Android - 2G versus 3G: Does it really save battery?

Fast answer:

Nes. Yo. Well, it all depends... on how you use your device.

Detailed answer

Self-Experiment

I just did a self-experiment to find out. Until now, I had all my devices fixed to 2G -- as that should "save juice", like the question suggests. So for 4 days now, I have 3G enabled. Surprisingly: No difference which could not be explained by measurement errors or slightly different situations (after all, I had a. no lab conditions and b. just tested with a single device). Humm. Time to go for some facts, and do some calculations.

Hard data

I already posted some hard data on battery consumptions of components as answer to the question What can I do to increase battery life on my Android device?, more than half a year ago. So now I picked the relevant values, made an average over the two devices, and -- I admit -- slightly adjusted things to reduce differences:

I took the best GSM (2G) speed possible to make 2G seem faster
I took worse conditions to "slow down" UMTS (3G) to conditions which should be met even in areas where it was not improved to the "the latest cutting-edge" (oh, speaking of "edge": 2G assumes EDGE here).

And now I've got some surprising answers: Whether 2G saves more juice than 3G, or vice versa, depends on how you use your device. I created 4 different "usage-groups", and added a 5^th special case at the end. Here are my results:

Few phone minutes, few data

Yeah -- that's me: 150..300 minutes talk per month, that makes not more than an average of 5..10 min per day. I have a 50 MB data flat, and hardly ever use it up (which makes for ~ 1.5 MB/day). So I'm mostly left with "standby":

~UMTS StdBy 15 mW
~GSM StdBy 10 mW

Consider that your screen uses 400..800 mW on average when turned on, a difference of 5 mW can safely be ignored. Besides, the little data even compensate -- but I get ahead of myself.

The 24/7 caller with no time for data

...better switches to 2G:

~UMTS Tel 800 mW
~GSM Tel 400 mW

As you can see, when on-call, 3G uses about twice as much juice than 2G would use (and you don't speak faster on 3G -- a call is a call, and will take the same time). So if you spend a lot of time in calls, and have only little data traffic, you're much better off with 2G/GSM.

Few phone minutes, but big data sucker

Here comes the big surprise: This group saves a lot when sticking with 3G/UMTS:

~UMTS XFER 1200 mW
~EDGE XFER 1000 mW

Doesn't look like a huge difference? OK: looking on "how much does it suck per minute" might not make the big deal. So we need a second component:

~UMTS speed dl/ul 7/1 MBit/s (HSDPA/HSUPA)
~EDGE speed dl/ul 200/100 kbit/s

You see I assumed max speed for EDGE, but not the state-of-the-art (20 MBit/s downstream) for UMTS. UMTS coverage might differ between areas, but the 7+ MBit/s should be available everywhere. Now let's look how long the download/upload of 1 MB data takes, and how much juice that corresponds to:

~UMTS 1MB dl/ul 1s/8s ~ 1200/9600 mWs
~EDGE 1MB dl/ul 40s/80s ~ 40.000/80.000 mWs

That's about a factor of 1:7 for upload, and even more than 1:30 for download!

So yeah: This group is definitely better off sticking to 3G!!!

Multi-Tasker: 24/7 on call while using full data

This now very much depends on what "full data" stands for, and if you're really a 24/7 on-call-boy (or girl). No clear answer for this, you must put your own data into the equation and do the calculation. It's rather 50:50 -- as with group #1, you might run a "zero growth" -- or tend to either group 2 or group 3.

Special case: The "hopper"

Especially in rural areas, coverage might be, humm, "slippery". Your device might be permanently toggling between 2G and 3G, always on the hunt for the better signal. Here all the above might become relative, as the toggling itself is drinking the juice. If you are in such a situation: Check to which of the 4 groups above you fit best, and fix your device to either 3G or 2G, depending on.

Btw: A similar situation would be if you've got the idea to toggle between 2G/3G whenever you switch your display off/on. As the toggling requires up to 10s full-power (to find the corresponding stations; 2G and 3G use different frequencies), you wouldn't really "save juice".

Epilog

Yepp, I completely forgot about 4G/LTE. But incidentally. First: I have no 4G/LTE device for a self-experiment. Second: I have no "hard data" on 4G/LTE. So all I could do is guessing. Doing that, OK, OK: It most likely will be the same as above -- just increment the numbers by one :)