How much electricity does the Internet use?

little bit. You know, you might remember several podcasts ago we we kind of compared the paper technology world and the computer technology world. Uh well, we started thinking about other topics we could talk about. They're kind of related to that. And one of those is an interesting question that a couple of people have asked, not just us, but in general, and that question is how much electricity does the Internet require to work? Lots? Yeah, but if

you wanted to get a little more specific, actually nobody knows. Yeah, we we can. We can get that out of the way. First thing, We're going to go ahead and be completely straight with you guys. It is practically impossible to say with any degree of certainty how much electricity the Internet requires to work. Part of the reason for that is that you're talking about such a massive machine, really the biggest machine in the in the world, and it's constantly changing.

It's growing all the time. Uh, the components are changing all the time. So what might be true today could be completely false in a year from now. It may be that perhaps eighteen more data centers go online in every region in the United States, and that could change things. It could be that some sort of massive energy saving technology comes out and that could that could impact it.

So when we talk about how much electricity the Internet uses, keep in mind we're talking about estimates and vague ones at that, because a lot of the data that exists out there that people basically these sort of estimates on is more than a decade old. Yeah, And um, it's one of those things too where people are sort of extrapolating based on numbers that they do know. So, I mean, they're educated guesses when we we do have an idea of what's going on. Yeah, I mean, think of it

this way. Even if we knew down to the very last device, how many computers and smartphones and everything else were connected to the Internet, including servers. Even if we knew all that, we don't necessarily know how often they're connected to the Internet. So just because a computer years on doesn't necessarily mean it's actually making use of the Internet,

so it's not putting a drain on any resources. So then you're talking about all right, well, now we have to estimate not only how many devices are out there, but how how much time they're spending on the Internet. Right, so this is really about as fast and loose as you can possibly get and still have something to talk about. Well, we do know that the machines that are using the most electricity on the Internet are the servers, the data centers,

which are sort of little internets in a way. Yeah, you could, you could say that the data centers on a on an individual basis, the server and a data center is using more electricity than any person's computer really, unless you're talking about I guess a supercomputer, in which case that's a little different. But if you're talking about

sheer numbers. There are so many computers out there that if you were to look at at least some of the figures that I looked at, the biggest drain on the the electric grid of the the world, really, not just the United States, comes from consumer and corporate computers, not from data centers. And that's because just from a sheer scale, you're talking about just a huge, uh difference

in numbers. So data centers, yeah, there are a lot of them, and there are thousands of servers in these data centers, but we're talking millions and millions of computers, or even perhaps one billion computers. But I think that was a figure that was arrived at maybe a year or two ago, and so again not necessarily accurate. Now they're probably even more. And also then we get into

what is a computer. Before we get too far into all of these kind of vague discussions that are going to take up our next twenty minutes, I think it might help to kind of talk about energy and electricity in general, about how that's measured. So we generally talk about electricity in a term called kill a watt hours, which is a unit of energy, and it's essentially the equivalent to one kill a lot of power expended over

the course of one hour. All right. So the reason why we do this is because energy is essentially a measurement of power times time. So it's the amount of power times the amount of time that that power is expended. That's how you get energy and kill what hour unit of energy? Now, when we're talking about global electricity use as far as the Internet goes, we're talking in the tarole wat range tarra watt hour range. And to give you guys an idea, Okay, so I kill a what

is a thousand watts? A mega watt is a thousand kill a watts? A giga what is a thousand megawatts? And a tara lot is a thousand giga watts? So huge number here, enormous number. It's too big for me to actually imagine. I know what, I know it exists, but I can't fit it into my head, right right. It's nice of them to use these terms that we would probably be familiar with from things like bites. Yeah, it's it's nice that it's not you know, a jot and then a bit tool and then four handspans and

then a furlong. Yeah, that would be really confusing. Yeah, thank you for the metric system. Um, So here's a here's a study that was done back in and it was done by Berkeley Labs Environmental Energy Technologies Division, the End Use Energy Forecasting group led by Dr Jonathan kumi who I have I've seen his notes on several different message boards about this subject. Dr Kumy is still very much active in this kind of study. Now back in the Berkeley group actually had money to conduct this study.

I think he said that it costs somewhere around the three dollar range to do the study because you're you're you're compiling lots and lots of data. You have to gather all this information and then you have to make sense of it and um and then actually does take a lot of time and money. It sounds like it shouldn't you know. It sounds like something you could plug into Wolf from ALFA and get an answer immediately. It's

not the way this works. So their analysis said that the total electricity used by office and network equipment is about seventy four tara watt hours per year, which was about two percent of the total electricity use in the United States. And then if you were to throw in telephone switching equipment and manufacturing energy for semiconductors and computers, uh, it knocked it up to about three of all electricity in use in the United States. So why would you

throw in the telephone switching. Well, because modems. The studies done in it was based on figures from ninety and we didn't all have DSL lines and cable modems at that point, right, and so we're talking seventy four Tara Lott hours at that time. That's a huge amount. Uh. And this was just for the United States to this was not a global study, so we're really talking about

just one nation. But the other problem that we have here is that if you're talking about figures that are based off of n equipment, a lot has changed in the years between and now we are using computers that have much more efficient monitors. Were using a lot more laptops, which in general consume less energy than a desktop computer would, but there are also lots and lots more of them. So you have to figure, all right, well, the energy efficiency angle, how much does that impact this equation the

number of computer is, how has that impacted? Uh? The and you start to see where the problem is here. You can't just easily adjust these numbers and say, oh, well, let's just you know, we'll fudge this one a little bit and we'll switch this one a little bit. Um without a full study. You can't say for sure how

much energy is being consumed. True enough, UM, I actually uh, I did a blog post on the tech stuff blog not too long ago, um that uh, in which I quoted Bobby Johnson who published in the Guardian, the paper from the United Kingdom, and UM, basically uh the information I had read. And there are actually a couple of articles in the Guardian about UH these topics. So apparently

they are among those who are interested in finding out more. Um. But there are a lot of people who are starting to become concerned by the uh energy usage from the Internet. I mean, UM, from from what I can tell, experts believe that data centers on the Internet U use about a hundred fifty two billion kilowatt hours of electricity per year, and that's two percent of the world's carbon dioxide emissions

as well. UM. As a matter of fact, uh, a thousand searches apparently create about the same uh CEO two emissions as an average European car going one kilometer according to the Guardian. So UM, you know that's just I figure a thousand searches in Google that's you know, probably as much as we do in a day at how stuff works with the editorial staff. Maybe Lamborghini one kilometer average. Yeah, I don't think Lamborghini is an average European car. Don't

get destroy my fantasies of how people live in Europe. Oh, you could ask the guys at high Speed stuff if that's accurate. I'll stop in next time they're recording. But the thing is um the Climate Group says emissions related to computers are probably going to go up around that's a hundred one point for gigatons of carbon dioxide by because the Internet is just growing at that sort of

a rate. That's pretty significant. Yeah, and back in two thousand seven, there were some people who attempted to take another stab at guestimating. Essentially. I mean, I'm it sounds like I'm I'm slapping them for their their efforts, But really, when you get down to it, when you don't have the data in front of you, it is a guestimation.

It's hard to quantify something like this too, because it's so you know, how would you break down the energy from somebody's electric meter on the side of their house and say, oh, well, you know your fridge used you know an eight to this, Yeah, and you have to figure out as well. Even with servers. I mean there there's a difference between the amount of energy a server will consume at a peak time versus a trough um.

That's all that's called your your load differential. There, you know your difference and load and the uh if you if you have a really efficient server, um, the difference is going to be minor, like within the five to ten percent range where your your server will be consistent and it's not gonna very wildly from full use to trough use. And that's just meaning that it's very efficient, not that it's not that it's the all underpowered or

anything like that. Um, you know, before we go on, just going to say too that I think that I'm just guessing, but I'm betting that the data centers, the facts and figures that we're seeing from the data centers, probably don't take into account the massive amount of air conditioning needed to keep those machines cool. Right, So you're not breaking down the different you're you're breaking down the

building as a whole. You know, say well, this Google data center uses x amount of electricity per year, Well it doesn't say well, part of that the lighting for the people who work in the building, and part of that is the you know, fifty three air conditioning units sitting on the roof and all that stuff. It doesn't it doesn't go any further than that. Yeah, well there's and there are some studies that specify whether or not they take that kind of thing into account. And of

course not all data centers use air conditioning. Some use water cooling systems instead. And I was going to mention something along those lines in a minute. All right, well, then I'll leave that for then. But the actually I should say that when I said a load differential, that's that's actually my brain throwing a wrong word, and it should have been load factor. I'm just gonna say that now because I'm sure people are already emailing me. But now that you've hit send, just know that I did

eventually get around to getting it right. Um so the uh but no, Back in two thousand seven, this other study that I was talking about, it's not even really a full study. Uh, there was a question online about well, how much electricity does the internet use? And people were talking about this study and trying to extrapolate from that how much energy was being used at that time At two thousand and two thousand seven. So the person who

tackled this came up with some interesting numbers. Uh. He decided that, UM, once he studied this carefully, said that the world consumption in billions of kilowatt hours for a for data centers was a hundred and twelve point five

billion kilowatt hours. All right, the PCs and monitors five eight billion kilowatt hours, bodem's routers and that sort of thing hundreds or hundred sixty seven billion kilowatt hours, and then phone network one billion kilowatt hours for a grand total of approximately eight hundred sixty eight billion kilowatt hours. And that was in two thousand seven. But here's the thing was that he was basing this off of figures for um, these older computers, So he's looking at power

consumption figures from years and years ago. So we're talking about old CRT monitors which were not as efficient necessarily as some of the l C D and plasma monitors that we have today. Yeah, you know, I saw a number on NPR that was interesting. Um. Their website said that, uh, this is not kill a lot hours, but um, they say that in the United States, more than one billion dollars has spent every year keeping on computer monitors that

are not really in use. Right, that's you know how accurate that is, But that's you know, I would I believe it. People leave their computers on and screens are are pretty they're pretty energy and hungry, and uh yeah, so put your computers to sleep, or at least your monitors the So yeah, the this eight h sixty eight billion. It's one of those figures that some people like to to say that's how much energy the Internet consumes per year, and other people point out, well, that's not really true

because we don't have enough information. We don't you know this this is taking into account old equipment. And yeah, while monitors have gotten better over the years, we've also had faster processors, which consume more powers. So as monitors have become more efficient, processors have demanded more energy. So

it's again one of these sort of seesaw things. You have to look at all the different factors and there's no easy way to say, all, right, well, how many high end computers are out there, how often are they accessing the Internet. How many low end computers are there, like netbooks, how many netbooks those are accessing the Internet all the time because they don't have the native capabilities to really do everything you need to do on the

computer itself. So that's why this this conversation is a difficult one to have, but sixty eight billion, it's good number to kind of keep in the back of your mind. That's sort of a best guess right now. But we're gonna talk a little bit more about energy, I assume because you had some stuff about water coolings the THEMS at least. Yeah, one of the uh, they're talking about supercomputers.

I ran across some information about IBMS Power five seven five supercomputer UM, which actually was linked to from one of the other posts, and I thought this was pretty cool because it's got it uses water chilled copper plates over each of its microprocessors UM and that's a lot because they're four processor course on each rack of the supercomputer UM. But the water cooling uses less electricity than you would using traditional cooling methods, so that's also much

more efficient. Yeah, because you know, if you think things like that could help us cut down Yeah, when you think about it, UM, water is just it's a it's a better cooling system because it as a fluid. It is more efficient at carrying heat away, and you can just run water across, uh, you know, using some sort of tube system. Obviously you're not just you're not just dousing serves and water necessarily. You know, I tried that. I just turned a hose on my computer in it.

It's it doesn't run that way at all. Why, Yeah, that's probably a bad way to do a water cooling system. But if you're doing it correctly, and the fire department was upset too, sure the water, the water will carry the heat away much more efficiently than air will. And you can then reclaim the water, you know, you you just cycle it exactly. You just you know, move it through so that it can release the heat, and you

reclaim the water. You cycle it back through, so it's not like you're using tons and tons of water and it's just going to waste. You're using the same water in the system, and then occasionally, I'm sure, replenishing it, but not like not like you just turned the tap on and let it go. You know, I wish I had made the turning the hose on the computer joke because I was going to say, yeah, and if you put the fish in there, that the heat kind of cooks the fish. But yeah, I've already made a joke,

so I won't make um. You know, one of the things that I found that was funny was and the reason I find it funny is because it's so contested. But um blogger and author Nicholas carr Uh said he he did a blog post that a lot of people disagree with a few months ago, but it said, um, the Second Life avatars themselves, just keeping the avatars running use one thousand, seven or fifty two kilowatt hours per year. The average human uses two thousand four and thirty six

kilowte hours per year. But uh, about the person the average person in Brazil, The average Brazilian uses one thousand four kilo one hours, So basically Second Life people in Second Life use almost as much electricity as the average Brazilian. He says, Now he's you know a lot of people disagree with him, but if it's even close to that, now they do they leave the company that that runs Second Life, leaves the servers running and the avatars are on for as long as that's the thing they got in.

They broke it down and said, well, you know, if the avatar is only on for two hours there. You know, such and such average people online, and it's not all the people on Second Life versus the average Brazilian who you know runs twenty four hours a day. Uh so, Yeah. I just think it's funny though, to compare a virtual

person to a human being. Um, you know, I would imagine that they're probably I don't know how accurate it is, but I would imagine that it's got to to run all those those data centers to keep the world of Second Life going. Um, you know, it would be at least reasonably comparable to what happening been in Second Life. See, they just need to hit that switch. They just need to hit that switch. I mean, you know, it's no

reason to keep that going, is all I'm saying. Alright, alright, so I've got nothing more to say about Second Life. I've never tried it. Yeah, well, you value your sanity, you just keep it that way. There are some cute

things about Second Life. I wrote how Second Life works, all right, and it's there are some people who are legitimately using that as a way to make friends and and have fun conversations and chat about all sorts of stuff, and they enjoy being able to modify their avatars, and so I guess they're the ones who are chalking up these kill of one hours because they're the ones who are going back over and over again. But then there's everything else, and there's a lot of everything else. Yeah,

and everything else. I mean like some pretty hardcore porn stuff. Yeah it's special, right. I didn't write too much about that in the article, but yeah, wow yeah, okay, alright, so moving on, So do you have anything else about electricity? No? No, Um, I just wish that we were able to deliver a hard and definitive numbers, and this is absolutely it. But

I feel the same way. So if there are any research institutions out there who wish to grant us three to four hundred thousand dollars to do a study, we will happily accept that money and we will see you in our research center, which happens to be on the big Island of Hawaii. Um. It's just that was the best place we could put it. So if you want to give us that, we will report back to you in ten maybe fifteen years. Okay, then sounds good to me.

So I guess that just brings us round two. Listener man, Oh no, today's listener mail comes from Chad from Montana, a Big Sky. Hello. First off, love the podcast. You guys do an excellent job of explaining things in a way easily understood. You make tech stuff interesting and not overwhelming.

Thanks you're welcome. I have a question for you. I don't really recall hearing too much about bits and PCs until Windows Vista came out, and it seemed like it was front and center, and how software hardware wouldn't be compatible anymore because one was on a thirty two bit

and the other was on the sixty four bit. I'll be honest, I really don't even understand what they mean when they say thirty two bit or sixty four bit, And when I tried to ask a computer salesman what it meant, he couldn't even give me an easy answer. So this seemed like an excellent question for the fab tech duo. How do bits work? And how is sixty four different from thirty two? Trying to fit it into the how theme? If possible, please address Mac users and

also snow Leopard and Windows XP Vista and seven. Thanks well, Chad. Here's the nice thing is, once you understand how this works, it applies to everything across the board. It's it's not something where it's different for Windows versus snow Leopard versus anything else. So let's start with bits. Now. Bits are your basic information units, right, and they come in two varieties, zero and one, unless you're using a quantum computer, in which case it can be both zero and one and

everything in between. That's a cubit uh. And we'll talk about quantum computers in some future podcast when both of us are ready to tackle quantum physics. But your basic bit is a zero or one. You can think of it like a switch, you know, it's like on or off. And then if you put two bits together, then suddenly you can have up to four integers represented. You have a quarter ha ha ha, two bits, nice shaven a haircut. So but you can have up to you can represent

up to four different states using two bits. And by that I mean you could have a zero, zero, zero, one, one zero or one one. And so every single time you add a bit, you increase the number of integers you can represent. Um and engine roll any for any number of bits, it results in two to that number in integers. So if you were to use two bits,

maybe two to the power of two, which is four. Right. Uh, Now, thirty two bit processor can handle up to thirty two zeros and ones all strung together in various states, and that results in oh, four point three billion integers. That's how many different integers you can have with a thirty two bit system. Yeah, but a sixty four bit system, now sixty four bit, that results in one point eight times ten to the nineteenth power integers, which I looked

this up is eighteen quintillion integers. Now, if you're wondering what a quintillion is, it goes billion, trillion, quadrillion, quintillion, So it's lots. Now, when we're talking about a thirty two bit system or sixty four bit system, we're really talking about the processor here, all right. So your processor, what it's doing is it has and it has a set of instructions that it has to uh complete upon a set of data. Now the thirty two bit or sixty four bit that refers to the amount the kind

of data that's coming into the processor. If it's a thirty two bit processor, that means it can accept digits of up to the thirty two bits. It can't accept anything more than that because it's just not capable of doing it. That's it's beyond its processing power. So a sixty four bit processor can process sixty four bit information. So if you have a thirty two bit processor, you cannot use an operating system that is the sixty four bit variety because it is asking too much of your processor.

Your processor can't handle that data flow, not normally. No, and why would you want you know, why is this important? What's the thirty two bit and sixty four bit? What's the deal? It becomes important when you're using really high end applications, things like three D modeling, video process, video processing. Maybe some really really heavy audio processing too, that can also do it. But video processing really because you're talking about rendering and things like that. Um, if you are

doing cryptography, cryptography involves factoring very large numbers. So the more the larger amount of information your processor can handle, the more complex you can make your your cryptography programs so that it is very difficult to crack. So essentially it can handle more calculations and some more advanced you're doing more advanced computing, you can handle word right, it can do bigger calculations in the same amount of time that a weaker processor would do smaller calculations. I guess

you could think of it that way. Okay, So yeah, it's not that the thirty two bit processor couldn't run the same applications. It probably could, it just would be much slower. So I hope that answers your question. Chad, Uh, we thought we'd take a stab at it. At any rate, If any of you have any questions, you can send us email at our email address, which is tech stuff

at how stuff works dot com. And if you want to learn more about bits and bytes or electricity or the internet, hey, guess what we have articles on that. They're all at how stuffports dot com. It is an awesome website and you should all go and visit and click on lots and lots of articles written by Johnson Strickland. So until then, I guess we will talk to you again really soon. For more on this and thousands of

other topics, does at how stuff works dot com. And be sure to check out the new tech stuff blog now on the house stuff works homepage. Brought to you by the reinvented two thousand twelve camera. It's ready, are you