: From Design to Volume Manufacturing
Qty (Millions Parts)
TimePRE
PRODUCTIONPROTOTYPING MASS PRODUCTIONDESIGN
Production Ramping
Up Review
Productcharacterisation
TAG SCAN
PG ReviewMaturityCheckpoint
Yield Data Review
1/19/2005 40®
Predictive Chip Yieldbased on IP content
0%
20%
40%
60%
80%
100%
DP
R2H
D
RO
8L
SP
S2H
D
SP
S6
AR
M
CO
RE
LIB
8DH
S
CO
RE
LIB
8DLL
PLL
81
IOLI
B_8
0
LF10
othe
r
------
-
CH
IP
YieldYield W/ rep
Design / Manufacturing Yield Optimization
Design For Test
FailuresFault Isolation
L1
L2
Analog
Digital
1/19/2005 41®
Automatic Test Equipment
Wafer
Reticle
ManufacturingDesign
FIBDesign Intelligence:
CD SEM
Critical Regions and SitesFailed Nets
Wafer InspectionCourtesy of Cadence, Applied Materials
1/19/2005 42®
SOC design challenge
1/19/2005 43®
Today’s Reality
Complex SoC projects are draining the companies resources and developments are lagging behind the technology capacity.Today SoC project non-recurring expenses (NRE) are 10-50M$ should have 50-500M$ revenues to balance an acceptable ROI.Several $B already invested on IPs and integrated functions during last years.Design re-use must be exploited at all levels included sub-system re-use.
$-
$5.00
$10.00
$15.00
$20.00
$25.00
$30.00
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
ASIC Revenue/Design Start Design Costs
Revenue per ASIC is Not Increasing as Fast as Fixed Costs of Cutting Edge Design
0.35µm 0.25µm
0.18µm
0.13µm
90nm
$ Millions
Source: Gartner Dataquest 2002/2003; I.B.S. 2002…Custom EDAC Study
1/19/2005 44®
1/19/2005 45®
Photomask Costs Are Increasing…
$102$167
$310
$985
$62826 2830 32 34
$-
$100
$200
$300
$400
$500
$600
$700
$800
$900
$1,000
0.35m 0.25m 0.18m 0.13m 0.09m
Process
($Th
ousa
nds)
05101520253035404550
Tota
l Mas
k La
yers
Mask Set Cost Mask Layers
Source: IC Insights, 1/2003
1/19/2005 46®
S/W and H/W Complexity Factors1012
19701960 1980 20001990 2010 2020
Gates/chip: 2x / 18 monthseSW/chip: 2x / 10 monthseSW productivity: 2x / 5 years
Lines of code
H/W: +50% / year
eS/W
: +14
0% / y
ear
Gates
1010
108
106
104
102
1
1/19/2005 47®
Closing the Design Productivity Gap
1981 1995 2009
CAGR 58%
100000
0.01
Pro
duct
ivit
y(K
)Tra
ns.
/Sta
ff-M
o.P
rodu
ctiv
ity
(K)T
ran
s./S
taff
-Mo.
CAGR 21%
System LevelSpecification
HW-SWCo-verification
IP Reuse
1000
Logi
cTr
ansi
stor
s pe
rC
hip
(M
)Lo
gic
Tran
sist
ors
per
Ch
ip (
M)
0.001
1/19/2005 48®
Embedded Software Development Requires as Much/More Design Effort Than
Hardware
Concurrent Hardware/Software Design
Standard Flow
Spec Archi Design Fab Breadboard Softwaredevelopment
SystemIntegration
SystemValidation
Time
GAINSystem
ValidationSoftware
development
Methodology Extensions
SystemIntegration
1/19/2005 49®
1/19/2005 50®
HW/SW fast prototype platform
Faithful representation of the final designAvailable much sooner than the final siliconGuaranties the real-time behaviorValidation of the fundamentals of the SoCHW/SW architecture
HW Emulation
CELARO
RTL Design
MMCI/F
FLASHI/F µ8051
MUX
AnalogBlocks
To MMChost
ToFLASH
RTL Design
MMCI/FMMCI/F
FLASHI/F
FLASHI/F µ8051µ8051
MUXMUX
AnalogBlocksAnalogBlocks
To MMChost
ToFLASH
• 1w setup-time• ~0.5MHz
• Real M58LW128FLASH chip in-circuit
• SW host running @100KHz• Connection of HW host ongoing
1/19/2005 51®
1/19/2005 52®
First Prototype Platform: Aptix
ST120 Module
ARM7 Module
Interface to RF/Tester
Virtex FPGA (GSM/GPRS modem)
Memory boards
1/19/2005 53®
FPGA-based Prototyping Environment
RF GSM/GPRS Board
JTAG Connection to SW Debuggers
STEP1 BaseBandBoard
1/19/2005 54®
IP Reuse ProgramOrganization
company program “Design Methodology & IP Reuse”, CEO sponsoredcorporate driven CR&D + cross-divisional Committeedomain-specific Work Groups RTL2Layout, AMS, SLD, DFT, Functional Verif., Power, …intranet information site CAD On Line portal > K9 IP Reuse Pages
Reuse standardsadherence to industry approach VSIA; RMM; Qualitydeliverables / views BlueBook + Unicad ExtensionIP packaging bbview (mapping from BB logic views to IP physical files)HDL coding style Design Conventions + HAL associated checking toolOn Chip Bus VCI-close STBus; AMBA
MethodologyDevelopment flow Synopsys based QuartetOCB support STBus, AMBA Platform kitsSystem Level SL model deliverables (TLM, BCA…)Verification dynamic, formal, H/W-S/W, integration
InfrastructureDesign Data Managt., Bug tracking Products from Synchronicity, RationalIP Quality (IP=Product) IPScreen Certification; LibYield Maturity trackingIP Procurement IP On Line catalog; Procurement, Exchange procedures
IP certificationUSB2.0 example
If you pass OK, you have the logo
ST USB2 PHY
1/19/2005 55®
1/19/2005 56®
NortelNortelNokiaNokia
AlcatelAlcatelSynopsysSynopsys
VirageVirage
IP IP ExternalExternalProviderProvider
ARMARM
IP IP DeliverDeliver
yy
Corporate IP Catalog Project
IP IP InternalInternalProviderProvider
IP IP InternalInternalDeliveryDelivery
1/19/2005 57®
Multi-Site Collaborative Design(Synchronicity-based)
Memory RandomLogic
CPU PCI
Graphics
Multiple Designers Multiple Designers Multiple Locations Multiple Locations
One ChipOne Chip
1/19/2005 58®
Network on ChipIn 45 nm technologies, the intra-chip propagation delay is predicted between six and ten clock cycles Wire issues increase reaching up to 20% of the project design effortsLatency • the intra-chip propagation delay• Processor clock cycle times / memory access times• Processors / Co-processor• Within five years, the large majority of SoCs will run on heterogeneous
embedded processors. Today simple Set-Top Box SoC host 6 different processors
On-chip traffic management between integrated functions and the associated methodology will be the forthcoming challengesFocus and complexity are moving • FROM GATES TO WIRES
1/19/2005 59®
Flexible and Reconfigurable SoC
1/19/2005 60®
SOC flexibility and Configurable LogicSOC emerging problems• rising cost of masks• shorter market windows• need to quickly adapt to new customer requirementsneed of• level of programmability• flexibility on-chip
Set of solutions to be provided to our designers• Laser-fuses• Electrically programmable fuses or anti-fuses • Embedded OTP• Embedded ROM• Embedded SRAM with external ROM/Flash• Natural, low-density, embedded Flash in standard CMOS, up to Kbits• Embedded FPGAs• Embedded mask-programmable sea-of-gates.
1/19/2005 61®
Reconfigurable IC with eASIC
0.13um HCMOS9 technologyIncludes 12 eASICores (300kraw gates)Single-mask configuration
Customization: 110k gates + 28kb DP-RAM • Area penalty: 3 - 3.5 x• Speed penalty: 2 - 2.5 x
12eASICores(300k raw
gates)
ARM9468k/4k
SRAM
SRA
M200k gates SC
PLL
1/19/2005 62®
eFPGA Reconfigurable SoC’s
0.18um Working SiliconSpeed: 110 MHzAverage Power @50MHz: 160mWPaper at ISSCC 2003
0.18um Working SiliconSpeed: 180 MHzAverage Power @ 50MHz: 140mW-> in use in 0.13um
1/19/2005 63®
The competition
1/19/2005 64®
TAIWANUn pays exemplaire :
•Ampleur et variété des formes d’aides
•Ampleur des résultats
9 secteurs prioritaires (plan de 1991)
· Semi-conducteur· Communication· Aérospatial· Informatique· Précision machines et automatisation· Advanced Materials Industry· Chimie et Pharmaceutique· Médical · Contrôle et traitement de la pollution
R&D
Fiscalité
Industrialisation
1/19/2005 65®
TAIWAN
Luchu
52 fabs
10 fabs
TSMC, UMC, VANGUARD, VISHAY, …
300 mm
Fab universitaire/Labo public
Centre de R&D
Parcs technologiques
En 2002, Taïwanassure près de 10% de la production mondiale
100 000 emplois
21 millions d’habitants
ITRI/ERSO
Source : SEMI
1/19/2005 66®
Impact sur les entreprises taïwanaises
Dépenses de R&D des grandes entreprises du semi-conducteur de Taiwan en 2000 :
TSMC: 164 M$ (3% du CA) , UMC: 180 M$ (5% du CA), soit moins de la moitié de ce qu'elles devraient être pour être en mesure avec leur chiffre d'affaires et leur activité.
Mécanismes fiscaux pour déduire les investissements :
Double déduction des investissementsFiscalité réduite (exemption de 5 ans d’impôts sur les bénéfices)
Alliances pour développements de procédés < 0.1 µm
ST
IBM
SONY
AMD
AGERETexasInst.
INTEL
TOSHIBATSMC
MOTOROLAPHILIPS
MATSUSHITA
MITSUBISHI FUJITSU
NEC HITACHI
SHARP SANYO
ROHMOKI
NANYA
MOSEL
WINBOND
SAMSUNG
HYNIX
MICRON
Chartered
LSI Logic
Accord R&D
Crolles
East Fishkill
Hsinchu
Séoul
Tsukuba
Singapour
Portland
ELPIDA
UMCINFINEON
1/19/2005 67®
1/19/2005 68®
Monsieur le President Jacques Chiracinaugure l’alliance Crolles2
1/19/2005 69®
Monsieur le President Jacques Chiracinaugure l’alliance Crolles2
1/19/2005 70®
ConclusionsAlliance and co-competition.Large portofolio of Technologies adapted to different SoC market needs: low power, highspeed, RF, analog, imagers…Strong interactions between:• Wafer Manufacturing• Technology Development• IP design• CAD Development
Low-power:• Leakage is here to stay!!!• Design techniques / Process variants can help• More tools needed to predict/implement power solutions at all
abstraction levels
1/19/2005 71®
ConclusionsCo-competition
Low-power:• Leakage is here to stay!!!• Design techniques / Process variants can help• More tools needed to predict/implement power
solutions at all abstraction levels