Click for next page ( 70


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 69
E Dennard Scaling and Implications T he following description was taken from the 2010 National Research Council Computer Science and Telecommunications Board (CSTB) report The Future of Computing Performance: Game Over or Next Level?1 decreases by 1/ because the smaller gates make all the wires shorter and capacitance is proportional to length. Therefore, the power requirements per unit of space on the chip (mm2), or energy per second per mm2, remain constant: "In a classic 1974 paper, reprinted in Appendix D, Power = (number of gates)(CLoad/gate)(Clock Rate)(Vsupply2) Robert Dennard et al. showed that the MOS Power density = NgCloadFclkVdd2 transistor has a set of very convenient scaling Ng = CMOS gates per unit area properties.10 The scaling properties are shown in Cload = capacitive load per CMOS gate Table 3.1, taken from that paper. If all the voltages Fclk = clock frequency in a MOS device are scaled down with the physical Vdd = supply voltage dimensions, the operation of the device scales in a Power density = (2)(1/)()(1/)2 = 1 particularly favorable way. The gates clearly become smaller because linear dimensions are scaled. That That the power density (power requirements per unit scaling also causes gates to become faster with space on the chip, even when each unit space contains lower energy per transition. If all dimensions and many, many more gates) can remain constant across voltages are scaled by the scaling factor ( has generations of CMOS scaling has been a critical property typically been 1.4), after scaling the gates become underlying progress in microprocessors and in ICs in (1/)2 their previous size, and 2 more gates can be general. In every technology generation, ICs can double placed on a chip of roughly the same size and cost as in complexity and increase in clock frequency while before. The delay of the gate also decreases by 1/, consuming the same power and not increasing in cost. and, most importantly, the energy dissipated each Given that description of classic CMOS scaling, one time the gate switches decreases by (1/)3. To would expect the power of processors to have remained understand why the energy drops so rapidly, note constant since the CMOS transition, but this has not been that the energy that the gate dissipates is the case. During the late 1980s and early 1990s, supply proportional to the energy that is stored at the output voltages were stuck at 5 V for system reasons. So power of the gate. That energy is proportional to a quantity density would have been expected to increase as called capacitance11 and the square of the supply technology scaled from 2 mm to 0.5 mm. However, until voltage. The load capacitance of the wiring recently supply voltage has scaled with technology, but power densities continued to increase." 1 NRC, 2011, The Future of Computing Performance: Game Over or Next Level?, Washington, D.C.: The National Academies Press (available online at http://www.nap.edu/catalog.php?record_ id=12980). 69

OCR for page 69