UC Regents Fall 2009 © UCBCS 250 L5: Power and Energy2009-9-10John Wawrzynek and Krste Asanovic! with John LazzaroCS 250 VLSI System DesignLecture 5 – Power and Energywww-inst.eecs.berkeley.edu/~cs250/TA: Yunsup Lee 1UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyEnergy and PerformanceSad fact: Computers turn electrical energy into heat. Computation is a byproduct.Air or water carries heat away, or chip melts.2UC Regents Fall 2009 © UCBCS 250 L5: Power and Energy+1V-1 Ohm Resistor1A0.24 Calories per SecondHeats 1 gram of water 0.24 degree CThis is how electric tea pots work ...1 Joule of Heat Energy per Second1 Watt20 W rating: Maximum power the package is able to transfer to the air. Exceed rating and resistor burns.3UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyCooling an iPod nano ...Like resistor on last slide, 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 ...4UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyPowering 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 :-).5UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyNotebooks ... now most of the PC market.Performance: Must be “close enough” to desktop performance ... most people no longer use a desktop.Heat: No longer “laptops” -- top may get “warm”, bottom “hot”. Quiet fans OK.Size and Weight. Ideal: paper notebook.1 in8.9 in12.8 in2006 Apple MacBook -- 5.2 lbs6UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyBattery: 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 more energy than iPod nano battery. And iPod lets you listen to music for 14 hours!7UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyThe CPU is only part of power budget!T.J. Watson Research Center© 2004, 2005 IBM Corporation6 Pradip Bose| Hot Chips 2005 Tutorial August 14, 2005Current Generation Laptop Power Pie15%4%5%1%8%26%1%3%8%29%CPU HDDPower Supply WirelessLCD LCD BacklightOptical Drive MemoryGraphics Rest of the system52%3%3%1%4%13%1%4%4%15%Idle PowerMax Power Workload(IBM Thinkpad R40)Data courtesy Mahesri et al., U of Illinois, 20042004-era notebook running a full workload.“Amdahl’s Law for Power”If our CPU took no power at all to run, that would only double battery life!T.J. Watson Research Center© 2004, 2005 IBM Corporation6 Pradip Bose| Hot Chips 2005 Tutorial August 14, 2005Current Generation Laptop Power Pie15%4%5%1%8%26%1%3%8%29%CPU HDDPower Supply WirelessLCD LCD BacklightOptical Drive MemoryGraphics Rest of the system52%3%3%1%4%13%1%4%4%15%Idle PowerMax Power Workload(IBM Thinkpad R40)Data courtesy Mahesri et al., U of Illinois, 2004CPULCD Backlight“other”LCDGPU8UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyServers: Total Cost of Ownership (TCO)Rack Stats•Weight: 1500 Lbs•Power: 98 Amps•Fans: 340 (2”) + 2 (8”)•Wire: 0.25 miles•Assembly and wiring time: 60 man-hoursMachine 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.9UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyProcessors and Energy10UC Regents Fall 2009 © UCBCS 250 L5: Power and Energy Switching Energy: Fundamental Physics!"#$%&'())* ++,!-.)'/ 012-)34$5$%&67&1'-)!"#$%&'(#)*(+,%-$*".(/01 2+.$0#$031 4546%,"#$3VddC 12C VddE1->0= 212C VddE0->1= 2VddEvery 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 limits the clock speed ...(3) Fewer circuits. But more transistors can do more work.(4) Reduce C per node. One reason why we scale processes.11UC Regents Fall 2009 © UCBCS 250 L5: Power and EnergyScaling switching energy per gate ...6665nm300mmDual CoreScaling: Scaling: The Fundamental Cost DriverThe Fundamental Cost Driver90nm300mm130nm200mm180nm200mm250nm200mm350nm200mmOROR==Twice theTwice thecircuitry in thecircuitry in thesame spacesame space(architectural(architecturalinnovation)innovation)The sameThe samecircuitry in halfcircuitry in halfthe spacethe space(cost reduction)(cost reduction)Half the die sizeHalf the die sizefor the samefor the samecapability thancapability thanin the priorin the priorprocessprocessIC process scaling (“Moore’s Law”)From: “Facing the Hot Chips Challenge Again”, Bill Holt, Intel, presented at Hot Chips 17, 2005.1616Process Advances Still Scale PowerProcess Advances Still Scale Powerbut the rate has slowed and collaboration is requiredbut the rate has slowed and collaboration is required..35!m35!m..25!m25!m..18!m18!m..13!m13!m90nm90nm65nm65nm45nm45nm32nm32nmCVCV22 Scaling ScalingDue to reducing V and C (length and width of Cs decrease, but plate distance gets smaller).Recent slope more shallow because V is being scaled less aggressively.12UC Regents Fall 2009 © UCBCS 250 L5: Power and Energy0V = = VddT.J. Watson Research Center© 2004, 2005 IBM Corporation29 Pradip Bose| Hot Chips 2005 Tutorial August 14, 2005Power-related issues in chip designTemperatureCapacitive (Dynamic) PowerStatic (Leakage) PowerMinimum Voltage20 cyclesDi/Dt (Vdd/Gnd Bounce)Voltage (V)Current (A) VOUT CL ISub VIN IGate Vin VoutCLVddSecond Factor: Leakage CurrentsEven 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 Ioffwe like, but a lower Ioff also results in a lower Ion, and thus a lower maximum clock speed.1717Leakage becomes