I worked on Android power perf (frequency scaling in particular) for the last 18 months and I'm not aware of any linkage with the battery state, except for the Battery Saver mode. So yeah same perf, just shorter life. There could be something in vendor Android builds I don't know about, but it isn't there in AOSP and almost certainly isn't in Pixel/Nexus.

I vaguely remember the dram and camera clock rates being reduced when the battery SoC went under 40% in the nexus 5, and then when users complained of early shutdowns when using the camera with HDR mode, an update changed it to 60%.

I think I even commented on the bugreport that it would be best to do the frequency scaling based on interrupts from the voltage regulators on board, but that piping all the necessary drivers/events to do that with low enough latency to not cause an under-voltage reset was far more work than was feasible.

Thats also why 60fps video dropped frames at the start of the video if you were recording video with less than 60% battery.

Anyone care to dig out the relevant commit...?

I have a Nexus 6P and its battery is going bad. It just shuts down when it reaches 20-30%. Kind of annoying because I never know quite how much battery life I actually have remaining.

I have a similar situation.

My Android (1st Gen Moto G, running Lineage OS, about 3 years old) shuts down once the battery level hits 15%. You can actually see it change from 15% to 1% for one brief moment before it shuts down.

Sure, the battery is old, but this behavior is unlike the behavior of any other OS/device I've seen. I've seen some laptops with seriously messed up batteries and as far as I can remember, the OS reports a smooth discharge profile without steep drops and without and strange "surprises".

I think Android is probably not taking battery health into account when it reports remaining capacity, but that is only a guess.

Can't it be a problem with the fuel gauge inacurately measuring the battery capacity? I have seen it only on one "cheap" Android device and I have been using Androids for a long time. Anyway, easily removable batteries are the way to go.

My old MacBook Pro battery got so bad that it would jump down and back up 20% randomly.

I replaced the battery on my 6P, and it's like a new phone.

It is a bit tricky, but well worth it.

Same here, the phone is otherwise fine. I don't really want to go to a Pixel 2XL because I like having a headphone jack.

Effectively the same, but it just quickly degrades from 60-0 for me (like 20 mins of just reddit/hn).

100-60 is still decent.

Same here. I have to carry around an extra battery booster. :-/

That's really strange. Because most/all modern systems build statistics about your battery as it charges and drains, to predict the life left. Without those statistics they'd have no idea how long it will last because it changes over the life of the battery. In Linux you can even see pretty graphs showing "confidence" for each percentage of battery vs "how confident" Linux is of remaining power.

https://www.makeuseof.com/wp-content/uploads/2012/02/power_s...

My Samsung Android phone shuts down at 5-7% battery life. And the battery dies much faster these days. But it's never "inaccurate." It doesn't jump from 50% to 2% over a course of five minutes.

Other than something else being wrong with your OS, my only guess there would be the battery got bad so quickly the OS hasn't adapted to the new life cycle yet by removing enough of the old datapoints.

Most Android devices have rather severe issues with flash degradation (due to use of cheaper chips, controllers and connection... cheap MMC controllers are still very standard) which makes them practically unusable after a few years.

Most? Absolute nonsense. The Nexus 7 have a small issue that was resolved when Google added TRIM support for Android. That was years ago.

I have a long history with almost every Android device. The device doesn't get slower, the battery life just gets terrible.

My Nexus 6P's performance has degraded considerably over the past couple of years. It's actually quite laggy now. A few of my friends report the same with their 6Ps. To be fair I haven't tried factory resetting it in a while, although I shouldn't need to do that in the first place..

The 2012 Nexus 7 is an outlier because it has especially bad flash controller - by 2015 most of the devices were unusable due to degraded flash. TRIM helped very little.

Other devices suffer from similar effects - not as bad, but most devices certanly lose quite a bit of I/O performance (especially writes) over time which makes OS run noticably slower.

Flash degradation literally only affects writes. Android has had TRIM since 2013. Writes in no universe make the OS run "noticeably slower" (or lead to "lag"), even if your claim was proven.

There are literally billions of Android devices. Benchmarking file system performance is trivial. Ergo, where are the benchmarks supporting your broad claims? This should be incredibly easy to demonstrate. We do know that it happened pre TRIM because people had factory new and after use benchmarks, and could show the decline.

I've owned two low cost Moto G phones and never had this issue.

Is this true for flagship Android devices too?

Depends really, a lot of them nowdays use better flash controllers and technology (e.g. UFS) but only time can tell how they cope with aging. Pixels seem to be doing quite well.

I had a Nexus 5 before which I replaced of about 2 years after buying it. Can definitely say performance improved after that (for a short while though as it was a cheap ebay battery).

Electric cars have fundamentally the same problem and Tesla decided to attack it by hiding some of the battery's capacity when the car is new. While the car might report to the user it is 100% full, it might only be actually 90% full. The battery can therefore lose a decent percent of its capacity before the user will ever notice the difference. This decision was made because cars obviously have a longer ownership cycle than phones and a drastic drop in range as a car ages is a huge flaw. But I do wonder if any phone companies might start to rethink this decision now that we are seeing some people use the same smartphone for 3-5 years.

There is absolutely no evidence that Tesla is doing what you say. In fact, if they were, this curve https://images.hgmsites.net/med/tesla-model-s-mileage-vs-rem... would stay at 100% for a while instead of immediately dropping.

Yep, it looks like you are right. I was passing along second hand information that I never bothered to verify. After some Googling I learned apparently Tesla only hides capacity at the low end to prevent a user from being stranded. The default settings also prevent charging to 100%, but that is fully transparent to the user.

Thank you for looking! I've seen this "Tesla sandbags at the top" repeatedly on HN but you're the first person to respond when challenged.

BTW the car won't actually let you drive at low charge, so yes, you can be stranded because the car refuses to damage its battery. That's why it comes with a towing service.

Li-ion batteries also last longer when kept between 30% and 90%. Several car manufacturers don't let you completely charge or expend your battery for this reason.

Tesla, for example, encourages owners to not charge beyond 90%, and refuses you let you drive below some number, which is around 20%, that it hides from the driver.

I wonder how much driving style effects that too, racing around vs Driving Ms. Daisy?

Discharging lithium batteries at a high C rate (C = Instantaneous Draw / Capacity) causes a different form of degraded life than storage at a high state of charge. And the major killer of lithium batteries is heat, which Tesla are known to manage well.

The biggest killer, though, is storage with the graphite electrode more than 50% lithiated - there is a gigantic step-up in capacity fade when batteries are stored with the State of Charge over around 90%, which Tesla are very smart to avoid.

http://jes.ecsdl.org/content/163/9/A1872.full

Nexus 6P the phone will die randomly at random percentages. Sometimes you can turn it back on hours later and it'll last thirty minutes. Sometimes you have to "charge" it a little bit first. This issue happens at all percentages and I've had it for over 8 months now.