How Much Power Is Consumed By A Load Drawing 5 A

Ability Consumption

Communication Mechanisms for Smart Objects

Jean-Philippe Vasseur , Adam Dunkels , in Interconnecting Smart Objects with IP, 2022

12.3.five Power Consumption

The ability consumption of IEEE 802.xv.iv is determined past the current draw of the electrical circuits that implement the concrete communication layer, and past the amount of time during which the radio is turned on. As shown in Chapter 11, there are several ways a radio can be switched off while maintaining communication abilities. Effigy 12.10 shows the power consumption of the electrical circuitry of the CC2420 IEEE 802.15.iv transceiver, as reported past the CC2420 data sheet. It shows that the idle power consumption is significantly lower than both the listen and the transmit power consumption. In the idle fashion, however, the transceiver is not able to receive any data. The power consumption in the transmit modes is lower than the power consumption in listen way. The ability consumption of the transmit mode depends on the output power, which is configurable via software on a per-packet basis.

Figure 12.10. The power consumption of the CC2420 IEEE 802.fifteen.4 radio transceiver.

Read total chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780123751652000120

Evaluation of Performance and Power Consumption of Storage Systems

Jalil Boukhobza , Pierre Olivier , in Wink Memory Integration, 2022

iii.3.2 Exploration of the power consumption of storage systems based on NAND flash memory through measurements

Power consumption metrics measurements tin exist performed for an exploratory purpose, in guild to understand and to study the power consumption profiles of the storage system or of ane of its components. Besides, ability consumption measurements can as well be performed during a report in order to validate the precision of a power consumption model, or to mensurate the efficacy of a proposition of a new storage organisation or an optimization that targets free energy saving. Here we concentrate on the studies that deal with the exploration of storage systems' power consumption based on flash retentiveness as their master subject area.

Every bit wink memory is a relatively recent engineering, numerous studies illustrate sequences of power consumption measurements that target this kind of memory. Their goal is to characterize the power consumption profiles of these systems. Moreover, some studies further develop this work by analyzing the measurements in order to identify the elements which have a meaning impact on power consumption and the elements which, on the reverse, take a negligible bear upon. Highlighting these elements is a showtime, essential step in whatever work for optimizing performance or consumption.

The measurements can be performed at wink flake level [GRU 09, MAT 09]. In this example, a specific hardware platform is required where the flash chip tin be inserted and equipped for power consumption measurement. This kind of platform was built by the authors of the 2 cited studies, and it includes a resistor along the power supply rail of the chip. By means of an oscilloscope, the current at the resistor's terminals is measured. The oscilloscope has to provide a data logger function in order to be able to exploit the data at a later time. This method of equipping the power supply line with a measuring device is also employed in many studies that address the ability consumption of SSDs [SEO 08, SHI 10, BJØ 10, YOO 11a], and the comparing between the power consumption of SSDs and hard disk drives [DAV 10, LEE 09b, SCH 10a].

Read total chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9781785481246500049

Rugged autonomous vehicles

R. Zalman , in Rugged Embedded Systems, 2022

2.iii Supply and Power Consumption

Power consumption is a key recurring thematic in automotive electronics. The push of the industry towards reducing the power consumption becomes even more disquisitional with:

–

Aggressive targets for emission reduction and

–

Constant increment of electronic functionality in the modern vehicles.

The emission reduction targets have a multiple influence on the modern ECU architectures. On i side in society to achieve the required emission levels (like particulates, NOx, etc.) there is a strong need for loftier performance computation in the engine control. This computation power, which for mod standards like EURO 6, needs the power of multicore microcontrollers and sophisticated algorithms (the need for the main microcontroller for powertrain applications needs more 1   Gips), comes with respective higher power consumption. On the other side, the need for CO₂ reduction implies a stringent need to lower the power consumption of the vehicle internal systems. Knowing that 100   Westward need in electric supply equates to effectually 0.1   L of fuel increase consumption, the need of electric power reduction is evident. Similarly, the equivalent of ane   1000 CO₂ is around 40   W electrical power need (as well equivalent with around 20   kg weight increase). These targets are extremely of import and cause a huge pressure on the automotive industry, as they are more than and more difficult to reach. Every bit an instance [vi,eight], in the EU, the boilerplate target for CO₂ emission is gradually decreasing from 130 to 95   grand   CO_ii/km in 2022 in the context of a dramatic increment in electronic functionality required by modern cars.

The modern need for more and more functionality covered by ECUs creates likewise a pressure on the power supply budget. With more than than 70–100 ECUs in today's heart-high stop cars (trend continues), each having an boilerplate consumption of around 200   mA, smart power saving strategies are needed.

One important constraint deriving from power consumption in automotive domain is the standby consumption. More and more ECUs demand certain functionality even exterior of the driving cycle. Depending on the original equipment manufacturer (OEM), requirements for standby consumption are varying between max 300 to max 100   μA. This target is difficult to attain due to strict requirements for cyclic wake-up:

–

Periodic wake-upward for analog inputs acquisition (as well watchdogs);

–

Wake-up on external events (e.thousand., switch detection);

–

Very fast consummate wake-up if needed for operating actuators (eastward.m., door unlock);

–

Periodic communication for antitheft units;

–

Blinking LEDs, etc.

The ECUs are highly interconnected in modern cars. Usually CAN, FlexRay, LIN, Well-nigh and most recently, Ethernet are the backbones of these circuitous interconnections. While there is a certain support for global wake features (all ECUs are waked up past specific letters on the omnibus), network architectures in which only some regions of the omnibus are powered downwards are non withal mainstream. Concepts like "fractional networks" and "pretended networks [five]" are addressing these ability requirements allowing certain "degraded" operation modes in which only a subset of functionality is available at a given moment in order to save power, merely the total functionality can exist available depending on the system needs—the network tin can wake up the consummate functionality of ECUs on demand.

Read full chapter

URL:

https://www.sciencedirect.com/scientific discipline/article/pii/B9780128024591000087

Energy Efficient Networking for Data Centers

Casimer DeCusatis , in Handbook of Fiber Optic Data Advice (Fourth Edition), 2022

Power consumption is recognized as a limiting factor in the blueprint of very big supercomputers; optical link engineering can help commencement this problem, enabling the design of significantly larger systems in the future. For instance, at 10  pJ per fleck (or 10   mW/Gbps) the ability consumed by internode communications in a 1 exaflop computer system is about 20   MW. This is equivalent to the entire system power upkeep for a system on this scale, showing the growing importance of controlling interconnect ability consumption if nosotros are to attain exaflop computers. Consider some typical estimates for exaflop compute power associated with data networking. If nosotros use VCSELs over multimode fiber, a one-stage fat tree network topology for an exascale network would cost around $840 million (this topology is probable not technically feasible for other reasons and is included hither equally a comparison point only). A more reasonable option is the three-phase fat tree, which increases the cost to around $2520 one thousand thousand. Nonetheless, if we employ a unlike topology such every bit a 4D torus, we can reduce the network cost to a more reasonable $420 million and 12   MW of power consumption. Higher levels of integration in the optical transceivers, combined with new technologies, are expected to reduce these costs even farther. As the industry moves toward this level of compute density, it will be important to pay increasing attention to the energy efficiency in all aspects of the optical data center network.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780124016736000398

Virtual bus structure-based network-on-chip topologies†

In Networks-On-Chip, 2022

four.five.4 Power consumption analysis

Figure iv.18 shows the ability consumption of the NoC designs. The power consumption results were obtained past executing application traffic in an 8 × viii mesh topology. As the figure indicates, the conventional NoC design has the largest power consumption, and is used as the baseline design. The NOCHI EVC blueprint reduces the ability consumption past an average of about 14%. This is mainly due to a reduction in buffer and crossbar ability consumption in bypassing routers. In a similar way, the proposed VBON pattern can also reduce the power consumption by about xx%. In summary, the proposed VBON design conspicuously outperforms the other two NoC designs beyond all benchmarks not merely in performance but likewise in ability consumption.

Figure four.18. Ability consumption results for application traffic in an 8 × 8 mesh.

Read full affiliate

URL:

https://www.sciencedirect.com/science/article/pii/B9780128009796000044

From Zippo to 1

Sarah L. Harris , David Harris , in Digital Design and Calculator Architecture, 2022

1.8 Power Consumption*

Power consumption is the corporeality of free energy used per unit of measurement time. Power consumption is of great importance in digital systems. The battery life of portable systems such as cell phones and laptop computers is limited by power consumption. For example, a cell phone battery holds near ten watt-hours (W-hr) of free energy, meaning that it could evangelize 1 watt (West) for 10 hours or ii   W for 5 hours, and then forth. For your telephone battery to last a full day, its average consumption should exist under one   W. Laptops typically have l to 100   W-hr batteries and consume under 10   W in normal performance, of which the screen is a large fraction. Power is also significant for systems that are plugged in because electricity costs coin and emissions and because the system will overheat if it draws as well much ability. A desktop calculator consuming 200   W for 8 hours each day would use approximately 600 kilowatt-hours (kW-hr) of electricity per yr. At an average toll of 12 cents and 1 pound of CO_two emission per kW-60 minutes, this is $72 of electricity each year every bit well as 600 pounds of CO_ii emissions.

You'll assistance our planet past making certain your reckoner sleeps when you lot aren't using information technology.

Digital systems describe both dynamic and static power. Dynamic power is the ability used to charge capacitance as signals change between 0 and ane. Static ability is the power used even when signals do non change and the system is idle.

Logic gates and the wires that connect them accept capacitance. The energy fatigued from the power supply to charge a capacitance C to voltage V_DD is CV_DD ². If the system operates at frequency f and the fraction of the cycles on which the capacitor charges and discharges is α (chosen the activity factor), the dynamic ability consumption is

(one.iv) $P_{\mathrm{dynamic}}=\mathrm{\alpha }C{V_{DD}}^{2}f$

Effigy 1.41 illustrates activity factors. Figure i.41(a) shows a clock indicate, which rises and falls once every bicycle and, thus, has an activeness factor of one. The clock flow from one rising edge to the side by side is chosen T _c (the cycle fourth dimension), and is the reciprocal of the clock frequency f. Effigy 1.41(b) shows a data signal that switches once every clock bike. The parallel lines on the timing diagram indicate that the signal might exist high or might exist low; we aren't concerned with its value. The crossover indicates that the signal changes once, early in each clock cycle. Hence, the activity factor is 0.v (rising half the cycles and falling one-half the cycles). Figure 1.41(c) shows a random information signal that switches in half the cycles and remains abiding in the other one-half of the cycles. Therefore, information technology has an activity factor of 0.25. Real digital systems ofttimes accept idle components that are non switching, so an action factor of 0.1 is more typical.

Figure 1.41. Illustration of activeness factors

Electrical systems describe some electric current even when they are idle. When transistors are OFF, they leak a pocket-size corporeality of current. Some circuits, such equally the pseudo-nMOS gate discussed in Section i.7.8, take a path from V_DD to GND through which current flows continuously. The full static electric current, I_DD , is also called the leakage current or the quiescent supply current flowing between V_DD and GND. The static ability consumption is proportional to this static electric current:

(1.5) $P_{\mathrm{static}}=I_{DD}{\mathrm{Five}}_{DD}$

Example 1.23

Power Consumption

A particular cell telephone has an 8   W-hour bombardment and operates at 0.707   5. Suppose that, when information technology is in utilise, the cell telephone operates at 2   GHz. The total capacitance of the circuitry is ten nF (10⁻⁸ Farads), and the activity factor is 0.05. When voice or data are active (ten% of its time in use), it also broadcasts 3   West of power out of its antenna. When the phone is non in apply, the dynamic power drops to almost zero because the point processing is turned off. Simply the telephone also draws 100   mA of quiescent current whether it is in use or non. Decide the battery life of the phone (a) if it is not being used and (b) if information technology is being used continuously.

Solution

The static power is P _static = (0.100   A)(0.707 Five) = 71 milliwatts (mW). (a) If the phone is not being used, this is the only power consumption, and then the battery life is (8  W-hr)/(0.071   Westward) = 113 hours (about five days). (b) If the phone is being used, the dynamic power is P _dynamic = (0.05)(10⁻⁸ F)(0.707   V)^two(2 × 10^nine  Hz) = 0.five   West. The average broadcast power is (three   W)(0.i) = 0.3   W.

Together with the static and broadcast power, the total active ability is 0.5   W + 0.071   W + 0.3   W = 0.871   W, so the battery life is 8   W-hour/0.0871   West = 9.2 hours. This case somewhat oversimplifies the actual performance of a cell telephone, but it illustrates the fundamental ideas of power consumption.

Read total chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780128200643000015

CPUs

Marilyn Wolf , in Computers as Components (Fourth Edition), 2022

iii.7 CPU power consumption

Power consumption is, in some situations, as important as execution time. In this department nosotros study the characteristics of CPUs that influence power consumption and mechanisms provided past CPUs to command how much power they eat.

Energy versus power

First, we demand to distinguish between energy and power. Power is, of course, energy consumption per unit time. Heat generation depends on power consumption. Battery life, on the other hand, almost directly depends on energy consumption. More often than not, we volition use the term ability as autograph for energy and power consumption, distinguishing betwixt them just when necessary.

3.7.1 CMOS power consumption

Power and energy are closely related but push different parts of the design. The energy required for a computation is independent of the speed with which we perform that work. Free energy consumption is closely related to battery life. Power is energy per unit time. In some cases, such as vehicles that run from a generator, we may have limits on the total power consumption of the platform. Only the well-nigh common limitation on power consumption comes from heat generation—more power burned means more heat.

CMOS power characteristics

The high-level ability consumption characteristics of CPUs and other system components are derived from the circuits used to build those components. Today, virtually all digital systems are congenital with CMOS (complementary metal oxide semiconductor) circuitry. The detailed circuit characteristics are best left to a written report of VLSI pattern [Wol08], simply we can identify 2 important mechanisms of power consumption in CMOS:

•

Dynamic: The traditional power consumption mechanism in CMOS circuit is dynamic—the logic uses nigh of its power when it is changing its output value. If the logic's inputs and outputs are non changing, then information technology does not consume dynamic power. This means that we can reduce ability consumption past freezing the logic's inputs.

•

Static: Mod CMOS processes besides consume power statically—the nanometer-calibration transistors used to make billion-transistor chips are subject to losses that are not important in older technologies with larger transistors. The almost important static power consumption mechanism is leakage—the transistor draws current even when it is off. The only way to eliminate leakage current is to remove the power supply.

Dynamic and static ability consumption require very different management methods. Dynamic ability may exist saved by running more than slowly. Decision-making static power requires turning off logic.

As a result, several power-saving strategies are used in CMOS CPUs:

•

CPUs tin be used at reduced voltage levels. For case, reducing the power supply from 1 to 0.9   V causes the power consumption to drib by 1ⁱⁱ/0.ix²  =   one.2.

•

The CPU tin exist operated at a lower clock frequency to reduce power (but non energy) consumption.

•

The CPU may internally disable certain function units that are not required for the currently executing function. This reduces energy consumption.

•

Some CPUs permit parts of the CPU to be totally disconnected from the power supply to eliminate leakage currents.

iii.7.2 Ability management modes

Static versus dynamic power management

CPUs tin can provide two types of power management modes. A static power management mechanism is invoked by the user but does non otherwise depend on CPU activities. An example of a static machinery is a power-down mode intended to save free energy. This way provides a loftier-level fashion to reduce unnecessary power consumption. The manner is typically entered with an pedagogy. If the mode stops the interpretation of instructions, then it conspicuously cannot be exited by execution of another instruction. Power-downwards modes typically end upon receipt of an interrupt or other event. A dynamic power management mechanism takes actions to control power based upon the dynamic activity in the CPU. For example, the CPU may plough off certain sections of the CPU when the instructions being executed do non need them. Awarding Example 3.2 describes the static and dynamic energy efficiency features of a PowerPC fleck.

Application Example 3.ii Free energy Efficiency Features in the PowerPC 603

The PowerPC 603 [Gar94] was designed specifically for low-power performance while retaining high functioning. It typically dissipates two.2   W running at 80   MHz. The architecture provides iii low-ability modes—doze, nap, and sleep—that provide static power direction capabilities for use by the programs and operating system.

The 603 likewise uses a variety of dynamic power direction techniques for power minimization that are performed automatically, without plan intervention. The CPU is a two-consequence, out-of-gild superscalar processor. Information technology uses the dynamic techniques summarized below to reduce ability consumption.

•

An execution unit of measurement that is not existence used can be shut downward.

•

The cache, an 8-KB, ii-manner set-associative cache, was organized into subarrays so that at well-nigh ii out of eight subarrays will exist accessed on any given clock wheel. A multifariousness of circuit techniques were also used in the enshroud to reduce power consumption.

Not all units in the CPU are agile all the time; idling them when they are not being used can save power. The table below shows the percentage of time various units in the 603 were idle for the SPEC integer and floating-signal benchmarks [Gar94].

Unit	Specint92 (% idle)	Specfp92 (% idle)
Data cache	29	28
Instruction enshroud	29	17
Load-store	35	17
Fixed-point	38	76
Floating-point	99	30
System register	89	97

Idle units are turned off automatically past switching off their clocks. Various stages of the pipeline are turned on and off, depending on which stages are necessary at the current time. Measurements comparing the fleck'south ability consumption with and without dynamic ability management evidence that dynamic techniques provide significant power savings.

A power-downwardly fashion provides the opportunity to profoundly reduce power consumption considering information technology will typically be entered for a substantial flow of fourth dimension. However, going into and especially out of a power-down mode is not free—information technology costs both time and energy. The power-down or power-up transition consumes time and energy to properly control the CPU's internal logic. Modernistic pipelined processors crave complex control that must be properly initialized to avoid corrupting data in the pipeline. Starting up the processor must also exist done carefully to avert power surges that could cause the chip to malfunction or even impairment information technology.

The modes of a CPU tin can exist modeled past a power country machine [Ben00]. Each state in the machine represents a dissimilar mode of the machine, and every state is labeled with its average power consumption. The example machine has two states: run mode with power consumption P _run and sleep mode with power consumption P _slumber. Transitions show how the machine can go from land to state; each transition is labeled with the time required to go from the source to the destination state. In a more complex example, it may not be possible to go from a particular land to another detail state—traversing a sequence of states may be necessary.

Application Example three.3 describes the power management modes of the NXP LPC1300.

Awarding Instance 3.3 Power Direction Modes of the ARM Cortex-R5

The NXP LPC1311 [[NXP12],NSP11] is an ARM Cortex-M3 [ARM11]. Information technology provides iv power management modes:

Mode	CPU clock gated?	CPU logic powered?	SRAM powered?	Peripherals powered?
Run	No	Yep	Yeah	Yes
Sleep	Yes	Yep	Yep	Yeah
Deep sleep	Yeah	Aye	Yep	Near analog blocks close downwards (power dip, watchdog remain powered)
Deep power-down	Shut downward	No	No	No

In sleep mode, the peripherals remain active and can crusade an interrupt that returns the system to run fashion. Deep ability-down mode is equivalent to a reset on restart.

Here is a power state machine for the LPC1311 (the manufacturer does not give the time required to enter one of the slumber or reset states).

The deep sleep mode consumes 13% of the power required by the slumber mode but requires 8X longer to return to run way. The deep power-down way uses 0.7% of the power required by deep sleep but takes 10X longer to return to run mode.

3.7.3 Program-level ability management

2 classic power management methods accept been developed, one aimed primarily at dynamic ability consumption and the other at static power. One or a combination of both can exist used, depending on the characteristics of the engineering in which the processor is fabricated.

DVFS

Dynamic voltage and frequency scaling (DVFS) is designed to optimize dynamic power consumption. DVFS takes advantage of the human relationship between speed and power consumption as a function of power supply voltage:

•

The speed of CMOS logic is proportional to the power supply voltage.

•

The ability consumption of CMOS is proportional to the square of the ability supply voltage (Fiveⁱⁱ).

Therefore, by reducing the ability supply voltage to the lowest level that provides the required performance, we can significantly reduce power consumption. DVFS controllers simultaneously adjust the ability supply voltage and clock speed based on a command setting from software.

Race-to-dark

Race-to-dark (also called race-to-sleep) is designed to minimize static power consumption. If leakage current is very high, then the best strategy is to run as fast equally possible and and then close downwards the CPU.

Assay

DVFS and race-to-dark can exist used in combination by selecting a moderate clock speed that is between the values dictated past pure DVFS or race-to-dark. We can understand the trade-off strategy using a model for total energy consumption:

(iii.3) $E_{tot}=\underset{0}{\overset{T}{\int }}P\left(t\right)dt=\underset{0}{\overset{T}{\int }}\left[P_{dy\mathrm{north}}\left(t\right)+P_{\mathrm{due\; south}tatic}\left(t\right)\right]dt$

The total energy consumed in a given interval is the sum of the dynamic and static components. Static free energy is roughly constant (ignoring whatever efforts past the CPU to temporarily turn off idle units) while dynamic power consumption depends on the clock charge per unit. If we plough off the CPU, both components get to zero.

Nosotros as well have to take into account the time required to modify ability supply voltage or clock speed. If mode changes take long enough, the free energy lost during the transition may exist greater than the savings given by the mode change.

Read total chapter

URL:

https://world wide web.sciencedirect.com/scientific discipline/commodity/pii/B9780128053874000030

Dark Silicon and Future On-chip Systems

Amin Rezaei , ... Hai Zhou , in Advances in Computers, 2022

2.iii Methodology

Experiments are performed on a cycle-authentic many-core platform implemented in SystemC. A pruned version of an open source simulator for mesh-based NoCs, called Noxim [25], is utilized equally its communication architecture. For power and temperature simulations, power and thermal models taken from Refs. [26,27] are integrated equally libraries into the simulator. By an optimistic assumption each sprinting menses is considered to be equal to the cooldown catamenia. Moreover, each core in sprinting status is supposed to gain performance past a factor of 4 and to lose life span by a factor of ii compared to the core in nominal condition. Maximum traffic injection rate and the average lifetime of the system running in idle status are called λ _full and γ _idle, respectively. Some multithreaded applications from the PARSEC [28] criterion suite are used in the experiments. Iii different network sizes of xvi (no dark silicon), 36 (55% dark silicon), and 64 (75% night silicon) cores are considered in the simulations. Comparisons are also made between SS and two state-of-the-arts architectures: CS [4] and NS [vi].

ii.3.1 Power Model

Power consumption of each cadre is modeled equally 2 major parameters, i.e., dynamic power due to transistor switching and static power due to leakage. Therefore, total power for each core is given by:

(1) $P_{\text{total}}=P_{\text{dynamic}}+P_{\text{leakage}}$

Generally, dynamic power is given past:

(2) $P_{\text{dynamic}}=\alpha C_{\mathrm{east}}{V}_{\mathrm{dd}}^{2}f$

where α is the activeness factor, C _e is the effective capacitance, V _dd is the supply voltage, and f is the running frequency of the core. Assuming both the activeness factor and the effective capacitance equally constants, dynamic power changes quadratically to supply voltage and linearly to frequency. According to Ref. [29], the static power of a system is given by:

(iii) $P_{\text{leakage}}=V_{\mathrm{dd}}{N}_{\mathrm{tr}}{k}_{\mathrm{d}}{I}_{\mathrm{s}}$

where N _tr is the number of transistors, k _d is a device-specific constant, and I _s is the normalized static current for each transistor that is proportional to the leakage current of a single transistor.

ii.iii.ii Thermal Model

For the thermal model, a well-known RC thermal network, divers in Ref. [27], is adopted in the simulator, which considers the duality betwixt thermal and electrical circuits. According to this model, the steady-state temperatures of the cores can be computed as follows:

(four) $T^{\mathrm{C}}=B{P}^{\mathrm{C}}+H_{\mathrm{one}}+H_2$

where T ^C is the steady-state temperatures of all the cores on the chip. Matrix B contains the amount of the oestrus contribution of all the cores. Thus, heat transfer among the cores is considered. Column vector H _i contains the heat contribution of the other thermal nodes to the cores, while H ₂ contains the heat contribution of the ambient temperature to the cores. Assuming only one application tin can be executed in a sprinted cadre, with respect to the mapping of applications, a binary matrix L  =   [L _{i, j} ] is defined. If application a _j is mapped to cadre c _i , [L _{i, j} ]   =   1; otherwise, [L _{i, j} ]   =   0.

Past involving the power vector of the applications and the mapping matrix in Eq. (4), nosotros become the following equation:

(5) $T^{\mathrm{C}}=\mathit{BL}{P}^{\mathrm{A}}+H$

where P ^A is a column vector containing the ability consumptions of all the applications and H is the sum of H ₁ and H ₂. Furthermore, Eq. (6) expresses the direct relation between application power consumption and the steady-land temperature of any core c _i :

(6) $T^{c_i}=\sum _{j=1}^k{b}_i{l}_j{p}_j+H$

where b _i is the row i from matrix B which corresponds to core c _i , 50 _j is the column j of matrix L which corresponds to application a _j , and p _j is the ability consumption of awarding a _j . Hence, the thermal model calculates the steady-state temperature for any core of the organisation given the mapping matrix of the applications.

Read total affiliate

URL:

https://world wide web.sciencedirect.com/science/article/pii/S0065245818300251

Dataflow Processing

Zivojin Sustran , ... Veljko Milutinovic , in Advances in Computers, 2022

eleven.5 Power Consumption

The ability consumption of the system was estimated using CACTI.four.ii. The results, with some reference values, are shown in Fig. 29. The modified STS system dissipated around x% more than power than the traditional cache organization used for reference.

Figure 29. Ability consumption of modified STS systems. Clarification: a modified STS organisation dissipates around 10% more power than the traditional cache organization used for reference.

We used CACTI tool likewise to verify that temporal cache latency could be at least 30% shorter than that of the spatial cache.

Results show an estimated access latency of 0.96 (ns) for the 64K spatial part of the cache, and 0.631 and 0.64 (ns) for 1K and 2K direct caches, respectively. The latter ones would be used for the temporal cache part.

The power consumption in most contempo efforts [22] was considered with highest priority. This inquiry exploits retentivity region semantics to partition data cache.

Read full affiliate

URL:

https://world wide web.sciencedirect.com/science/article/pii/S0065245814000060

Techniques to Measure, Model, and Manage Power

Bhavishya Goel , ... Magnus Själander , in Advances in Computers, 2022

4.ii.1 Temperature Effects

Processor power consumption consists of both dynamic and static elements. Among these, the static power consumption is dependent on the core temperature. Equation (half dozen) shows that the static power consumption of a processor is a function of both leakage current and supply voltage. The processor leakage current is, in turn, afflicted by process engineering science, supply voltage, and temperature. With the increase in processor power consumption, processor temperature increases. This increase in temperature increases leakage current, which, in turn, increases static processor power consumption. To study the furnishings of temperature on ability consumption, nosotros ran a multithreaded program executing MOV operations in an infinite loop on our Intel® Core™ i7 automobile. The behavior of the program over its unabridged run remains very consistent. This indicates that the dynamic ability consumption of the processor changes little over the run of the programme. Figure 14a shows that the total ability consumption of the machine increases during the program'south runtime, and information technology coincides with the increase in bit temperature, while the CPU load remains constant. Thus, the gradual increase in ability consumption during the run of this program can be attributed to the coincidental, gradual increase in temperature. The full power consumption increases by almost x% due to the change in temperature. To account for this increase in static ability, it is necessary to include temperature.

Fig. 14.. Temperature furnishings on power consumption.

(half-dozen) $P_{\text{static}}=\sum I_{\text{leakage}}\ast {\mathrm{Five}}_{\text{core}}=\sum I_{\text{s}}({\mathrm{due\; east}}^{{\mathit{qV}}_{\text{d}}/\mathit{kT}}-\mathrm{ane})\ast V_{\text{cadre}}\text{,}$

where $I_{\text{s}}$   =   reverse saturation current; $V_{\text{d}}$   =   diode voltage; chiliad  =   Boltzmann's constant; q  =   electronic charge; T  =   temperature; ${\mathrm{Five}}_{\text{core}}$   =   cadre supply voltage.

As per Eqn (6), the static power consumption increases exponentially with temperature. We confirm this empirically by plotting the cyberspace increment in ability consumption one time the program starts execution at the higher temperature, as shown in Fig. 14b. The non-regression analysis gives us Eqn (vii), and the bend fit shown in Fig. fourteenb, which closely follows the empirical data points with decision coefficient $R^2=0.995$

(7) $P_{\text{staticInc}}=1.4356\times {\mathrm{i.034}}^T\text{,}\text{when}{\mathrm{Five}}_{\text{cadre}}=\mathrm{ane.09}\text{V.}$

Plotting this guess of increment in static ability consumption, as in Fig. 14c, explains the gradual rising in full power consumption when the dynamic behavior of a program remains abiding.

Goel et al. [21] include the temperature effects in their power model to account for the increase in static power consumption every bit flake temperature increases. Instead of using a nonlinear function, they approximate the static power increment equally a linear function of temperature. This is a fair approximation because that the nonlinear equation given in Eqn (7), can exist closely approximated with linear equation given in Eqn (8) with determination coefficient $R^{\mathrm{ii}}=0.989$ for the range in which die temperature changes occur. This linear approximation is a trade-off for avoiding the added price of introducing an additional exponential term in the model.

(8) $P_{\text{staticInc}}=0.359\times T-16.566\text{,}\text{when}{V}_{\text{cadre}}=\mathrm{ane.09}\text{V.}$

Modern processors allow programmers to read temperature information for each core from on-dice thermal diodes. For example, Intel platforms written report relative cadre temperatures on-die via Digital Thermal Sensors (DTS), which can be read past software through Model Specific Registers (MSRs) or the Platform Surroundings Control Interface (PECI) [vi]. This information is used past the system to regulate CPU fan speed or to throttle the processor in instance of overheating. 3rd-party tools like RealTemp and CoreTemp on Windows and open-source software like lm-sensors on Linux can be used to read data from thermal sensors. Every bit Intel documents indicate [half-dozen], the accuracy of temperature readings provided by thermal sensors varies, and the values reported are not exactly equal to the actual core temperatures. Because of manufactory variation and individual DTS calibration, accuracy of readings varies from fleck to chip. The DTS equipment also suffers from slope errors, which means that temperature readings are more than accurate near the T-junction max (the maximum temperature that cores tin can reach before thermal throttling is activated) than at lower temperatures. DTS circuits are designed to be read over reasonable operating temperature ranges, and the readings may not show lower values than twenty   °C fifty-fifty if the actual core temperature is lower. Since DTS is primarily created as a thermal protection mechanism, reasonable accuracy at high temperatures is acceptable. Simply this affects the accurateness of power models using cadre temperature. Researchers and practitioners should read the processor model datasheet, design guidelines, and errata to understand the limitations of their respective thermal monitoring circuits and take corrective measures for their power models, if required.

Read full chapter

URL:

https://world wide web.sciencedirect.com/science/article/pii/B978012396528800002X

Source: https://www.sciencedirect.com/topics/computer-science/power-consumption

Posted by: ellenderbeely1967.blogspot.com

Share this post