• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Battery gate - the breakdown

GC2:CS

Junior Member
Jul 6, 2018
4
0
11
Just feeling so bored these days...

Hello comunity. If you ever heard that phones have some problems with batteries, you know about famous Apple “battery gate”. And i have some “testing” of this feature my iPhone 7 unit. I never saw any kind of data to examine this "phenomenon" in detail so I made some bits up.
At the time of the test I had a relativelly fresh iPhone 7 unit. Personally i have not experienced any performance degradation or reduction in battery life compared to a new unit.. The test was done around the time the whole afair bursted out (january 2018). After the battery life update (march 2018) the battery health said “97%” battery capacity. And the throtling behavior was turned off by default.

In other words the throtling was up and running on a pretty much pristine iphone 7 unit with 97% battery health left.

So aggain before the battery health update, the throttling mechanism for the CPU was up and running all the time, even if the phone was brand new. Was it ? People complained about poor performance and some low benchmark scores, that was not the case for my half-a-year-then-old unit. How to test the throttling ? Well if it doesn’t affect day to day perf, what happens when the battery has low charge ? How does this compare say to low power mode, which also cuts on perf ?

For reference I run Geekbench and GFX bench. Nothing scientific, but quick. Geekbench 3 reports 1.02 Ghz for iPhone 7 in low power mode, compared to 2.34 Ghz normally. Quite a big cut, but i honestly have problems noticing that in day to day usage.

So for the test. I made a few repeated runs of Geekbench 3, while allowing the phone to cool down between them. I started with relatively low charge and written down the performance as the battery was dropping. I choose geek bench 3, because it executes quicker and thus gives more datapoints. I also forced display brightness low to help preserve more power. Results:


Snímka obrazovky 2020-04-20 o 19.21.22.png


So we can see perf gradually droping off. Each datapoint represents a single run which is more or less a fixed amount of workload. Notice the dramatic slow in battery drain as perf and CPU clockspeed is going down. There is prety much no effect until about 5% battery left. The low power mode score is about 3500 points. This was also the first time I really noticed the throttling as an user - the test took progressively more to load. The CPU throtles to about 1/3 the perf and then shuts down.

Now to the GPU. I used GFX bench and the heaviest test avalable at that time - 1440p manhattan 3.1 offscreen. The phone manages 24 fps at first run when cool, 17 fps with low power mode and about 13fps thermally sustained. Phone was left to cool down between the runs.



Snímka obrazovky 2020-04-20 o 19.23.20.png

Note this time i plotted against the number of runs for a different perspective on performance deficit. Battery power in blue, perf in green. This time the number of runs is not a fixed workload, but a fixed time - 1 minute. We have a behaviour practically identical to the CPU. Throttling starts at 5%, drops sharply, the battery % lasts longer and phone shuts down at about 1/3 the full performance. The phone was also heating much less as the battery went down. This might be because CPU cannot feed enough frames as it throttles down.
For low power mode the behaviour is about the same, only the initial perf. is lower, further drop happens later and is far less noticeable before shutdown.

So what is the conclusion. We have a smooth throttle down, based on battery voltage down to about 1/3 perf, this (was happening in 2018) happens for all the phones regardless of age if their battery cannot supply enough volts. For new phones this practically means automatic low power mode as battery level drops. Have a worn out or frozen battery and the threshold will probably work its way up to 100% and reduce the perf until the phone won’t even turn on with fully charged battery after 5 years of heavy use.

So at last let's look at how batteries age according to Apple for 2 iPhone 7 units i had access to.

Snímka obrazovky 2020-04-20 o 19.26.22.png


First this is subjective to the level of what cell the phone has got, how much it was used etc. I am not a heavy user, the phone could last me 2 days sometimes, 1.3 days seems like median to me (new phone). I use fast charging and don't mind letting the phone charge overnight.
Then we see the first unit which had older software is just flying around with the estimates, the newer software being much smoother and consistent on the second unit.
The first unit was at 90% battery health after almost 16 months of use.
We also know that every new generation iPhone somehow got “better” with this power management and is supposed to retain performance for longer according to Apple.

Now I would like to know how a worn out battery would perform. But the only phone I have with me and it’s old too is an iPhone 7 unit with battery health at 86%. The throttling still has not turned on, there was no unexpected shutdown. The perf is great for a 2016 mobile. But the shorter battery life is definitely noticeable.

Maybe I should put it into a freezer to force a restart and throw up some tables ? :D

Have a nice day for now, until i get bored again.
 

ASK THE COMMUNITY