Low Power Design Prof. Dr. J. Henkel CES - Chair for Embedded - - PowerPoint PPT Presentation

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Low Power Design

• Prof. Dr. J. Henkel

CES - Chair for Embedded Systems KIT, Germany

• I. Introduction and Energy/Power Sources

2

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Overview

Reason for Low Power Design: motivation Specific need for low power in embedded systems: examples Battery issues (re-chargeable batteries) Power/energy sources

3

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Why design for low power/energy?

n

Portable Systems

ä

Notebooks, smartphones, tablets, cameras, etc.

32% of PC market, and growing ä

Battery-driven - long battery life crucial

ä

System cost, weight limited by batteries

40W, 10 hrs @ 20-35 W- hr/pound = 7-20 pounds

Slow growth in battery technology n

Must reduce energy drain from batteries

n

Thermal Considerations

ä

10 oC increase in operating temperature => component failure rate doubles

ä

Packaging: ceramic vs. plastic

ä

Cooling requirements

n

Increasing levels of integration / clock frequencies make the problem worse

ä

10cm2, 500 MHz => 315Watts

n

Reliability Issues

ä

Electro-migration

ä

IR drops on supply lines

ä

Inductive effects

n

Tied to peak/average power consumption

n

Environmental Concerns

ä

EPA estimate: 80% of office equipment electricity is used in computers

ä

“Energy Star” program to recognize power efficient PCs

ä

Power management standard for desktops and laptops

n

Drive towards “Green PC”

LOW POWER (Src: A. Raghunathan, NEC)

4

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

(Src: F. Pollack, Intel

5

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power consumption: motivation

Pentium Crusoe

Pentium 4 Crusoe Processor

(source: www.transmeta.com)

6

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Temp max: 125.0°C Temp min: 81.5 °C Thermal variation: 43.5 Spatial thermal gradient: ~1.81 °C/mm

resources LUTs 2000 FFs 2000 DSPs 12 BRAMs DCM 1 Frequency 550 MHz Area 13% of chip Properties of the tested region

Src: Henkel, Amrouch, Ebi

Xilinx Virtex-5 FX100t, package ff1136, speed -3 (65nm)

Exploring different FPGA chips (The worst stress-scenario)

Dennard Scaling, Circuit Heat

Folgerung?

“Circuit heat generation is the main limiting factor for scaling of device speed and switch circuit density”

By Jeff Welser, Director SRC Nanoelectronics Research Initiative, IBM, Opening Keynote Address ICCAD 2007

Classical scaling

Device count S2 Device frequency S Device power (cap) 1/S Device power (Vdd) 1/S2 Utilization 1

Leakage limited scaling

Device count S2 Device frequency S Device power (cap) 1/S Device power (Vdd) ~1 Utilization 1/S2

(Src: “Dennard Scaling”)

8

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

65nm 45nm 32nm 22nm 2 Cores @ < 2GHz

Power Wall: Is Multi-Core Scaling Promising?

4 Core @ >= 2GHz 8 Core @ > 2GHz 16 Core @ >= 2GHz 8 Dark Cores 22nm 16 Core @ >= 4GHz 12 Dark Cores

Assumption: Scaling Factor S=2 Tradeoff between #Cores and Frequency

9

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power Wall

• H. Esmaeilzadeh, E. Blem, R. St. Amant, K. Sankaralingam, D. Burger, “Dark

Silicon and the End of Multicore Scaling”, in International Symposium on Computer Architecture (ISCA), 2011

• The Dark Silicon Problem emerged due to the

utilization wall as a result of

• Memory wall
• Parallelism wall
• Power Wall => Chips cannot be driven with

a power greater than their power budgets given as TDP (thermal design power)

• Power dissipation, peak power, and power

density are the ultimate limiting factors, thus determining the amount of Dark Silicon

• Power is more expensive than area
• Regardless of available parallelism, chip
• rganization & topology, multicore scaling is

power limited

• @22 nm: Dark Silicon≈20%-50%
• @8 nm: Dark Silicon > 50%-70%
• Dark Silicon is must be kept powered-OFF
• J. Allred, S. Roy, K. Chakraborty, “Designing for Dark Silicon: A

Methodological Perspective on Energy Efficient Systems”, in ISLPED, 2012

10

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power Wall: Trends

10

• H. Esmaeilzadeh, E. Blem, R. St. Amant, K. Sankaralingam, D. Burger, “Dark Silicon and the End of Multicore Scaling”, in International Symposium on Computer Architecture (ISCA), 2011

Scaling Limits when Dark Silicon Dominates

Even if there is unlimited Parallelism, The Speedup is limited by the Power Constraint

11

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power and Heat in FPGAs

45 48 51 54 57 60 63 1 67 133 199 265 331 397 463 529 595 661 727 793 859 925 991 1057 1123 1189 1255 1321 1387 1453 1519 1585 1651 1717 1783 1849

Temperature [°C]

Time [sec] Core1 Core2

Thermal Camera Virtex-5 FPGA without packaging

Activity migration between two cores at the interval of 154 MCycle

Scr: Amrouch, Ebi, Henkel

Problem: vertical heat flow

Only one layer directly interfaces with the heat sink Heat needs to dissipate through multiple layers The heat sink is located on top of the chip Hot cores distant to the heat sink dissipate their heat through other layers Silicon has a low thermal conductivity! 150 W/(m*K) (Silicon) 401 W/(m*K) (Copper)

Temperature in 3D

13

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Energy vs. Programmability

Large (100X – 1000X) gap in energy efficiency between fully programmable and fully custom implementations

Ample scope for tradeoffs

Source: Rabaey et. al., IEEE Computer, July 2000

14

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power consumption by processing type

Technology Operations/Watt [MOPS/mW] 1 0.1 0.01 0.13µ Ambient Intelligence 0.07µ DSP-ASIPs µPs 10 0.25µ 0.5µ 1.0µ poor design generation techniques

(Src:[Marw03])

15

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Relationship between Power and Energy

∫

= dt P E

t P E

Energy: 1 Ws = 1 VAs = 1 Joule = 1 Nm

16

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power vs. Energy

Minimizing the power consumption is important for

the design of the power supply the design of voltage regulators the dimensioning of interconnect short term cooling

Minimizing the energy consumption:

Limited availability of energy (mobile systems, try to maximize the amount of computation that can be accomplished with a given amount of energy) through:

limited battery capacities (only slowly improving) very high costs of energy (solar panels, in space)

cooling

high costs limited space

dependability

long lifetimes, low temperatures

(Src:[Marw03])

17

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

HW Power Consumption

Behavior level Register-transfer level Logic level Transistor level Power analysis iteration times seconds - minutes minutes - hours hours - days Decreasing design iteration times

High-level synthesis, RTL optimizations Architecture-level power analysis Logic synthesis Logic-level power analysis Transistor-level/ Layout synthesis Transistor-level power analysis Power models for macroblocks, control logic Power models for gates, cells, nets

1 Power Cap Switching _Power Leakage/Static Power + … = ( . _ +

L dd

2 C V A f . . .

2

)

(src: A. Raghunathan, NEC)

18

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power/Energy-Conscious Applications

• Some examples-

19

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Example 1: E-Textiles

• Smart Shirt -

Source: [Marc03]

20

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Example 2: Medical Diagnostics

(source: Jan Madsen DTU)

21

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Example 3: Sensor Networks

Manufacturing plants & Power distribution

• Improve reliability, operating efficiency

Health care

• Unwired operating

rooms

• Early detection of

cardiac attacks Energy-efficient buildings

• $55 B / year
• pportunity in the

US

Disaster Prevention & Emergency Response

Traffic control

• Reduce commute time

by 15 min => $15B/year in California alone “Smart” environments

• Homes, Offices, Schools, …
• Convenience, Productivity, Security

(source: A. Raghunathan, NEC)

22

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

More examples

Banking & Money transfer smart cards, … Consumer cell phone, MP3 player, PDA, … Clothing electronic textiles Environment sensor networks Healthcare hearings aids, pace maker, … Telecom Systems satellite, …

…

23

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power/Energy Sources

24

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Problem of battery capacity in comparison

1 10 100 1000 10000 100000 1000000 10000000

Algorithmic Complexity (Shannon’s Law) Processor Performance (Moore’s Law) Battery Capacity 1G 2G 3G

(src: A. Cuomo, ST Micro, Stockholm, Sept.8, 2004)

25

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Primary/Secondary Batteries

Primary batteries

+ availability + no re-charging required + often higher density compared to secondary batteries (later)

• cannot be re-charged (replacement of cartridge etc. instead)
• user always needs to carry replacement batteries
• form-factor often unfavorable (not flat as desired)

Secondary batteries

Ni-Cd (nickel-cadmium), NiMH (nickel-metal-hydride, Lithium-Ion, Lithium-polymer + can be re-charged

• lesser energy density compared to primary (it is constantly

increasing but “plateauing” i.e. cannot be significantly improved any more. Lithium-Ion: has increased around 8-10% in the last 10 years (every year)

26

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Metrics: Energy density:

• gravimetric, volumetric -

1994 200 wh/kg 1996 1998 2000 2002 2004 2006 150 100 50 Sanyo Toshiba 1994 500 wh/l 1996 1998 2000 2002 2004 2006 300 100 200 400 Sanyo Toshiba

shown Lithium-Ion technology

 Gravimetric: Wh/kg -> Watt * hours / kg  Volumetric: Wh/l -> Watt * hours / liter

(src: [Blo04])

27

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Metrics: cost

• secondary batteries -

 Average cost of Lithium-Ion technology (currently 2005) is ~0.5 USD/Wh  Will decrease further but curve is predicted to flatten in the near future

0,5 1 1,5 $0,55 $0,55 $0,45 $1,27 $ per Wh NiCd average price NiMH average price Li-ion cyl. average price Li-ion prism. average price

1 2 3 4 5 6 7 8 9 10 US $/cell 1999 2000 2002 2001 2003 2004 2005 2006 Li-ion (average) Li-ion Cylindrical Li-ion Prismatic Li-ion Polym r e

(src: [Blo04])

28

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Improving gravimetric, volumetric characteristic

Improving the gravimetric and/or volumetric by embedding part of the battery into the final device and obtain the

• xidant, for example, from the environment. This way, it

does not have to be carried by the user Example: metal-air system. Basic idea: positive electrode is the ambient air. Metal-air system: Reaction:  Negative electrode:

 Zn + 4 OH- -> Zn(OH-)4

2- + 2 e- E0 = -1.266V

 Positive electrode:

 ½ O2 + H2O +2 e- -> 2 OH- E0 = 0.401V

 Allover reaction:

 Zn + H2O + ½ O2 -> ZnO E0 = 1.667V

29

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Metal-air system (cont’d)

Oxidation reaction of zinc with oxygen produces very high energy density: 1370 Wh/kg (theoretical) Reaction begins by presence of air and continues until zinc has been used up

 - For continuous use only  - No re-charging (then energy density would drop)  + low cost  + simple to use  + environmentally OK (no heavy or noble metals nor hazardous compounds involved)

30

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Fuel Cells

Technologies: Solid oxide fuel cells (SOFC):

Needs 800-850 degrees centigrade

Polymer exchange membrane (PEM)

Reaction positive electrode:

½ O2 + 2 H3O+ + 2e- -> 3 H2O

Reaction negative electrode:

H2 + 2 H2O -> 2H3O+ + 2 e-

Overall reaction:

H2 + ½ O2 -> H2O E0 = 1.229V

31

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Fuel Cell Principle

 Functionality:

• Core parts are two electrodes

separated by an ion-conducting polymeric membrane (electrolyte)

• Fuel (i.e. H2) is transformed on

catalytic sites at the negative electrode and form protons (H+) which cross the membrane and electrons on the other hand which produce a current outside the cell

• electrical energy is obtained when

electrons recombine at the positive electrode with protons (H+) coming from the negative electrode and

• xygen from the air
• chemical reaction results in:

electricity, water heat

• a whole system is shown on the

next page

H*

EME

• +

Heat O (air)

2

H O + O

2 2

Cathode Solid Polymer Elektrolyte Anode H

2

H O + H

2 2

Current collector Bipolar plate

Proton exchange membrane fuel cell principle (PEM)

(src: [Blo04])

32

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

A whole fuel cell system

Whole system contains besides the core (stack): a) electrical, b) thermal, c) and fluidic management systems

Methanol Methanol Reformer Humidi- fication water PEMFC Stack H Loop

2

Water pump Heat exchanger Water condenser exhaust Cooling Loop

DC DC

Air + H O

2

Regulation Compressor Air H

2

Buffer battery

DC DC

• rAC

Electric device

(src: [Blo04])

33

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

About fuel cells

Estimated to be more energy efficient in converting chemical energy to work (via electrical and afterwards mechanical energy conversion) Environmentally clean (byproduct is water) Efficiency of 50% (globally) is claimed to be achievable (incl. peripherals like water/heat/fuel management and fuel storage)

Ex: H2 heating value: 33.3 Wh/g , 600g H2

600g * 33.3 Wh/g * 50% = 10,000 Wh (e.g. 10kW for 1 hour)

However: large-power fuel cells are not likely to be mass- produced before probably 2020 But: miniature fuel cells are on their way …

34

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Miniature fuel cells

Application domain: portable electronic devices (PDA, cell ph, cameras, etc.) Two approaches:

A) “bipolar” technology

Built with bipolar plates forming the fuel cell stack Typically 20-500W Smaller stacks seem not to be competitive with Lithium-Ion batteries www.smartfuelcell.de and many others

B) Various approaches with new concepts e.g. micro fabrication techniques

Typically 0.1-25W substrate (thin-film)-based

35

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Silicon fuel cell

• +

Current Collector Positive electrode Elektrolyte Diffusion layer Catalyst layer Hydrogen or methanol Air H O

2

e- e- H+

O 2 H 2

 Silicon wafer; grown and treated with lithographic techniques  Often less than a centimeter wide  By various companies/institutions like: Neah Power; Integrated Fuel Cell Technologies, French Atomic Energy Commission, Case Western University

(src: [Blo04])

36

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Human-generated power for portable devices

Source: [StaPa04]

37

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Increases in a computing system by component

Improvement multiple since 1990

1990 1992 1994 1996 1998 2000 2002 1 10 100 1000 Year Disk capacity CPU speed Available RAM Wireless transfer speed Battery erargy density

 (Note: different physical units for different components are given)  Battery capacity only grew by 3x since 1990  On contrary: storage size, as an example, grew by 4000x during same time frame  Problem: how can these largely increased system components appropriately fed with electrical energy?

38

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Human power consumption for various activities

Sleeping Lying quietly Sitting Standing at case Conversation Eating a meal Strolling Driving a car Playing the violin or piano Housekeeping Carpentry Hiking, 4 mph Swimming Mountain climbing Long-distance run Sprinting 70 80 100 110 110 110 140 140 140 150 230 350 500 600 900 1400 81 93 116 128 128 128 163 163 163 175 268 407 582 698 1048 1630 Activity Kilocal/hr Watts Human Energy Expenditures for Selected Activities

Source: Derived from D. Morton. Human Locomotion and Body Form. Williams & Wilkens, Baltimore, MD.1952

A span of ~20x !

• However: power may

not be easily harvested

• But even then: for usage

the power/energy stored, converted (DC/DC, impedance, etc)

• For acceptance,

harvesting needs to be completely non-obtrusive

39

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Power/energy from humans

Body Heat 2.4 - 4.8 W (Carnot efficiency) Blood pressure 0.37W (0.93 W) Arm motion 0.33 W (60 W) Footfalls 5.0 - 8.3 W (67 W) Exhalation 0.40 W (1.0 W) Breathing band 0.42 W (0.83 W) Finger motion 0.76 - 2.1 mW (6.9 - 19 mW)

40

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Human power sources

Body heat:

(T_body-T_ambient)/T_body = 310K-293K)/310K = 5.5%

• > little efficient

From breath

Principle: uses diff. in from breath pressure and atmospheric pressure -> only 2%

From blood pressure Capturing energy from vibrations, motion etc.

41

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Human power sources (cont’d)

Power from typing

 Ex: 50g key pressure, depress by 0.5cm  (0.05kg/key-stroke) / (9.8m/s2) * 0.005m * (7.5 key-strokes / sec) =  = 19 mW -> too less to power a whole portable system; plus, user is not continuously typing  Idea: keyboard can at least announce its character to the rest of the system through own energy

Inertial micro systems

 Used for hundred of years in watches

Electrical version (next slide)

 Functionality:

 the mass winds a spring  when enough mech. (spring) energy is accumulated, a micro generator is driven at 15,000 rpm (rotations per minute)  yields 6mA and 16V for 50ms

42

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Self-winding electric watch

43

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Human power sources: walking

Walking (68kg human, 3.5mph) costs ~324Watt of power

Most of this power is used to move legs

Power through the fall of the heel:

68kg * (9.8m/sec2) * 0.05m * (2 steps/sec) = 67W This power cannot simply converted in electrical power w/o significant intrusion Converting to electrical power: e.g. via piezoelectric device (e.g. Quartz)

Mechanically stressed axis during fabrication Piezoelectric Material

Electrostatic Poling Direction (across electrodes)

44

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Other power/energy sources

Energy Source Power/Energy Density Batteries (Zinc-Air, primary) 1050-1560 mWh/cm3 Batteries (Li, rechargeable) 300 mWh/cm3 Solar (outdoors) 15 mW/cm2 (direct sun) 1 mW/cm2 (24 hour avg) Solar (indoors) 0.006 mW/cm2 (office desk) 0.57mW/cm2 (<60W desk lamp) Vibrations 0.01-0.1 mW/cm3 Acoustic (noise) 3 e-6 mW/cm2 @ 75dB 9.6 e-4 mW/cm2 @ 100dB Miniature Fuel cells 0.1-500W Nuclear Reaction 80 mW/cm3, 1 e+6 mWh/cm3

(Src.(modified): A. Raghunathan, NEC)

45

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Vibration -> Electricity

[Src: Hande, Dallas]

46

• Prof. Jörg Henkel, Low Power Design, SS2014 ces.itec.kit.edu

Reference and sources

[Piguet04] Ch. Piguet (Ed.), “Low Power Electronics Design”, CRC Press, ISBN 0-8493-1941-2, 2004. {Marc03], Marculescu, D.; Marculescu, R.; Park, S.; Jayaraman, S.; “Ready to ware”, Spectrum, IEEE ,Volume: 40 , Issue: 10 , Oct. 2003, Pages:28 – 32. [StaPa04] Th. Starner, J. Paradiso, “Human-generated power for mobile electronics”, appeared in “Low Power Electronics Design”, CRC Press, 2004. [Blo04] D. Bloch, “Miniature fuel cells for portable applications”, appeared in “Low Power Electronics Design”, CRC Press, 2004. [Marw03] P. Marwedel, “Embedded System Design”, Kluwer, 2003. [Raghunathan] A. Raghunathan, Tutorial on low power design, held at various CAD conferences