Dennard scaling

Last updated

Dennard scaling, also known as MOSFET scaling, is a scaling law based on a 1974 paper co-authored by Robert H. Dennard, after whom it is named. [1] Originally formulated for MOSFETs, it states, roughly, that as transistors get smaller, their power density stays constant, so that the power use stays in proportion with area; both voltage and current scale (downward) with length. [2] [3]

Robert H. Dennard American engineer and inventor

Robert Dennard is an American electrical engineer and inventor.

MOSFET Transistor used for amplifying or switching electronic signals.

The metal–oxide–semiconductor field-effect transistor (MOSFET, MOS-FET, or MOS FET), also known as the metal–oxide–silicon transistor (MOS transistor, or MOS), is a type of field-effect transistor that is fabricated by the controlled oxidation of a semiconductor, typically silicon. It has an insulated gate, whose voltage determines the conductivity of the device. This ability to change conductivity with the amount of applied voltage can be used for amplifying or switching electronic signals. The MOSFET is the basic building block of modern electronics. Since its invention by Mohamed M. Atalla and Dawon Kahng at Bell Labs in November 1959, the MOSFET has become the most widely manufactured device in history, with an estimated total of 13 sextillion (1.3 × 1022) MOS transistors manufactured between 1960 and 2018.

Power density is the amount of power per unit volume.

Contents

Derivation

Dennard observes that transistor dimensions are scaled by 30% (0.7x) every technology generation, thus reducing their area by 50%. This reduces the delay by 30% (0.7x) and therefore increases operating frequency by about 40% (1.4x). Finally, to keep the electric field constant, voltage is reduced by 30%, reducing energy by 65% and power (at 1.4x frequency) by 50%. [note 1] Therefore, in every technology generation the transistor density doubles, the circuit becomes 40% faster, and power consumption (with twice the number of transistors) stays the same. [4]

Relation with Moore's law and computing performance

Moore's law says that the number of transistors doubles about every two years. Combined with Dennard scaling, this means that performance per watt grows at this same rate, doubling about every two years. This trend is referred to as Koomey's law. The rate of doubling was originally suggested by Koomey to be 1.57 years [5] (somewhat faster than the doubling period of Moore's law), but more recent estimates suggest this is slowing. [6]

Moores law Heuristic law stating that the number of transistors on a integrated circuit doubles every two years

Moore's law is the observation that the number of transistors in a dense integrated circuit doubles about every two years. The observation is named after Gordon Moore, the co-founder of Fairchild Semiconductor and CEO of Intel, whose 1965 paper described a doubling every year in the number of components per integrated circuit, and projected this rate of growth would continue for at least another decade. In 1975, looking forward to the next decade, he revised the forecast to doubling every two years, a compound annual growth rate (CAGR) of 41.4%.

In computing, performance per watt is a measure of the energy efficiency of a particular computer architecture or computer hardware. Literally, it measures the rate of computation that can be delivered by a computer for every watt of power consumed. This rate is typically measured by performance on the LINPACK benchmark when trying to compare between computing systems.

Koomeys law

Koomey's law describes a trend in the history of computing hardware: for about a half-century, the number of computations per joule of energy dissipated doubled about every 1.57 years. Professor Jonathan Koomey described the trend in a 2010 paper in which he wrote that "at a fixed computing load, the amount of battery you need will fall by a factor of two every year and a half."

Breakdown of Dennard scaling around 2006

The dynamic (switching) power consumption of CMOS circuits is proportional to frequency. [7] Historically, the transistor power reduction afforded by Dennard scaling allowed manufacturers to drastically raise clock frequencies from one generation to the next without significantly increasing overall circuit power consumption.

Since around 2005–2007 Dennard scaling appears to have broken down. As of 2016, transistor counts in integrated circuits are still growing, but the resulting improvements in performance are more gradual than the speed-ups resulting from significant frequency increases. [2] [8] The primary reason cited for the breakdown is that at small sizes, current leakage poses greater challenges and also causes the chip to heat up, which creates a threat of thermal runaway and therefore further increases energy costs. [2] [8]

Thermal runaway situation where an increase in temperature changes the conditions in a way that causes a further increase in temperature, often leading to a destructive result

Thermal runaway occurs in situations where an increase in temperature changes the conditions in a way that causes a further increase in temperature, often leading to a destructive result. It is a kind of uncontrolled positive feedback.

The breakdown of Dennard scaling and resulting inability to increase clock frequencies significantly has caused most CPU manufacturers to focus on multicore processors as an alternative way to improve performance. An increased core count benefits many (though by no means all) workloads, but the increase in active switching elements from having multiple cores still results in increased overall power consumption and thus worsens CPU power dissipation issues. [9] [10] The end result is that only some fraction of an integrated circuit can actually be active at any given point in time without violating power constraints. The remaining (inactive) area is referred to as dark silicon.

Central processing unit power dissipation or CPU power dissipation is the process in which central processing units (CPUs) consume electrical energy, and dissipate this energy in the form of heat due to the resistance in the electronic circuits.

In the electronics industry, dark silicon is the amount of circuitry of an integrated circuit that cannot be powered-on at the nominal operating voltage for a given thermal design power (TDP) constraint. This is a challenge in the era of nanometer semiconductor nodes, where transistor scaling and voltage scaling are no longer in line with each other, resulting in the failure of Dennard scaling. This discontinuation of Dennard scaling has led to sharp increases in power densities that hamper powering-on all the transistors simultaneously at the nominal voltage, while keeping the chip temperature in the safe operating range. According to recent studies, researchers from different groups have projected that, at 8 nm technology nodes, the amount of Dark Silicon may reach up to 50–80% depending upon the processor architecture, cooling technology, and application workloads. Dark Silicon may be unavoidable even in server workloads with abundance of inherent client request-level parallelism.

See also

Notes

  1. Active power = CV2f

Related Research Articles

Processor design is the design engineering task of creating a processor, a key component of computer hardware. It is a subfield of computer engineering and electronics engineering (fabrication). The design process involves choosing an instruction set and a certain execution paradigm and results in a microarchitecture, which might be described in e.g. VHDL or Verilog. For microprocessor design, this description is then manufactured employing some of the various semiconductor device fabrication processes, resulting in a die which is bonded onto a chip carrier. This chip carrier is then soldered onto, or inserted into a socket on, a printed circuit board (PCB).

Integrated circuit electronic circuit manufactured by lithography; set of electronic circuits on one small flat piece (or "chip") of semiconductor material, normally silicon

An integrated circuit or monolithic integrated circuit is a set of electronic circuits on one small flat piece of semiconductor material that is normally silicon. The integration of large numbers of tiny MOS transistors into a small chip results in circuits that are orders of magnitude smaller, faster, and less expensive than those constructed of discrete electronic components. The IC's mass production capability, reliability, and building-block approach to circuit design has ensured the rapid adoption of standardized ICs in place of designs using discrete transistors. ICs are now used in virtually all electronic equipment and have revolutionized the world of electronics. Computers, mobile phones, and other digital home appliances are now inextricable parts of the structure of modern societies, made possible by the small size and low cost of ICs.

N-type metal-oxide-semiconductor logic uses n-type MOSFETs to implement logic gates and other digital circuits. These nMOS transistors operate by creating an inversion layer in a p-type transistor body. This inversion layer, called the n-channel, can conduct electrons between n-type "source" and "drain" terminals. The n-channel is created by applying voltage to the third terminal, called the gate. Like other MOSFETs, nMOS transistors have four modes of operation: cut-off, triode, saturation, and velocity saturation.

CMOS Technology for constructing integrated circuits

Complementary metal–oxide–semiconductor (CMOS), also known as complementary-symmetry metal–oxide–semiconductor (COS-MOS), is a type of MOSFET fabrication process that uses complementary and symmetrical pairs of p-type and n-type MOSFETs for logic functions. CMOS technology is used for constructing integrated circuits (ICs), including microprocessors, microcontrollers, memory chips, and other digital logic circuits. CMOS technology is also used for analog circuits such as image sensors, data converters, RF circuits, and highly integrated transceivers for many types of communication.

Static random-access memory Semiconductor memory

Static random-access memory is a type of semiconductor random-access memory (RAM) that uses bistable latching circuitry (flip-flop) to store each bit. SRAM exhibits data remanence, but it is still volatile in the conventional sense that data is eventually lost when the memory is not powered.

BiCMOS is an evolved semiconductor technology that integrates two formerly separate semiconductor technologies, those of the bipolar junction transistor and the CMOS transistor, in a single integrated circuit device.

The CPU core voltage (VCORE) is the power supply voltage supplied to the CPU, GPU, or other device containing a processing core. The amount of power a CPU uses, and thus the amount of heat it dissipates, is the product of this voltage and the current it draws. In modern CPUs, which are CMOS circuits, the current is almost proportional to the clock speed, the CPU drawing almost no current between clock cycles.

Power optimization is the use of electronic design automation tools to optimize (reduce) the power consumption of a digital design, such as that of an integrated circuit, while preserving the functionality.

Electronics industry global industry

The electronics industry, especially meaning consumer electronics, emerged in the 20th century and has now become a global industry worth billions of dollars. Contemporary society uses all manner of electronic devices built in automated or semi-automated factories operated by the industry. Products are assembled from integrated circuits, principally by photolithography of printed circuit boards.

In computer architecture, frequency scaling is the technique of increasing a processor's frequency so as to enhance the performance of the system containing the processor in question. Frequency ramping was the dominant force in commodity processor performance increases from the mid-1980s until roughly the end of 2004.

Edholm's law, proposed by and named after Phil Edholm, refers to the observation that the three categories of telecommunication, namely wireless (mobile), nomadic and wired networks (fixed), are in lockstep and gradually converging. Edholm's law also holds that data rates for these telecommunications categories increase on similar exponential curves, with the slower rates trailing the faster ones by a predictable time lag. Edholm's law predicts that the required data rates double every 18 months, which has proven to be true since the 1970s. The trend is evident in the cases of Internet, cellular (mobile), wireless LAN and wireless personal area networks.

Dynamic frequency scaling is a technique in computer architecture whereby the frequency of a microprocessor can be automatically adjusted "on the fly" depending on the actual needs, to conserve power and reduce the amount of heat generated by the chip. Dynamic frequency scaling helps preserve battery on mobile devices and decrease cooling cost and noise on quiet computing settings, or can be useful as a security measure for overheated systems. Dynamic frequency scaling is used in all ranges of computing systems, ranging from mobile systems to data centers to reduce the power at the times of low workload.

Dynamic voltage scaling is a power management technique in computer architecture, where the voltage used in a component is increased or decreased, depending upon circumstances. Dynamic voltage scaling to increase voltage is known as overvolting; dynamic voltage scaling to decrease voltage is known as undervolting. Undervolting is done in order to conserve power, particularly in laptops and other mobile devices, where energy comes from a battery and thus is limited, or in rare cases, to increase reliability. Overvolting is done in order to increase computer performance.

The term die shrink refers to the scaling of metal-oxide-semiconductor (MOS) devices. The act of shrinking a die is to create a somewhat identical circuit using a more advanced fabrication process, usually involving an advance of lithographic nodes. This reduces overall costs for a chip company, as the absence of major architectural changes to the processor lowers research and development costs, while at the same time allowing more processor dies to be manufactured on the same piece of silicon wafer, resulting in less cost per product sold.

Low-power electronics are electronics, such as notebook processors, that have been designed to use less electric power.

Pollack's Rule states that microprocessor "performance increase due to microarchitecture advances is roughly proportional to [the] square root of [the] increase in complexity". This contrasts with power consumption increase, which is roughly linearly proportional to the increase in complexity. Complexity in this context means processor logic, i.e. its area.

References

  1. Dennard, Robert H.; Gaensslen, Fritz; Yu, Hwa-Nien; Rideout, Leo; Bassous, Ernest; LeBlanc, Andre (October 1974). "Design of ion-implanted MOSFET's with very small physical dimensions" (PDF). IEEE Journal of Solid-State Circuits. SC-9 (5).
  2. 1 2 3 McMenamin, Adrian (April 15, 2013). "The end of Dennard scaling" . Retrieved January 23, 2014.
  3. Streetman, Ben G.; Banerjee, Sanjay Kumar (2016). Solid state electronic devices. Boston: Pearson. p. 341. ISBN   978-1-292-06055-2. OCLC   908999844.
  4. Borkar, Shekhar; Chien, Andrew A. (May 2011). "The Future of Microprocessors". Communications of the ACM. 54 (5): 67. doi:10.1145/1941487.1941507 . Retrieved 2011-11-27.
  5. Greene, Katie (September 12, 2011). "A New and Improved Moore's Law: Under "Koomey's law," it's efficiency, not power, that doubles every year and a half". Technology Review . Retrieved January 23, 2014.Italic or bold markup not allowed in: |publisher= (help)
  6. http://www.koomey.com/post/153838038643
  7. "CMOS Power Consumption and CPD Calculation" (PDF). Texas Instruments. June 1997. Retrieved March 9, 2016.
  8. 1 2 Bohr, Mark (January 2007). "A 30 Year Retrospective on Dennard's MOSFET Scaling Paper" (PDF). Solid-State Circuits Society. Retrieved January 23, 2014.
  9. Esmaeilzedah, Hadi; Blem, Emily; St. Amant, Renee; Sankaralingam, Kartikeyan; Burger, Doug (2012). "Dark Silicon and the end of multicore scaling" (PDF).
  10. Hruska, Joel (February 1, 2012). "The death of CPU scaling: From one core to many — and why we're still stuck". ExtremeTech . Retrieved January 23, 2014.