CONTRIBUTION OF COMPUTERS HEAT IN GLOBAL WARMING

Contribution of Computers Heat in Global Warming

Virender Kumar

Asst .Prof IIMS Bareilly

virendra.k@invertis.orgSunil Kumar Sharma

Lecturer INC Bareilly

sunil.sharma.icfai@gmail.comABSTRACT

Computers convert electricity in to heat as they operate. As they switch on and off, transistors produce heat, this heat must be dissipated in order to keep these components within their safe operating temperatures Components which gain heat are susceptible to performance loss and damage. These components usually includes integrated circuits such as CPUs, chipset and graphics cards, along with hard drives. Hence to encounter this heat at least equal amount of cooling is required by some external source such as Air Conditioners or Coolers, which too in turn release abundant amount of heat which is ultimately transferred to environment. The exponential rates at which computers are increasing world wide have raised concerns of environmentalist. According to a survey conducted by Forrester Research marketing company, the number of personal computers in the world has reached one billion already by the end of 2008, and it will reach two billion - by 2015. Forrester Researchs forecast is based on the assumption, that from 2003 to 2015 the total number of personal computers in the world will annually increase by 12 percent. In this paper we will present a model that will show contribution of computers heat to global warming in current year (2009) also, on the bases of research conducted by Forrester Research marketing company, well also show the contribution of computers heat in global warming world wide by 2015 .Beside this we have also generated an equation which will calculate heat generated by computers world wide and their subsequent contribution in global warming in any year.Keywords: Green Computing, Global Warming, Computers Heat,1. INTRODUCTION

We love our computers for all the ways they make our lives (and the world) better, and easier. Any work which earlier seems to be impossible is now possible and that too by a single click of mouse. Performance-wise, computer design has progressed staggeringly well and astonishingly fast but looking at it from a green perspective, the work has barely begun. Computer consumes a good amount of energy, and major fraction of this energy is dissipated in the form of heat. Hence, these computers are also contributing very actively in Global Warming like any thing else.By the law of conservation of energy, all the electricity coming into the computers has to go out somehow. It does so as heat. Obviously, the largest by-product of power is heat.This means that there should be a close correlation between watts of power consumed by the CPU and the heat it releases. Normally, heat (which is just a form of energy) is not measured in watt-hours, but in units such as calories or joules or BTUs. (A BTU (British Thermal Unit) is the amount of energy (heat) required to raise the temperature of 1 pound of water 1 degree Fahrenheit). The conversion from watt-hours is straightforward and is defined as:

1 watt-hour = 3.41 BTUs

Now if we knew how much amperage a CPU was consuming by use of a multi-meter in line with the CPU, we can calculate the wattage it is consuming by the formula P(watts) = I (current[amps]) x E (electromotive force [volts]) and by that could calculate the theoretical CPU heat released. As a matter of fact all microprocessors are about 10% energy efficient, this means of all the electricity that we pump in, 10% does the job (like running a program), while the rest of the 90% is dissipated as heat. Let us take 2 examples to calculate heat dissipated firstly in case of 100 watt bulb than in case of a P4 processor:

Case 1 100 w light bulb

For example, a 100 watt light bulb that is on for 1 hour consumes 100 watt-hours of electricity and generates about 100 watt-hours of heat (assuming 100%(total) conversion of power in to heat energy )

Now we have 1 watt hour = 3.41 BTUs

Hence; heat generated = 100 watt-hours x 3.41 BTUs/watt-hour

Heat generated = 341 BTUs

Case 2 Intel Pentium IV 672

For example, a Intel Pentium IV 672 processor consumes 115.0 w of power, if this is working for one hour than it will consume 115.0 w of electricity and generates about 103.5 watt-hour of heat (assuming 10 % energy efficiency)

Now we have 1 watt hour = 3.41 BTUs

Hence; heat generated =103.5 watt-hour * 3.41 BTUs/watt-hour

Heat Generated = 352.935 BTUs

2. Model for calculating -Total number of computers in the world In general we can calculate the total number of computers in the entire world, (2008 onwards) on the basis of general formula (model)Total number of Computers = 1,000,000,000(1+r/100) tWhere

r = rate of growth, which is evidently 12% as per Foresters survey

t = time

t = 0, for 2008

t = 1, for 2009

By applying above model total number of computers in various years (2008-onwards) comes out to be:

2008 = 1,000,000,000 2009 = 1,120,000,0002010 = 1,254,400,0002011 = 1,404,928,0002012 = 1,573,519,3602013 = 1,762,341,6832014 = 1,973,822,6852015 = 2,210,681,407And so on, these results are represented clearly, Figure 1.0 represents the results by using a bar diagram, whereas Figure 1.1 represents same results by using a line The total heat emitted in Btu/hr, can be further calculated by multiplying average heat emitted from a single CPU to the total number of CPU in the world.3 .Moores Law and CPU HEAT DISSIPationThe numbers of transistor, incorporated in a chip will approximately double every 24 months (MOORES LAW). This law is clearly demonstrated in Figure 2.0 .Moores law shows one side of the coin. Technology advancements focus on output performance CPUs, despite advancements, consume more power and subsequently dissipate more heat.

Heat dissipated is mainly affected by the CPU utilization

Heat dissipation is linear to the CPU utilization [Impact-2008]P = a U+ b

As the number of transistors with in a chip increases and physical size of the chip is reduced , clock speed is increased all these has boosted the heat dissipation across the chip to a dangerous level , these increased number of transistors results in a good power consumption, which in turn releases abundant heat. This is the basic reason for Intel Pentium 75 requiring as low as 6.0 w for its functioning whereas Pentium D 840 requiring 130 w for its functioning. However the heat dissipated does not strictly follow Moores law, this is evident from various types of processors and their subsequent power consumption (Figure 2.1) in which even though the number of transistors are increased manifold, even than power consumption and heat dissipation has subsequently reduced.Power and performance relationshipIncrease in power consumption has occurred despite dramatic and revolutionary improvements in process technology and circuit design. The primary reason behind the increase in power has been the continued emphasis on higher performance[Shameen Akhter-2006].As complexity and performance of processors has increased over the year to provide unprecedented level of performance the power required to supply these processors has increased steadily too. A simplified equation that demonstrates power performance relationship for the CMOS circuits on which all modern processors are based is:P=ACV2f

The first component in the equation captures dynamic power of charging and transistor circuits from which basic functional blocks of processor are built. The power is directly proportional to switching frequency (f), square of the supply voltage (V), and total capacitance(C).Since not all functional blocks are used at any given time by a processor workload and not all gates are switched , A represents activity factor or the number of switched transistors on a die.The equation above demonstrates the tradeoff between performance and power. As processor frequency increases so does power consumption. As processors architecture becomes more complex to support greater levels of instruction level parallelism and increased performance, capacitance of system increases and so does dynamic power. Another and more subtle point in that concurrent processing can also lead to significant power reduction. Splitting the workload in to multiple threads and running them in parallel can significantly increase processor power efficiency .This is the basic reason why a subsequent power reduction has been observed in multi core and other advanced processors.

4 Heat dissipated by a specific processorThe waste heat also causes reliability problems, as CPU's crash much more often at higher temperatures. The central processing unit is the most significant source of heat in a modern PC. Heat dissipated by a specific processor depends on many factor, with in a specific category of CPU(say Pentium II), the power consumption varies to a great extent, for example: some model of Pentium II takes as low as 14.0 w where as other model of Pentium II consumes 34.5 w. Heat Dissipation in year 2009 and in 2015:

Average heat dissipated globally by all PCs is dependent primarily on many factors such as, the type and model of processors used, operation time, idle time etc. Hence by calculating the power consumption of that specific processor one can easily calculate heat dissipated. This calculation is possible only in situations when we are aware about the fact that what fraction of worlds population is using which type (model) of processor. Once these figures are available calculation of heat dissipated in any year is an easy task. And on the basic of our model discussed in section 2 of this paper we can easily calculate the global heat in any year. Table 1.0 shows average heat dissipated by a specific category of processors. We have taken average power consumption of majority of Intel processors. Within a specific group (such as Pentium Core 2 duo) there are various models whose power consumption is variable to a great extent (for Pentium core to duo it is from 5.5 w to 150 w) hence average power is taken. Now as 10 % of power is used by system, for running programs etc, therefore if we subtract this 10 % power from total power of a CPU (or average here), than we will get the power that will be converted in to heat.Table 1 shows total heat generated globally (per hour for year 2009 and 2015). This is only applicable if we consider that world is using a specific type of processor. S.NoProcessor TypeAverage Power(w) consumedPower as heat dissipatedGlobal heat generated(2009)Global heat generated(2015)

1Pentium12.49 11.2511.25*3.41*1,254,400,000 =48121920000 Btu/hr11.25*3.41*2,210,681,407 = 84807265476 Btu/hr

2Pentium II27.2524.5324.53*3.41*1,254,400,000 =104927173120 Btu/hr24.53*3.41*2,210,681,407=184917530855 Btu/hr

3Pentium III25.5322.9822.98*3.41*1,254,400,000 = 98297041920 Btu/hr22.98*3.41*2,210,681,407=173232974279 Btu/hr

4Pentium IV82.5674.3174.31*3.41*1,254,400,000 = 317861322240 Btu/hr74.31*3.41*2,210,681,407=560180257557 Btu/hr

5Pentium IV Ex. Ed105.1794.6794.67*3.41*1,254,400,000 = 404951303680 Btu/hr94.67*3.41*2,210,681,407 =713662562010 Btu/hr

6Pentium D102.7792.5792.57*3.41*1,254,400,000 = 395968545280 Btu/hr92.57*3.41*2,210,681,407=697831872454 Btu/hr

7Pentium Dual Core5347.747.7*3.41*1,254,400,000=204036940800 Btu/hr47.7*3.41*2,210,681,407=359582805618 Btu/hr

8Pentium Core 2 Duo4944.144.1*3.41*1,254,400,000=188637926400 Btu/hr44.1*3.41*2,210,681,407=332444480666 Btu/hr

Table 15. Heat Dissipation in Idle and in max use stateIn reality, processors in mobile desktop and server platforms spend a significant amount of time doing nothing or being idle. Ensuring that processors power consumption is minimal in this state is critical for overall power efficiency.Heat dissipated in idle and max used(100% load) state (Figure 3.0 and 3.1 respectively) can be easily calculated if we have the power consumption reading .To determine the maximum power consumption of our review sample, we stressed all four cores using Prime95. Compared to its direct quad-core predecessor with the Kentsfield core (Conroe), the new Penryn-based CPU with its High-K Metal Gate technology draws 39.25% less power. This is because of the fact that the Penryn processor is able to signal the motherboard when it is idling. The motherboard then switches off parts of the voltage plane in order to conserve power further. 6. Heat dissipated when a cpu is not fully used All CPUs are not always utilized to the max state, neither they remain idle always, hence in these situations when CPU is used to a fixed percentage [Green grid (2008)] (say 10%,20% etc) than calculating heat dissipated in that situation require some alternative model.

For n% CPU utilization, the power used Pn can be calculated using the following formulaPn = Pidle + (Pmax Pidle) / 100 * nSo for an example where Pidle= 200W and Pmax = 300W at 10% CPU utilization

P10 = 200 + (300 - 200) / 100 * 10 = 210W

Now once we have calculated the power than by subtracting 10% of the power (which is used eg.. for running a program), we can get the rest of the power as heat dissipated.Figure 4.0 shows heat dissipated v/s CPU utilization; it is evident that heat dissipated increases with CPU utilization.

7. Solution for reducing global warming by computersComputer dissipates a good amount of heat. In fact a major cost of running an organization comes from supporting the PC infrastructure. Virtual desktop computing based on NComputing saves money up front and over time. NComputing consumes less power, generates less heat, lasts longer, and produces less e-waste. Ultimately, the NComputing green advantage helps organizations pursue their missions while they lessen their environmental impact.

According to latest researches over 850 million PCs are turned on every day. If NComputing systems were used instead (at a ratio of six access devices to each PC) there would be substantial immediate and long-term environmental benefits. The impact on the environment of adopting NComputing solutions would be enormous. Energy use would decline by a good amount (83%).

The electricity usage decline would save nearly approx 15 million metric tons of coal each year and would eliminate the need for 120 megawatts of coal power capacity.

CO2 emissions would decrease by 96 million metric tons. This is equivalent to planting nearly 460 million trees.

Disposing of NComputing devices (0.33 lb each), rather than disposing of an equal number of PCs (21 lbs each) would save over 6.7 million metric tons of e-waste. Heat Dissipation can be reduced to minimum level

And that is just for the PCs in use today. There are another billion users who will join the digital world by 2015, than this ill effect will be clearly visible.8. CONCLUSIONSPower is a challenge to the entire semiconductor industry and nothing new to CPU. Smaller transistors consume less power, but as transistor density and speed rise, the overall chip consumes more power and generates more heat. People seldom count this heat, and its contribution to global warming. But as the entire world has already crossed 1 billion PC in count, and this count will reach more than 2 billion in 2015, hence this area cannot be neglected. Processors manufacturers should focus on the development of processors which consumes less power and dissipates less heat. Intel and AMD have already enrolled in this mission with the introduction of multi core processors. Beside this NComputing can also be adapted as a good measure to reduce global warming by computers9. AcknowledgementSI acknowledge all those who have played a major role in this research paper; firstly I would like to start of with the coauthor Mr. Sunil Kumar Sharma, who has played a major role in completion of this paper. I would like to thank Mr. Umesh Gautam Chancellor Invertis University for his continuous motivational and financial support at every level. I would like to acknowledge Dr.YDS Arya Director IIET Bareilly, as well as Prof. Arpan Khastagir, Director IIMS for sharing valuable experiences with me. At last I would also like to acknowledge my Father, Mother, Brother, My Wife and my Kids- Gunjan and Vansh, and of course my valuable friends without whose support this work would have been daunting task. For any one I may have missed please accept my apologies.

10. References:[Impact-2008] Georgios Varsamopoulos, Sandeep Gupta Thermal-Aware Task Placement in Data Centers[Shameem Akhter-2006]Shameem Akther , Jason Roberts, Multi Core programming increasing performance through software multithreading ,Intel Press -ISBN-09764832-4-6[Green grid(2008)]Five Ways to Save Server Power Mark Blackburn, 1E Operations Work Group-green grid(2008)[Cluster 2007]Thermal Aware Task Scheduling for Datacenters through MinimizingHeat Recirculation

[GreenCom2007 A]Thermal Aware Task Scheduling for Datacenters through MinimizingHeat Recirculation,

[GreenCom2007 B]Thermal Aware Task Scheduling for Datacenters through MinimizingHeat Recirculation,