How do you measure the energy consumption of a mobile phone?

Mobile phones are really incredible devices. They’re some of the most efficient computers that we have, with a decade-plus of engineering that’s gone into making sure they use as little (battery) power as possible, while simultaneous developments in battery technology have also improved how much we can cram into these devices every day. And yet, because they sip on such tiny amounts, our methods for measuring the power they do use (which still has to be generated in a power plant somewhere) and the resulting GHG emissions have somewhat lagged behind other devices. We have tended to rely on averages and assumptions, extrapolating from testing that has established the typical range of power consumption expected (3-5 watts, usually).

If you’re not on LinkedIn, or subscribed to the SGA Newsletter (fill out the form at the bottom of the page), you might have missed the announcement last week that we’ve released an exciting new component of the SGA GHG Standard - featuring some new guidance on measuring Scope 3, Category 11 ‘Use of sold products’ for mobile devices. I’ve embedded the video below that I recorded for the announcement.

That’s a nice, short overview, but I wanted to use the newsletter to explain a bit more about some of the technical choices we’ve made.

As with other sections of the SGA GHG Standard, there’s multiple methods to choose from, that can help meet the capacity for data collection at games businesses while also showing the direction of travel. There’s basic estimates we can produce based on playertime metrics, and an option for even less data than that.

But the main goal is to get to a point where we are able to measure battery levels at the start and end of a session directly. The difference between the charge level at the start and the end of a game session is essentially the total power consumption attributable to gameplay. Neat, right? It’s quite a simple approach that sidesteps all sorts of mucking about with device models, iOS v Android, different battery capacities of different devices, even the different regional hardware differences for the same kind of phone. Did you know that the same Galaxy S23 sold in the US often doesn’t even have identical internal components as a Galaxy S23 sold in India, for example. What a minefield! But by going straight to the (power) source and measuring it, we simplify things so, so much.

Once we have the power consumption, we can calculate the GHG emissions fairly simply. All we need to know is what region a player is in – say France – and then add in an allowance for power losses due to the conversion from AC>DC power that happens whenever players charge their phones and apply an accurate regional emissions factor (for France last year it was a mere 44.18 grams of CO2 for every kWh used).

Most phone chargers are pretty good these days, but their efficiency is still typically only around 70-75% – that means to charge a battery from 0-100%, you actually end up drawing about 140% of the equivalent battery capacity. That extra is lost in the conversion process (mainly as heat) – though GaN chargers are getting much better, so this figure will improve over time. We can (and will!) adjust this figure for charger losses as more research about average charger efficiency (or GaN charger use) becomes available.

So, how easy is it to get to battery readings? Not too hard, it seems – at least, from a technical standpoint. But there are things to consider on the way, like legal implications and terms of service updates, ensuring users have informed consent, and so on. Also, because we want to see this practice adopted widely, and systematically, that means making it simple to do, like flicking a switch in your mobile game. We have some ideas on that, but I can’t say more just quite yet. But because the SGA GHG Standard is free and open to anyone and everyone, service providers can now use it as a guide for the development of future features, and maybe some enterprising business wants to figure out how to make a plugin or something that collects the data for you. Who knows what people will come up with! Such is the power of an open Standard and the ecosystem of solutions it can enable.

One last thing I wanted to discuss and possibly head off is a certain potential objection to just simply measuring the battery difference and taking that as the figure. A reasonable observer might ask “aren’t you also measuring other processes running on the phone in the background at the same time as the game?” Which is true in a sense – in fact, this is exactly why the GeSI and Carbon Trust ICT Sector guidance (which the GHGProtocol seems to tacitly endorse, hosting a link to it on its own website) suggests that when measuring the power consumption of software a baseline measurement taken of the operating system should be subtracted as well (by measuring what the OS does when nothing else is actively running, for instance).

I think that approach doesn’t work for mobile phones for two reasons. First, a lot of those background services handled by the OS are integral to the game working properly. A mobile OS actively manages resources in a way that keeps a game running smoothly, from cache memory to handling TCP/IP communications, and probably a whole lot more. At least on iOS I don’t think it’s possible to treat the OS as separate, it’s as integral as the game (maybe I’m wrong about that – happy to be corrected by those sufficiently in the know).

The other reason is that, unlike software running on a typical desktop PC, when you run the “baseline” test measuring the power of the phone while idle, you’re going to inevitably be measuring the power used by the screen as well. There’s just no simple way to measure and subtract an iOS/Android background OS energy consumption figure without also subtracting the energy used on the screen as well – and screens eat up a lot of total power.

(Now I’m thinking about it more though – there might be a very convoluted procedure you could set up, where you get a phone, disable absolutely everything, all notifications, wifi, cell connection, turn on airplane mode, make sure nothing at all is running in the background, and leave the screen off… until it runs out of battery, then divide the energy level at the start and end by the time it took… but how do you know when it’s about to turn off without turning on the screen? How do you turn all background processes off? Would the amount of power draw even be material in this state? How long does an iPhone last in airplane mode these days… probably a pretty long time! I don’t have an answer. Phones and their operating systems are just so incredibly complex these days.)

All that said, if you develop a mobile phone game especially: please go check out the SGA GHG Standard components , including the new S3.11 mobile documentation and data input sheet that we’ve released. Feedback is always welcome. We’ve also got plenty more components on the way, based on our research and the input from SGA members and other experts in the field. I am super excited about this work. I think we’re laying out a framework that will help the entire industry to see more clearly than ever before what a roadmap for real GHG emissions reduction looks like for these next crucial few years. The SGA Standard is already starting to clarify where and how different elements of the games industry produce GHG emissions, and what we can do about it. Things like identifying mobile hardware and software combinations that are using more energy than they should be, or finding places in individual games (like the Xbox ‘eco modes’ have done for consoles) to shave off small amounts of energy consumption for our massive gaming audiences. All of these approaches will make specific games drain less power, in turn making players’ batteries last longer, and reducing pressure on the charge cycle. Batteries, after all, are consumable components, making them last even a little bit longer is a win-win for everyone.

Oh, and as a bonus – because it’s not a GTG post without some sort of data or visualisation, here’s the average battery size in phone models released each year, from 2018 to now. I sourced the data for this from this incredible Wikipedia page that someone (or someone’s) have spent a huge amount of work compiling – the power of open source and collective effort, eh? It just goes to show how much these tiny computers we carry around in our pockets have improved in just a few short years.

Thanks for reading GTG! Join us as we work on the SGA GHG Standard. We'd love to have your input as well.