Upload
lamdang
View
212
Download
0
Embed Size (px)
Citation preview
© 2012 Altera Corporation—Public
The Coming Silicon Convergence:Programmability is the Answer
Ty GaribayVP, Embedded Processing
© 2012 Altera Corporation—Public
Altera at a Glance
2
Founded in Silicon Valley, California in 1983
Industry’s first reprogrammable logic semiconductors
$2.06 billion in 2011 sales
Over 2,900 employees
Leading supplier of FPGAs, ASICs, and CPLDs
© 2012 Altera Corporation—Public
Altera’s Embedded Processing COE
Altera launched the Austin Technology Center in Q2, 2011 Focused on building a strong team
to develop processor-based systems and software
3
Altera finds its Austin homeCalif. chipmaker to employ hundredsPremium content from Austin Business Journal - by Cody Lyon , ABJ Staff Date: Friday, June 10, 2011, 5:00am CDT - Last Modified: Thursday, June 9, 2011, 2:55pm CDT
California chipmaker Altera Corp. signed a major lease in Southwest Austin so it can establish a research and development center that will employ hundreds,
according to Transwestern.
© 2012 Altera Corporation—Public
The Dilemma: Flexibility vs. Efficiency
4
Source: “High-performance Energy-Efficient Reconfigurable Accelerator Circuits for the Sub-45nm Era” July 2011by Ram K. Krishnamurthy, Circuits Research Labs, Intel Corp.
MO
PS/m
W
Programmable Processing
© 2012 Altera Corporation—Public
The Range of IC Options Today
5
Std-cell/Custom Multiple customers Buy & program SW
µP DSP
Std-cell/Custom Multiple customers Buy & program SW Optimized for DSP
C
Software ProgrammabilityµP: C-code DSP: C-code or Models
W W
µPC
DSP FPGA
Programmable Fabric Multiple customers Buy & program HW
Hardware ProgrammabilityRTL code (VHDL, Verilog)
FPGA
W
N
W Supports wide range of applications and unit volumes
Only for narrow /single applicationswith high unit volumes
SoftwareInstruction-based C-code programmingor behavioral modelingHigher abstraction, less silicon efficiency
RTL-based detailed design“Wired speed” efficiency
IP
ASIC
Std-cell One customer Provide design,
buy devices
ASSP
Std-Cell Multiple customers Buy complete
product
Limited or No Software ProgrammabilityEmbedded µP, Embedded DSP: C-code or Models
N N
DSP
May containMay containembeddedembedded
µµP,DSPP,DSPµP
IPIPIPIP
DSP µP
Hardware
© 2012 Altera Corporation—Public
Optimization Drives “Best Chip For The Job”
6
FPGA?DSP?ASSP?ASIC?
DSP?ASSP?
FPGA?µP?ASSP?
Intel®Atom™ processorE600C series
Intel®ATOMTM
E600 Processor
Series
AlteraFPGA
Internet Protocol Surveillance Camera
ImageSensor Image
Processing
IP Block 1 Analytics
VideoEncoding
IP Block 3
FPGA
DDR-SDRAM
EthernetPHY
IP Block 2
DDR-SDRAM
The UMTS Infrastructure
Chipset
© 2012 Altera Corporation—Public
Silicon Convergence
7
Std-cell/Custom Multiple customers Buy & program SW
µP DSP
Std-cell/Custom Multiple customers Buy & program SW Optimized for DSP
C
Software ProgrammabilityµP: C-code DSP: C-code or Models
W W
µPC
DSP FPGA
Programmable Fabric Multiple customers Buy & program HW
Hardware ProgrammabilityRTL code (VHDL, Verilog)
FPGA
W
N
W Supports wide range of applications and unit volumes
Only for narrow /single applicationswith high unit volumes
SoftwareInstruction-based C-code programmingor behavioral modelingHigher abstraction, less silicon efficiency
RTL-based detailed design“Wired speed” efficiency
IP
ASIC
Std-cell One customer Provide design,
buy devices
ASSP
Std-Cell Multiple customers Buy complete
product
Limited or No Software ProgrammabilityEmbedded µP, Embedded DSP: C-code or Models
N N
DSP
May containMay containembeddedembedded
µµP,DSPP,DSPµP
IPIPIPIP
DSP µPNeed for IP Speed& Efficiency
Need for Greater ROI
Hardware
© 2012 Altera Corporation—Public
ASIC
Std-cell One customer Provide design,
buy devices
ASSP
Std-Cell Multiple customers Buy complete
product
Limited or No Software ProgrammabilityEmbedded µP, Embedded DSP: C-code or Models
N N
DSP
May containMay containembeddedembedded
µµP,DSPP,DSPµP
IPIPIPIP
DSP µP
Silicon Convergence
8
Std-cell/Custom Multiple customers Buy & program SW
µP DSP
Std-cell/Custom Multiple customers Buy & program SW Optimized for DSP
C
Software ProgrammabilityµP: C-code DSP: C-code or Models
W W
µPC
DSP FPGA
µP, DSP, application-specific IP & programmable fabric Multiple customers Buy & program HW
and SW
Software & Hardware Programmability
C-code, Models, RTL code
W
Need for IP Speed& Efficiency
Need for Greater ROI
IP
µP DSP
“Mixed System Fabric”
µP + DSP + application-specific IP + programmable fabric
IP
FPGA
© 2012 Altera Corporation—Public
Augmenting Fine-Grained Fabric with Coarse-Grained Programmable Functions in FPGAs
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
IPI/ORAMLOGIC
© 2012 Altera Corporation—Public
Growth of Processors in FPGAs
10
0
20,000
40,000
60,000
80,000
100,000
120,000
Without CPU With CPU
Source: Gartner September 2010
50%Des
ign
Star
ts
© 2012 Altera Corporation—Public11
SoC FPGA
Glue Logic HeterogeneousCapabilities
Stratix I130nm process
High Integration/ Bandwidth
Stratix IV40nm process
Hardened Subsystems
Stratix V28nm process
Cortex-A9 MPCore
SoC FPGA28nm process
Flex 60003µ process
FPGA Architectural Progression
1990’s >50K LEs
Logic
2010 & beyond >1M LEs
RAM
2000’s >500K LEs
Logic RAM
LogicXCVR
DSP
PLL
ALM
RAMPPT
PLLExt Mem Int
FPGAs have evolved well beyond logic fabric!
XCVR
DSP
PLL PLLExt Mem Int
FPDSP
2 X8 CG V
R
HardµPMem Cntl
ALM
PPT
EHB
PCS
© 2012 Altera Corporation—Public
“More Moore” Projections
Sources: ITRS 2010, ITRS 2011, Altera projections beyond 2026
12
Year 2012 2014 2017 2020 2023 2026 2029 2032
Node 20nm 14nm 10nm 7nm 5nm 3.5nm 2.5nm 1.8nm
# FETs per die (B) 8 14 28 56 113 222 453 887
M1 1/2 pitch (nm) 32 24 16.9 11.9 8.4 6 4.2 3
Lgate (nm) 22 18 14 10.6 8.1 5.9 4.2 3
* Normalized to 20nm
© 2012 Altera Corporation—Public
ADC / DAC
Optical
ASSP
3D Integration Technology Opportunities
13
Memory
ASIC
M-core CPU
XCVR + FPGA
3D In
tegr
atio
n Po
ssib
ilitie
s
XCVR + FPGA
XCVR + FPGA
© 2012 Altera Corporation—Public
Programmable Platforms in 2012
Moore’s law has enabled a range of high density programmable platforms
14
Fine‐GrainedMassivelyParallel HeterogeneousArrays
FPGAsDSPsCPUs
Single Cores Coarse‐GrainedMassivelyParallelProcessorArrays
Multi‐CoresCoarse‐GrainedCPUs and DSPs
Multi-Cores Many-Core Arrays
CONVERGENCE
© 2012 Altera Corporation—Public
Mixed System Fabric
The Best of All Worlds
15
Wide ApplicationScope
Wired SpeedWired Efficiency
•Support for legacy code•Familiar C-code
methodology
•Support for high volume applications
• Application-specific IP
• Off-the-shelf availability• Support for changing
standards / reconfigurable systems
DSP•Optimized for DSP
operations
µP ASIC /ASSP
Domain of FPGA Suppliers
IPIP
FPGA
© 2012 Altera Corporation—Public
It’s More Than the Silicon
16
Embedded Initiative Qsys System Integration Tool
Software Programming Support
Video andImage
ProcessingOTN Memory
ControllersFloating
PointOpen standard for
cross-platform design (µP, DSP, GPU)
DSPdesign
Cortex-M1MP32 ColdFire V1
Atom E6XXC
Cortex A9
© 2012 Altera Corporation—Public
Providing the Best of Two Worlds
17
SoC FPGA
Dual Core ARM Cortex-A9 MPCore Processor
Hard Memory
ControllerPeripherals
ARM Processor System28-nm FPGA
ARM + Altera = SoC FPGAs
© 2012 Altera Corporation—Public
SoC FPGA
Processor Dual ARM Cortex-A9
SDRAM Controller, Peripherals Other Hard IP
Serial protocols, memory interfaces
FPGA programmable fabric Multiple density options
Programming model: C/C++ for ARM Common operating systems APIs for hardware accelerators
developed in HDL (Verilog, VHDL, System Verilog), or C/C++ by using high-level-synthesis
OpenCL
* Integrated DMA logic
ARM Cortex-A9NEON / FPU
L1 Cache
ARM Cortex-A9NEON / FPU
L1 Cache
L2 CacheL2 Cache
USBOTG *USB
OTG *
WD(x2)WD(x2)
DMADMA
FPGAFPGA
Multiport DDR SDRAM Controller
Multiport DDR SDRAM Controller
Ethernet
(x2) *
Ethernet
(x2) *
JTAG Debug / Trace
JTAG Debug / Trace
QSPI & NANDFlash Controller *
QSPI & NANDFlash Controller *
ARM Cortex-A9NEON / FPU
L1 Cache
ARM Cortex-A9NEON / FPU
L1 Cache
SD / MMCSD / MMC
I2C(x2)I2C(x2)
CAN(x2)CAN(x2)
GPIOGPIO
UART(x2)
UART(x2)
SPISPI
PCIeHIP
PCIeHIP
Multi-port DDR SDRAM Controller (optional)
Multi-port DDR SDRAM Controller (optional)
Timer(x4)
Timer(x4)
SOC 2FPGASOC 2FPGA
FPGA2 SOCFPGA2 SOC
FPGAConfigFPGAConfig
© 2012 Altera Corporation—Public
System-Level Benefits of SoC FPGA
19
Before
FPGAFPGA
Dev
ices
Dev
ices Memor
yMemor
y
CPUCPU
Dev
ices
Dev
ices Memor
yMemor
y
After
FPGA+ CPUFPGA+ CPUD
evic
esD
evic
es Memory
Memory
Increased system performance 4,000 DMIPS for under 1.8W Up to 1,600 GMACS, 300 GFLOPS DSP >125 Gbps processor to FPGA interconnect Cache coherent hardware accelerators
Reduced power consumption Up to 30% power savings vs. 2-chip solution
Reduced board size Up to 55% form factor reduction As few as two power rails
Reduced system costs Lower component cost Reduction in PCB complexity and cost
Less routing with fewer layers
HighLow
© 2012 Altera Corporation—Public
Summary: Enabling Convergence
Effective integration FPGA programmable fabric, standard cell ASIC /
ASSP, HardCopy ASIC, application-specific IP, DSP, µP cores, analog transceivers, memory
Flexibility Hardware and software programmability Supports wide range of applications
Mass innovation for developers Quartus II, Qsys – enable hardware engineering
productivity OpenCL , Advanced DSP Builder – enable
software engineering productivity
20
Mixed System Fabric
© 2012 Altera Corporation—Public
What Does This All Mean for DFT?
More of the same, of course: Bigger chips More devices Same amount of test time (or less) Same number of pins (or less)
New challenges: 3D IC testing?
Who owns what, when? Higher level redundancy
FPGA Tiles Processing units
What is the boundary between DFT and DV for highly programmable products?
21
© 2012 Altera Corporation—Public ALTERA, ARRIA, CYCLONE, HARDCOPY, MAX, MEGACORE, NIOS, QUARTUS and STRATIX words and logos are trademarks of Altera Corporation and registered in the United States and are trademarks or registered trademarks in other countries.
Thank You
22
© 2012 Altera Corporation—Public
23