The full report is here: https://groenlinks.nl/sites/groenlinks/files/2021-09/CE_Delf...

I don't believe their numbers. They say that data use over wi-fi consumes 150 watt-hours per gigabyte transferred, or 220 watt-hours per gigabyte when including the energy used in data centers.

Downloading Call of Duty: Black Ops Cold War, at 250 GB [1], would consume 37.5 to 55 kilowatt hours depending on whether you use the lower or upper figures. That's more energy than you need to fully charge a Nissan Leaf. Watching 8k video on YouTube can transfer up to 22.5 GB per hour [2]. At 220 watt-hours per GB transferred that implies 4,950 watts of power -- comparable to running a large air conditioner or an electric oven. If data transfer really had these huge power demands, you'd see electrical brownouts whenever a hot new streaming series debuts.

When I've investigated reports like this before, the flaw of the methodology is usually something like:

1) Take a well-sourced megabits-per-watt study of networking equipment that somebody published long ago -- say, in 2003.

2) Measure today's data consumption.

3) Multiply the 2022 data consumption number by the 2003 energy intensity number. Get a huge result.

Figuring out what happened in step 1 can be difficult because of indirect citations. A current paper will cite "Davy, 2018" but you look at the Davy references and it's getting numbers from "Smith, 2012" while Smith sources from "Chu, 2008" and Chu finally references the person who measured it: "Graham, 2003."

[1] https://gamerant.com/pc-games-file-size-hd-space-biggest-hug...

[2] https://www.androidauthority.com/how-much-data-does-youtube-...

Yeah, many things in the report seems a bit off. Like they assume everyone uses the same amount of data regardless of what kind of network they're on. So since 5% of users are on 2G networks, they assume 5% of total data is transferred over 2G (which is ~400x less efficient per byte than 5G).

> The EU average energy consumption of data traffic depends on the distribution of mobile networks used for cellular data traffic. The Ericsson mobility report also provides information on the type of smartphone subscription. This data is summarised in Table 3. We make the assumption that the distribution of the total data-use over the different mobile networks is the same.

According to their numbers, 2G would account for nearly two thirds of total mobile energy use, which probably should've made them question their assumptions.

There seems to be a lot here that doesn't pass even a cursory sniff test. Just thinking about the power consumption of the networking equipment involved and doing some quick napkin math should have at least made them suspicious of their results.

You can spot-check a lot of claims with freshman level knowledge of the natural sciences, unit conversion, and arithmetic. But this combination of simple, powerful skills is less common than I would like.

Out of curiosity, I followed the wifi citation [1], and at a first reference it comes from the 1byte model on the page, which gives a simple Wh/byte figure. Which in turn appears be derived from the 5th spreadsheet by taking the total yearly energy consumption of devices that use wifi and dividing it by the amount of data said devices use over the year.

Which I have absolutely zero idea what use that ratio even could have, let alone be important enough for an entire spreadsheet. But it certainly is complete garbage for calculating the energy cost of data transfers.

(well, infrastructure kWh/GB calculated that way could be interesting, but still utterly useless for calculating a marginal energy cost)

[1] https://theshiftproject.org/wp-content/uploads/2019/10/Lean-...

> Downloading Call of Duty: Black Ops Cold War, at 250 GB [1]

The rest of your comment is spot on but in the spirit of similar accuracy it's worth noting that that 250GB on the drive isn't how much is actually transferred.

That said, while I'm sure energy improvements have happened and those numbers aren't correct I do think there's something to consider in the full equation of sheer number of pieces of equipment between you and the server.

If we assumed 1wh per device per GB, that could sometimes be 10-20wh just for routers on your path. That doesn't include switches at the destination, server energy usage, etc. I have no idea what the number actually is on a per GB basis but just using that sample assumption as a way of illustrating that it could add up quite quickly. I suspect in reality it's nowhere near 1wh per device per GB transferred for routers, as even reasonably efficient consumer routers are 5-15W devices.

You also need to take into account the fact that most of this equipment is always going to be turned on and processing some non trivial level of load whether you're sending requests or not. There's generally going to be a difference in power consumption between different levels of load above idle state, but an n% increase in bandwidth is almost always going to result in < n% increase in power consumption.

Also, any section of the network that's at full throughput all the time, and just evenly divides that throughput amongst users as it gets more load is not going to consume /any/ more power per GB no matter how many requests you throw at it.

> Also, any section of the network that's at full throughput all the time, and just evenly divides that throughput amongst users as it gets more load is not going to consume /any/ more power per GB no matter how many requests you throw at it.

If it's at full throughput, you can't throw more load at it (well, smaller packets is more load than larger packets for the same throughput).

I meant more at an individual level - like if I use 100gb of bandwidth today, and 200gb of bandwidth tomorrow there's a good chance that the extra power consumption on at least some of the network from my extra bandwidth is zero if what I described in my previous comment is the case.

If everyone attempted to double their bandwidth usage tomorrow, then yes equipment is going to have to be upgraded. (Although even that could be a case of upgrading equipment that provided N throughput and cost Y watt hours before to newer equipment that provides 2N throughput, and still costs Y watt hours)

My friends are surprised when I show them my little raspberry PI on which I have running two social media instances, two search engines, Nextcloud, XMPP, email and some other services for a dozen people. I explain that my services don't have any advertising, blockchain, surveillance, or data mining which massively reduces the amount of resources needed. We are all in the same town which shortens the data transfers. My phone also stays powered for days since it is using primarily my local services and no Goopple. This is a secure, private, and lower carbon solution I hope more people employ.

Do you have any write-up about your setup? I've been interested for a while in self hosting myself at home, but things like security and such are quite confusing and conflicting at times, I find.

Here is a guide I hope helps get you started: https://growyourown.services/ I am using https://yunohost.org, but https://pibox.io/ looks like a cool setup that takes care of some of the more difficult things such as tunneling and backups.

That's amazing! Thanks so much for sharing

I'm not totally convinced tracking/data mining is the key distinction here. I'm going to assume your dozen users are friends/family and 'well behaved'. Global cloud services do not have that luxury. They have to deal with wide ranging usage, sophisticated attacks, bots etc that you simply don't have to deal with.

In short, they have a vast set of requirements you don't have, each one of which adds complexity.

Dont get me wrong, I definitely agree with your message. However I think we need to be clear that a raspberry pi running some services is not the equivalent of cloud services without 'advertising, tracking etc'

Yes it is a different scale with different needs. It is noncommercial for friends and family. It is private cloud services at a small scale with appropriate security and backup measures.

While not reflecting on the carbon footprint, I was surprised to see just how many queries from my phone were blocked by PiHole once I set it up. Every minute, even when sitting there not being used. What an eventual waste of battery, data and processing power.

I do wonder how many of those queries may be repetitions upon the first one being blocked but even still.

What really sucks about the more recent ios versions is that they make it a little counterintuitive to shut off your radios when you don't need them. If you want to actually shut off your wifi or blutooth its in your Settings app, the buttons in the control center don't actually shut things off.

Sometimes I just throw the device in airplaine mode for a while if I am busy or overnight. This is a lot better than low power mode; the device just sips power and can probably be on standby for a week versus hopefully a day with all the services on the device phoning home in the background. I flick it on when I have a chance to actually follow up on text or calls, and they come in a few seconds after.

> What really sucks about the more recent ios versions is that they make it a little counterintuitive to shut off your radios when you don't need them. If you want to actually shut off your wifi or blutooth its in your Settings app, the buttons in the control center don't actually shut things off.

This annoys the absolute hell out of me.

> Sometimes I just throw the device in airplaine mode for a while if I am busy or overnight.

I have made a habit of doing the same when I'm going somewhere remote that doesn't have service, else the phone will thrash the battery to death vainly searching for signal.

The phone looking for signal issue could have been solved when cellphones first came out even. The phone internally knows what cell tower its pinging. You flag ones around the fringes of known no service areas like a natural park. If the user hikes off into the woods, the phone is capable of knowing its going into a no service area based on the id of last cell tower its pinged and the fact that its just lost signal for longer than 30 seconds. At which point it could switch to hunting for signal every 10 minutes or something with better compromises for battery life and realistic expectations given the area.

This annoyed me too, but I worked out a solution to having to go into Settings or airplane mode every time you want to disable wifi: Make a Shortcut for it and make that shortcut an icon on the home screen.

I made a Shortcut icon named All Off that fully turns off WiFi, Data, and Bluetooth with one press. Easy to set up in the Shortcuts app.

And now apps are using encrypted DNS to get around that :-)

Tax carbon and move on. These (inaccurate) calculations and the resulting guilt trips are the real waste of resources.

Isn't the cost of electricity basically that, or do you mean increase the price to where it actually hurts hard to even exist?

Current electricity pricing treats the cost of dumping the waste products (carbon dioxide, and worse) into the atmosphere as an externality.

My work MacBook Pro had been running hot for months. Whenever I check the app power consumption it would list Chrome, but I felt that was to be expected. Eventually I got curious and looked at the Task Manager within Chrome to see which of the many tabs was actually using a lot of CPU/memory/energy. Turns out this niche Scottish ancestry website, with a mostly-text layout and a 1990s advertising banner was hogging almost an entire core, constantly, in the background. And this had been going on for months!

Every web developer and designer should be forced to experience their site on a low-power Chromebook or rPi before they can launch it.

In an ideal world, web browsers would enforce some reasonable runtime memory cap (let's say 5MB for now) and CPU limit on websites, and users must give permission for the site to break the cap.

Or perhaps a search engine that establishes some reasonable resource caps and de-ranks websites that break the cap.

I think this is a great idea.

At the same time, a lot of pages are still broken in Firefox since they introduced tracking protection 2-3 years ago.

So I don't see devs fixing, or users caring enough about this issue to cause a change in devs attitude.

Lot of web devs know their website is slow but can't do anything about it, they don't choose where to spend time, the manager do.

Maybe we can force Google to do this through ESG activism.

So I wonder how much that text-mostly website consume without the Ad banner (ie, content blocked).

This reminds me of the obscure campaign to change the Google homepage background to black to save electricity. There are much higher value things to spend our time and attention on.

It’s not the biggest waste in the world, but it’s a particularly galling one. “You want to spy on me, use my data cap, use my battery, and generate CO2? No thanks.”

Black is a high power state for LCDs.

Not for OLEDs. I just tested my LG C1, 64 watt difference between full screen black #0000000 and full screen white #FFFFFF at 120hz

The campaign referenced by GP is very old. It dates from before OLEDs. LCDs were the most common screen type and CRTs second most common at the time.

How much does it draw in total before / after? Curious how much of a percentage change this is

228w - total system draw with LG C1 off/standby

270w - #000000 fullscreen

332w - #FFFFFF fullscreen

270w-228w = 42w #000000 fullscreen power draw

332w-228w = 104w #FFFFFF fullscreen power draw

( 104w - 42w ) / 42w = 2.4761

so ~248% more power usage from the screen for full #FFFFFF.

One thing to note is I have the OLED pixel brightness set to 50/100 and the brightness set to 50/100. 2.2 Gamma, black level: full. The LG C1 has some auto brightness baked into the firmware and it fluctuates, I tried to measure it at full brightness but as you can see today it's at 62w over #000000 compared to before so there is some deviation. However, I think it's safe to say that darker colors do actually save substantial power on OLED displays. Plus the black levels really are just amazing. Black is actually black and not a weird grayish with backlight shining through like TN/IPS LCDs.

edit: formatting.

What was the difference?

64 watts?

The website allows you to load images individually by clicking on them, it goes well with the title :)

Indeed! I just ran it through the inspector, it only transfers 257kb in the initial load! Impressive!

The images are super tight, too.

Buuut, it loads the "Two girls sitting on couch" image even though it doesn't display it by default.

I cannot imagine the sheer amount of energy wasted by a whole bunch of NPM packages running on my phone...

I, too, struggle to imagine "infinitesimal".

Maybe there should be a carbon tax on an app's typical resource usage once they have more than a million users.

Microsoft, for example, has how many hundreds of millions of Teams users? If anybody has the knowledge and resources to build an efficient application, it's Microsoft. But instead, to save pennies per user, they build with Electron and waste a huge amount of energy.

In all seriousness - there are already taxes on exactly this in many countries, and even the US has considered carbon taxes on this.

The big secret? Electricity costs money, and is taxed.

Basically - why fucking bother with trying to measure and assess these applications when there are already measurements on the root consumption - electric power.

Long story short - it's a pretty useless thing to be focused on - it costs well under $5 to charge a smartphone for a year's worth of usage. There are simply much, MUCH more impactful changes we could be focused on.

This is a waste of time.

And how many people have or will have handsets in the near future? If it is 3 billion people and we can reduce the power consumption by 25%, it would be a lot. Reduction in data center usage of power is the other part. $5 a year of electricity per phone sounds low, but maybe you can share where you got that number?

> If it is 3 billion people and we can reduce the power consumption by 25%, it would be a lot.

Honestly - not really. Yes, it's large in the abstract sense that it's a very tiny number multiplied by a large number, so it appears large compared to the original tiny number - but in terms of practical impact, it's SO VERY FAR down the list of approaches to reducing climate change.

A year's worth of smartphone usage will generate about ~5 lbs of CO2. Driving a car ~5 miles (at an average of ~21mpg, or 411 grams CO2/mile) will generate the same CO2.

The average commute in the US is 41 miles round trip - so every day, the average commuter burns 8 years worth of smartphone usage in a CO2 budget by driving their car.

This is getting really back of the napkins-esque (since not all people work/commute) but if we just assume that in general the US population is driving - at ~330 million people, times 8... we get 2.63 billion.

Basically - each day, driving in the US is generating close to the same CO2 that all 3 billion of those smartphone users will generate in a YEAR.

Looks like roughly 1/3 of US pop actually commutes, so we'll be generous and say that every 3 days, we blow through the yearly carbon generation for ALL 3 billion of those smartphone users.

It's just a really, really tiny number compared to other initiatives. You're MUCH better off tackling basically anything else. There is SO much better low hanging fruit.

A week long "don't drive" mandate in the US earns the whole world years of smart phone usage.

Go argue against "back to the office" policies instead - you'll do far more good.

$5/year is an extremely high estimate.

A phone with a 5000mAh battery (more than any iPhone) takes about 20 watt hours of power to charge.

Charging it from empty to full each day would use about 7.3 kWh per year.

Electricity costs about 15c per kWh.

Correct - I used $5 because I KNEW it would cost less even in countries with very high electric costs, and for power hungry phones.

The last good data I can find (2019) puts the average right around $0.90/year for US based phone users.

Another bonus of turning off these advertising activities on phones and networks is that people are not as manipulated into buying more crap they don't need. Mass consumerism is a major cause of CO2 production.

I wonder if the respondents understood the consequences of disabling advertising.

Like was the question “do you want to pay to use Kwazy Kupcakes, lose access to Kwazy Kupcakes, or allow Kwazy Kupcakes to show you ads?” or was it just “Should Kwazy Kupcakes be allowed to spy on you?”

Because I have a hunch it would be far fewer than 60% if they understood the consequences of destroying the revenue model of their shitty free to play games they use to numb their brains at every available opportunity.

I also bet that using less performance oriented languages like Php, Ruby, Python in the datacenter instead of Rust and C++ also has a huge Carbon footprint.

Honestly, the only way forward to really fight climate change is to decouple carbon from power generation.

Everything else is just wasting time re-arranging deck chairs on the Titanic.

That means we need to be all in on zero emissions power generation (wind, solar, hydro, nuclear) as well as conversion of combustion engines to electric (see electric cars).

(see electric cars^h^h^h^hbikes).

There, fixed it for you.

Given how many cars exist globally, I do wonder where 100's of Kgs of batteries per car is going to come from.

A bike requires much less and indeed most people could get by with one so I'd like to see more of this too. But it doesn't cater to the lowest common denominator of population use cases, it doesn't signal wealth, and convenience is king. So I don't think it will come about.

Datacenter energy usage is irrelevant to our global situation. A large cloud datacenter uses a similar amount of power as 1 large aircraft. All U.S. datacenters combined consumed about 1% of U.S. electric power, meaning they constitute much less than 1% of all energy demand. And despite the fact that cloud capacity has exploded in the last 10 years, electric power consumption of datacenters has barely increased, because the cloud is so dramatically more efficient than other datacenters.

source? Bitcoin network allegedly uses as much electricity as Argentina based on headlines (and... those are headlines). Yes that's global, but still.

I'm not sure of the source (as I'm not the parent), but I wouldn't classify crypto farms as "datacenters". They have very low needs in terms of bandwidth and reliability. Many of them are barns on consumer ISP connections filled with bare mobos sitting on pieces of cardboard, or other similar arrangements that are presumably OK given the economics of crypto, but wouldn't be sellable even as cloud compute for most purposes in the current market (it might be OK for batch-type stuff that doesn't involve tons of transfer -- some kinds of scientific workloads like Folding@Home maybe?).

How can all US data centers make up 1% of US electrical power and at the same time only use the same amount of energy as 1 large aircraft?

Each.

Please take care with using the word "Carbon Footprint", it was invented by the advertisement company Ogilvy for a BP public relation disinformation campaign to draw away responsibility from companies and hand it over to individuals in 2005. - Everytime the term is used, someone thinks he can make a difference, while he cannot as long as the fossils are still digged up to,the meat industry grows and grows, concrete is used on bigger and bigger scale each year and so on.

Other than that, good article. Data Transmission should definitely be fueled with low carbon emitting energy sources or avoided if possible. But I'm not sure if 30x 8GB DVDs plus packaging and logistics would be a better way to distribute Call of Duty. Why do all the fun things have a downside? Meh

They say that this is 3-8 Metric tons of carbon per year. For comparison a 747 flying from New York to London one way emits 500 Metric tons of carbon. Europe is emitting around 2 BILLION Metric tons of carbon per year.

I personally don't think this is an effective use of resources if our goal is to stop climate change.

They're measuring in Mega Tonnes, not metric tons.

GP needs to fix their comment, it's completely inaccurate and presents to people who don't read the article an entirely wrong conclusion.

I agree that we should focus on the major uses of power, but the trajectory of computing and data center usage is accelerating quickly. We have the opportunity to get ahead of it rather than try to rip out and replace later as we are doing with power plants and transportation, all very costly.

747 maximum takeoff weight is 320 tons. Even assuming it is entirely made of carbon and it is all burnt (helluva flight for the passengers!), it won't make another 180 tons of carbon appear out of nowhere.

I am not even close to trusting your statement. Dial back the hyperbole.

I installed Lineage os on my phone and now my battery lasts four days (it lasted only one day before)

A nice comparison would be to bulk mail also known as junk mail in some places. A lot of paper and ink.

"Unwanted" only by the user.

Carbon-live.