managing performance and efficiency of a processor
DESCRIPTION
Managing Performance and Efficiency of a Processor. Aditi Shinde Vishwani D. Agrawal Department of Electrical and Computer Engineering Auburn University, Auburn, AL 36849. 10000. 1000. Rocket. Nozzle. 100. Nuclear. Power Density (W/cm 2 ). Reactor. 8086. 10. 4004. P6. 8008. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/1.jpg)
Managing Performance and Efficiency of a Processor
Aditi ShindeVishwani D. Agrawal
Department of Electrical and Computer Engineering
Auburn University, Auburn, AL 36849
March 11, 2013 SSST 2013
![Page 2: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/2.jpg)
400480088080
8085
8086
286386486
Pentium®P6
1
10
100
1000
10000
1970 1980 1990 2000 2010Year
Pow
er D
ensi
ty (
W/c
m2 )
Hot Plate
NuclearReactor
RocketNozzle
Sun’sSurface
Source: Patrick P. Gelsinger , Keynote, ISSCC, Feb. 2001
MOTIVATION: THE POWER PROBLEM!
March 11, 2013 SSST 2013 22
![Page 3: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/3.jpg)
CONTRIBUTION Propose a new cycle efficiency rating for
energy similar to the maximum clock frequency rating for time.
Analogy: Cycle efficiency is similar to miles per
gallon (mpg). Maximum clock frequency is similar to miles
per hour (mph)
![Page 4: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/4.jpg)
PERFORMANCE IN TIME Performance is measured with respect
to a program.
Performance = TimeExecution
1
D. A. Patterson and J. L. Hennessy, Computer Organization & Design, the Hardware/Software Interface, Fourth Edition, San Francisco, California: Morgan Kaufman Publishers, Inc., 2008.
March 11, 2013 SSST 2013 44
![Page 5: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/5.jpg)
PERFORMANCE IN ENERGY (EFFICIENCY)
Efficiency is also measured with respect to a program.
Efficiency nconsumptio Power
ePerformanc
D. A. Patterson and J. L. Hennessy, Computer Organization & Design, the Hardware/Software Interface, Fourth Edition, San Francisco, California: Morgan Kaufman Publishers, Inc., 2008.
March 11, 2013 SSST 2013 55
dissipatedEnergy 1
![Page 6: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/6.jpg)
TWO PERFORMANCES
Time performance
Energy efficiency
D. A. Patterson and J. L. Hennessy, Computer Organization & Design, the Hardware/Software Interface, Fourth Edition, San Francisco, California: Morgan Kaufman Publishers, Inc., 2008.
March 11, 2013 SSST 2013 66
dissipatedEnergy 1
Time Execution1
![Page 7: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/7.jpg)
TIME PERFORMANCE OF A PROCESSOR Speed of a processor is measured in
cycles per second or clock frequency (f).
Execution time of a program using C clock cycles = C/f
Time performance = f/C
![Page 8: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/8.jpg)
ENERGY PERFORMANCE OF A PROCESSOR Efficiency of a processor may be
measured in cycles per joule or cycle efficiency (η).
Energy dissipated by a program using C clock cycles = C/η
Energy performance = η/C
![Page 9: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/9.jpg)
CHARACTERIZING DEVICE TECHNOLOGY Consider 90nm CMOS technology. Use predictive technology model (PTM). Example circuit: Eight-bit ripple carry adder. Nominal voltage = 1.2 volts. Simulated for varying operating conditions (VDD =
100mV through 1.2V) using Spice: With random vectors for energy per cycle (EPC = 1/η). With critical path vectors for clock period (1/f).
Reference: W. Zhao and Y. Cao, “New Generation of Predictive Technology Model for Sub-45nm Early Design Exploration,“ IEEE Trans. Electron Devices, vol. 53, no. 11, pp. 2816–2823, 2006.
![Page 10: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/10.jpg)
ENERGY PER CYCLE OF 8-BIT ADDER
K. Kim, “Ultra Low Power CMOS Design,” PhD Dissertation, Auburn University, Dept. of ECE, Auburn, Alabama, May 2011.
March 11, 2013 SSST 2013 1010
![Page 11: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/11.jpg)
CLOCK CYCLE TIME OF 8-BIT ADDER
K. Kim, “Ultra Low Power CMOS Design,” PhD Dissertation, Auburn University, Dept. of ECE, Auburn, Alabama, May 2011.
March 11, 2013 SSST 2013 1111
![Page 12: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/12.jpg)
PENTIUM M PROCESSOR Published data: H. Hanson, K. Rajamani, S. Keckler, F.
Rawson, S. Ghiasi, J. Rubio, “Thermal Response to DVFS: Analysis with an Intel Pentium M,” Proc. International Symp. Low Power Electronics and Design, 2007, pp. 219-224.
VDD = 1.2V Maximum clock rate = 1.8GHz Critical path delay, td = 1/1.8GHz = 555.56ps Power consumption = 120W Energy per cycle, EPC = 120/(1.8GHz) = 66.67nJ
![Page 13: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/13.jpg)
CYCLE EFFICIENCY AND FREQUENCY FOR PENTIUM M
March 11, 2013 SSST 2013 1313
![Page 14: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/14.jpg)
EXAMPLE For a program that executes in 1.8 billion
clock cycles.Voltage VDD
Frequency f
MHz
Cycle Efficiency,
η
Execution Time second
Total Energy
Consumed
Powerf/η
1.2 V1800
megacycles/s
15 megacycles/jo
ule1.0 120 Joules 120W
0.6 V277
megacycles/s
70 megacycles/jo
ule6.5 25 Joules 39.6W
200 mV54.5
megacycles/s
660 megacycles/jo
ule33 2.72 Joules 0.083W
March 11, 2013 SSST 2013 1414
![Page 15: Managing Performance and Efficiency of a Processor](https://reader036.vdocuments.mx/reader036/viewer/2022062521/56816643550346895dd9ba3a/html5/thumbnails/15.jpg)
CONCLUSION New performance rating: cycle efficiency η has
the unit cycles per joule. Clock frequency f in cycles per second is a
similar rating with respect to time. Similarity to other popular ratings:
η → mpg f → mph Power = f/η
Two ratings allow effective time and energy management of an electronic system.
March 11, 2013 SSST 2013 1515