1. CS 150 - Spring 2007 – Lec #28 – Power - 1 Power zMotivation for design constraints of power consumption zPower metrics zPower consumption analysis in CMOS zHow can a logic designer control power?

2. CS 150 - Spring 2007 – Lec #28 – Power - 2 Automobiles 700 Million Telephones 4 Billion Electronic Chips 60 Billion X-Internet “X-Internet” Beyond the PC Forrester Research, May 2001 Revised 2007 500 Million 1.5 Billion Internet Computers Internet Users Today’s Internet

3. CS 150 - Spring 2007 – Lec #28 – Power - 3 “X-Internet” Beyond the PC Forrester Research, May 2001 Millions Year X Internet PC Internet

4. CS 150 - Spring 2007 – Lec #28 – Power - 4 Cell Phones Siemens SL45i Ericsson T68 zPhone w/voice command, voice dialing, intelligent text for short msgs zMP3 player + headset, digital voice recorder z“Mobile Internet” with a built-in WAP Browser zJava-enabled, over the air programmable zBluetooth + GPRS zEnhanced displays + embedded cameras

5. CS 150 - Spring 2007 – Lec #28 – Power - 5 Shape of Things zPhone + Messenger + PDA Combinations yE.g., Blackberry 5810 Wireless Phone/Handheld xIntegration of PDA + Telephone xPLUS Gateway to Internet and Enterprise applications x1900 MHz GSM/GPRS (Euroversion at 900 Mhz) xSMS Messaging, Internet access xQWERTY Keyboard, 20 line display xJAVA applications capable x8 MB flash + 1 MB SRAM

6. CS 150 - Spring 2007 – Lec #28 – Power - 6 Shape of Things to Come zDanger “Hiptop” yFull-featured mobile phone w/Internet Access yEmail + attachments/instant messaging + PIM yDigital camera accessory yEnd-to-end integration of voice + data apps yMedia-rich UI for graphics + sound yLarge screen + QWERTY keyboard yData nav: keyboard or push wheel yAffordable (under $200) yMIDI synthesizer for quality sound yMulti-tasking of user actions yCustomizable ring tones and alerts to personalize hiptop experience

7. CS 150 - Spring 2007 – Lec #28 – Power - 7 Important (Wireless) Technology Trends “Spectral Efficiency”: More bits/m 3 Rapidly declining system cost Rapidly increasing transistor density

8. CS 150 - Spring 2007 – Lec #28 – Power - 8 In the Physical World: Sensor Devices

9. CS 150 - Spring 2007 – Lec #28 – Power - 9 Important (Wireless) Technology Trends Speed-Distance-Cost Tradeoffs Rapid Growth: Machine-to- Machine Devices (mostly sensors)

10. CS 150 - Spring 2007 – Lec #28 – Power - 10 Why Worry About Power? zPortable devices: yHandhelds, laptops, phones, MP3 players, cameras, … all need to run for extended periods on small batteries without recharging yDevices that need regular recharging or large heavy batteries will lose out to those that don’t. zPower consumption important even in “tethered” devices ySystem cost tracks power consumption: xPower supplies, distribution, heat removal yPower conservation, environmental concerns zIn 10 years, have gone from minimal consideration of power consumption to (designing with power consumption as a primary design constraint!

11. CS 150 - Spring 2007 – Lec #28 – Power - 11 zPower supply provides energy for charging and discharging wires and transistor gates. The energy supplied is stored & then dissipated as heat. zIf a differential amount of charge dq is given a differential increase in energy dw, the potential of the charge is increased by: zBy definition of current: Power: Rate of work being done wrt time Rate of energy being used total energy Units: Watts = Joules/seconds A very practical formulation! If we would like to know total energy Basics

12. CS 150 - Spring 2007 – Lec #28 – Power - 12 Basics zWarning! In everyday language, the term “power” is used incorrectly in place of “energy” zPower is not energy zPower is not something you can run out of zPower can not be lost or used up zIt is not a thing, it is merely a rate zIt can not be put into a battery any more than velocity can be put in the gas tank of a car

13. CS 150 - Spring 2007 – Lec #28 – Power - 13 + 1V - 1 Ohm Resistor 1A 0.24 Calories per Second Heats 1 gram of water 0.24 degree C This is how electric tea pots work... 1 Joule of Heat Energy per Second 1 Watt 20 W rating: Maximum power the package is able to transfer to the air. Exceed rating and resistor burns.

14. CS 150 - Spring 2007 – Lec #28 – Power - 14 Cooling an iPod nano... Like a resistor, iPod relies on passive transfer of heat from case to the air Why? Users don’t want fans in their pocket... To stay “cool to the touch” via passive cooling, power budget of 5 W If iPod nano used 5W all the time, its battery would last 15 minutes...

15. CS 150 - Spring 2007 – Lec #28 – Power - 15 Powering an iPod nano (2005 edition) Battery has 1.2 W-hour rating: Can supply 1.2 W of power for 1 hour 1.2 W / 5 W = 15 minutes Real specs for iPod nano : 14 hours for music, 4 hours for slide shows 85 mW for music 300 mW for slides More W-hours require bigger battery and thus bigger “form factor” -- it wouldn’t be “nano” anymore!

16. CS 150 - Spring 2007 – Lec #28 – Power - 16 0.55 ounces 12 hour battery life $79.00 1 GB

17. CS 150 - Spring 2007 – Lec #28 – Power - 17 12 hour battery life 24 hour battery life for audio 5 hour battery life for photos 20 hour battery life for audio, 6.5 hours for movies (80GB version) Up from 14 hours for 2005 iPod nano Up from 4 hours for 2005 iPod nano Thinner than 2005 iPod nano

18. CS 150 - Spring 2007 – Lec #28 – Power - 18 Notebooks... now most of the PC market Performance: Must be “close enough” to desktop performance... many people no longer own a desktop Heat: No longer “laptops” -- top may get “warm”, bottom “hot”. Quiet fans OK Size and Weight: Ideal: paper notebook 1 in 8.9 in 12.8 in Apple MacBook -- Weighs 5.2 lbs

19. CS 150 - Spring 2007 – Lec #28 – Power - 19 Battery: Set by size and weight limits... Almost full 1 inch depth. Width and height set by available space, weight. Battery rating: 55 W-hour At 2.3 GHz, Intel Core Duo CPU consumes 31 W running a heavy load - under 2 hours battery life! And, just for CPU! At 1 GHz, CPU consumes 13 Watts. “Energy saver” option uses this mode... 46x energy than iPod nano. iPod lets you listen to music for 14 hours!

20. CS 150 - Spring 2007 – Lec #28 – Power - 20 Battery Technology zBattery technology has developed slowly zLi-Ion and NiMh still the dominate technologies zBatteries still contribute significantly to the weight of mobile devices Toshiba Portege 3110 laptop - 20% Handspring PDA - 10% Nokia 61xx - 33%

21. CS 150 - Spring 2007 – Lec #28 – Power - 21 55 W-hour battery stores the energy of 1/2 a stick of dynamite. If battery short-circuits, catastrophe is possible...

22. CS 150 - Spring 2007 – Lec #28 – Power - 22 CPU Only Part of Power Budget 2004-era notebook running a full workload. If our CPU took no power at all to run, that would only double battery life! CPU LCD Backlight “other” LCD GPU

23. CS 150 - Spring 2007 – Lec #28 – Power - 23 Servers: Total Cost of Ownership (TCO) Machine rooms are expensive … removing heat dictates how many servers to put in a machine room. Electric bill adds up! Powering the servers + powering the air conditioners is a big part of TCO Reliability: running computers hot makes them fail more often

24. CS 150 - Spring 2007 – Lec #28 – Power - 24 Thermal Image of Typical Cluster Rack Rack Switch M. K. Patterson, A. Pratt, P. Kumar, “From UPS to Silicon: an end-to-end evaluation of datacenter efficiency”, Intel Corporation

25. CS 150 - Spring 2007 – Lec #28 – Power - 25 How Do We Measure and Compare Power Consumption? zOne popular metric for microprocessors is: MIPS/watt yMIPS, millions of instructions per second xTypical modern value? yWatt, standard unit of power consumption xTypical value for modern processor? yMIPS/watt reflects tradeoff between performance and power yIncreasing performance requires increasing power yProblem with “MIPS/watt” xMIPS/watt values are typically not independent of MIPS Techniques exist to achieve very high MIPS/watt values, but at very low absolute MIPS (used in watches) xMetric only relevant for comparing processors with a similar performance yOne solution, MIPS 2 /watt. Puts more weight on performance

26. CS 150 - Spring 2007 – Lec #28 – Power - 26 Metrics zHow does MIPS/watt relate to energy? zAverage power consumption = energy / time yMIPS/watt = instructions/sec / joules/sec = instructions/joule yEquivalent metric (reciprocal) is energy per operation (E/op) zE/op is more general - applies to more that processors yalso, usually more relevant, as batteries life is limited by total energy draw. yThis metric gives us a measure to use to compare two alternative implementations of a particular function.

27. CS 150 - Spring 2007 – Lec #28 – Power - 27 Power in CMOS Switching Energy: energy used to switch a node Energy suppliedEnergy dissipatedEnergy stored Calculate energy dissipated in pullup: An equal amount of energy is dissipated on pulldown

28. CS 150 - Spring 2007 – Lec #28 – Power - 28 Switching Power zGate power consumption: yAssume a gate output is switching its output at a rate of: zChip/circuit power consumption: activity factorclock rate Therefore: number of nodes (or gates) (probability of switching on any particular clock period)

29. CS 150 - Spring 2007 – Lec #28 – Power - 29 Other Sources of Energy Consumption z“Short Circuit” Current: 10-20% of total chip power ~1nWatt/gate few mWatts/chip Transistor drain regions “leak” charge to substrate. zJunction Diode Leakage :

30. CS 150 - Spring 2007 – Lec #28 – Power - 30 Other Sources of Energy Consumption z Consumption caused by “DC leakage current” (Ids leakage): z This source of power consumption is becoming increasing significant as process technology scales down z For 90nm chips around 10-20% of total power consumption Estimates put it at up to 50% for 65nm Transistor s/d conductance never turns off all the way Low voltage processes much worse

31. CS 150 - Spring 2007 – Lec #28 – Power - 31 Controlling Energy Consumption: What Control Do You Have as a Designer? zLargest contributing component to CMOS power consumption is switching power: z Factors influencing power consumption: yn: total number of nodes in circuit   : activity factor (probability of each node switching) yf: clock frequency (does this effect energy consumption?) yVdd: power supply voltage z What control do you have over each factor? z How does each effect the total Energy? Our design projects do not optimize for power consumption

32. CS 150 - Spring 2007 – Lec #28 – Power - 32 Scaling Switching Energy per Gate Moore’s Law at work … From: “Facing the Hot Chips Challenge Again”, Bill Holt, Intel, presented at Hot Chips 17, 2005. Due to reduced V and C (length and width of Cs decrease, but plate distance gets smaller) Recent slope reduced because V is scaled less aggressively

33. CS 150 - Spring 2007 – Lec #28 – Power - 33 Device Engineers Trade Speed and Power From: Silicon Device Scaling to the Sub-10-nm Regime Meikei Ieong, 1* Bruce Doris, 2 Jakub Kedzierski, 1 Ken Rim, 1 Min Yang 1 We can reduce leakage (P standby ) by raising V t We can increase speed by raising V dd and lowering V t We can reduce CV 2 (P active ) by lowering V dd

34. CS 150 - Spring 2007 – Lec #28 – Power - 34 Customize processes for product types... From: “Facing the Hot Chips Challenge Again”, Bill Holt, Intel, presented at Hot Chips 17, 2005.

35. CS 150 - Spring 2007 – Lec #28 – Power - 35 Intel: Comparing 2 CPU Generations... Clock speed unchanged... Lower Vdd, lower C, but more leakage Design tricks: architecture & circuits Find enough tricks, and you can afford to raise Vdd a little so that you can raise the clock speed!

36. CS 150 - Spring 2007 – Lec #28 – Power - 36 Switching Energy: Fundamental Physics V dd C 1 2 C V dd E 1- >0 = 2 1 2 C V dd E 0- >1 = 2 V dd Every logic transition dissipates energy Strong result: Independent of technology How can we limit switching energy? (1) Slow down clock (fewer transitions). But we like speed... (2) Reduce Vdd. But lowering Vdd lowers the clock speed... (3) Fewer circuits. But more transistors can do more work. (4) Reduce C per node. One reason why we scale processes.

37. CS 150 - Spring 2007 – Lec #28 – Power - 37 0V = Second Factor: Leakage Currents Even when a logic gate isn’t switching, it burns power … Igate: Ideal capacitors have zero DC current. But modern transistor gates are a few atoms thick, and are not ideal. Isub: Even when this nFet is off, it passes an Ioff leakage current. We can engineer any Ioff we like, but a lower Ioff also results in a lower Ion, and thus the lower the clock speed. Intel’s current processor designs, leakage vs switching power A lot of work was done to get a ratio this good... 50/50 is common. Bill Holt, Intel, Hot Chips 17.

38. CS 150 - Spring 2007 – Lec #28 – Power - 38 Engineering “On” Current at 25 nm... I ds VsVs VdVd V g 0.7 = V dd 0.25 ≈ V t I ds 1.2 mA = I on I off = 0 ??? We can increase I on by raising V dd and/or lowering V t.

39. CS 150 - Spring 2007 – Lec #28 – Power - 39 Plot on a “Log” Scale to See “Off” Current I ds VsVs VdVd VgVg I off ≈ 10 nA We can decrease I off by raising V t - but that lowers I on 0.25 ≈ V t 1.2 mA = I on 0.7 = V dd