The energy consumption of avatars

.
Are virtual worlds environmentally sustainable? Based on such thoughts, Nicholas Carr wrote a blog post some three years ago, in December 2006, about how much power we use when we use virtual worlds. It provoked strong reactions, not the least because the title of his text was "Avatars consume as much electricity as Brazilians". Still three years laters, texts (such as this one :-) about avatars, eletricity, climate impact and Brazilians show up like a jack-in-the-box. I will here go through Nick’s line of reasoning and the criticism he encountered before I go on and analyze the ways in which we can think about these issues. I recently wrote about the energy footprint of Google searches and will eventually come around and write about the power consumption and carbon footprint of PCs and data centers. Computer servers consume one percent or so of the world electricity supply. That might not sound like much, but their power consumption grows by 15-20% per year (which is equivalent to a fivefold increase in 10 years). 



At the time when Nick Carr posted is question, the virtual world Second Life was visited by somewhere between 10 000 and 15 000 avatars at any one time. To run it all, no less than 4000 servers were required. In the absence of actual figures as to the electricity consumption of the company that runs Second Life - Linden Lab - Nick made a few assumptions: 
- Each server in Linden Lab’s data center burns through 200 watts and then uses an additional 50 watts to cool the data center. 
- Every home computer that is connected to Second Life uses 120 watts. 



This would mean that the 4 000 servers in question use (4 000 servers) x (250 watts) x (24 hours) = 24 000 kWh (kilowatt hours) each day. Additionally, the power consumption of all home computers is (12 500 PCs on average) x (120 watts) x (24 hours) = 36 000 kWh each day. Altogether these computers and servers would thus use 60 000 kWh per day and if we divide this electricity consumption between 12 500 avatars, each of them would use 4.8 kWh for each 24 hours of existence in the game/virtual world Second Life. 



How much is 4.8 kWh per day then? Well, it adds up over the days and months and becomes 1750 kWh per year which is comparable with the electricity consumption per capita in Brazil (according to the 2003 data that Nick had access to).



In the ensuing discussion, Nick was quickly corrected by a person who was employed by Linden Lab. Previously, each computer server ran a "region" in the game but now, a server may run up to four "regions". The correct figures for Linden Lab’s electricity consumption is therefore (1 000 servers) x (225 watts) x (24 hours) = 5 400 kWh per day - that is, less than 1/4 of the original estimate (which, however was based on a fuzzy statement by the CEO of Linden Lab). The new figures gives that an avatar consumes approximately 1 200 kWh per year instead of 1 750 kWh, and that Linden Lab’s servers account for a relatively small part of that power consumption while the home computers account for more than 85% of the total power consumption. 



Almost six months later (May 2007), Nick is once more corrected when a new, better-informed (?) employee from Linden Lab presents new figures. To begin with, the average number of avatars are now 30 000, and the number of servers has risen to 2 000. In addition, both servers and home PCs draw significantly more power when the run Second Life. Now, Linden Lab’s power consumption is instead (2 000 servers) x (500 watts) x (24 hours) = 24 000 kWh. The rule of thumb is that for every watt that a server uses, the same amount of energy is needed to cool the data center where the server is housed. Power consumption at home is estimated to be (30 000 computers) x (250 watts) x (24 hours) = 180 000 kWh per day. In total, these 204 000 kWh divided into 30 000 avatars becomes 6.8 kWh per day. That is equivalent to 2 500 kWh per year and the home computer accounts for almost 90% of the total power consumption. Latvia, Romania and Argentina are a few countries that had a power consumption in the neighborhood of 2 500 kWh per capita in 2005. In Sweden, we used more than 15 000 kWh per person in 2005.



Taking all of this conflict (and constantly changing) information into account, what conculsions can be drawn so far?



- The Internet changes constantly. To get current figures is like chasing a moving target. What are the figures for Second Life right now? According to the latest figures (Jan 2010) there are currently 18 million accounts (avatars) registered in Second Life, but only 750 000 of them (5%) log in to Second Life each month. These avatars spent a total of 118 million hours (!) in Second Life during the third quarter of 2009.



- Information about the number of servers and their power consumption varies widely and therefore seems not be that reliable (see above). Power consumption can obviously not have been one of the heavier costs when running virtual worlds - or they would have kept better track of the figures. The same has probably been true also for other companies that rely on data centers such as Google, Flickr, Blizzard etc., but things might be changing now as the energy prices have been marching upwards during the last couple of years.



- A computer at work uses 120-150 watts, but a computer that runs Second Life (or World of Warcraft or any other computer games) can use up to twice as much power as these applications make use of your computer's capabilities to the max. Data center use a lot of power, but you home computer that utilizes these services draw a lot more and get less work (computer cycles) done per unit of energy used.



- It is difficult to determine the usefulness (or damage) of using virtual worlds. On the one hand, you use a lot less energy (and generate considerably less pollution) if you cancel a trip and instead meet in a virtual world. But a computer uses a lot of electricity - if the option is an electricity-free activity (take a walk, talk to a neighbor, help your children do their homework). 



- Ideas are hard to kill. Although Nick’s figures were refuted and modified immediately, the "meme" about Second Life and the electricity consumption of Brazilians remains alive and pops up now and then to the chagrin of some.



The main objections raised against Nick’s argument above was that no real person is connected to Second Life 24 hours a day and that Second Life actually had 700 000 "active user" (whatever that means) at the time. So the power consumption of each person who used Second Life would have been just a 50th of Nick’s original calculation. Furthermore, any computer that is used for 24 hours a day 365 days per year uses more energy than the average Brazilians whatever that computer is used for (playing Second Life or doing something entirely different). 



Both ways of looking at this problem is correct, but these different perspectives choses to focus on slightly different things. Any individual physical person who plays Second Life did that for less than an hour a day on average and thus uses a moderate amount of energy. But each avatar in Second Life has the same (or higher) power consumption (per hour, per day or per year) as many people on earth have. 



I think Nicholas perspective is interesting, not the least because some information technology pundits sometimes tend to completely ignore that computers are physical objects that have required resources (raw materials, energy) for their manufacture, that consume electricity throughout their lifetime, and that one day will be scrapped/recycled. Computers obviously have an ecological footprint and the size of that footprint should naturally be explored further. This is a topic I will come back to later.

.