Steam Deck low battery health (% of original capacity) and formatting
I believe a few of us have Steam Decks, thus I wanted to do this kinda public announcement.
TL;DR
- If your Steam Deck reports low battery health (low % of original capacity) drain the battery until it shuts down and then fully charge and check again
- Use your Steam Deck until it shuts down from time to time (say once every 6 months?) to keep your battery level indicator (and remaining time) precise
Long version
I have my Steam Deck since May 2022 and I put certainly over 1000 hours in gaming on it. I would believe if it was even approaching 2000 hours.
Lately I played demanding game and battery was discharging rather fast with remaining time on full charge being under 1:30 hours, which I wasn't used to just a few months ago when it lasted over 2 hours even in demanding games (I limit to 30 fps and I also limit TDP/power). This weekend I jumped into desktop mode and checked the battery life which showed me what I feared - 65% of original capacity.
I went on iFixit page and the price for new one is hefty 95€, but since Steam Deck got me so much enjoyment, I was ready to pay for it - if it was in stock, that is. I'm glad it wasn't!
Since I have a bit of electronics and software background (hobby level), I realized that the charging chip (or whatever keeps the info about state of charge) was running since day 1 on relative data. What I mean: I have never discharged my Steam Deck lower than 10% and most of the times charging anywhere between 30-80%. And since the charging chip likely measures last fully charged capacity (and thus battery health) based on, well, how much it was charged and discharged all over again, it probably skewed its measurements in those three years.
So I went on a "quest" to play until dead. And I was surprised when Steam Deck reached 3% battery and kept running for another hour (ligthweight game) until it was finally dead. Then I fully charged it and voila - battery health 90%!
I have already said how it likely happened, but once more and in short: the charging chip needs to reach both limits, 0% and 100% of battery, from time to time = You have to let it drain fully here and there if you want your battery level indicator (and remaining time) to be precise or if you want to get the real state of your battery.
A bit offtopic: I've had laptop that had 50% of original capacity. I have changed the battery cells inside the battery and let it fully discharge and charge again but the vendor locked the chip from "learning" the real capacity making the new cells useless because the chip still reported 50% thus telling me when I booted it up that the battery has to be changed and also telling me non-relevant remaining time based on this 50% battery health... I'm glad that Steam Deck is capable of re-learning this data and not playing dumb.
Edit: As some of you might know, I've been a professional electronics technician for about a decade now. I don't like pretending to be an authority, but it seems germaine to this conversation, so I thought I'd mention it.
Battery calibration used to be something manufacturers pushed a little more often, but I've rarely seen any mention of it when getting parts shipments for a while now. I think it's partly because it takes a fair amount of time to do correctly, and usually the effects are negligible.
The correct way to calibrate a battery is just a touch more complicated than your description. Ideally it would involve two full charging cycles: you let it discharge until the device shuts itself off, then you let it charge fully without using the device or otherwise letting it discharge in any way. Once at full, you discharge the battery completely again without connecting it to power until the device shuts itself off, then let it fully charge without disturbing it one last time. After that, the battery is calibrated and you can use the device normally.
In my experience, built-in battery health monitoring isn't worth paying much attention to. I've seen too many devices report excellent battery health even though the battery can't hold a charge for more than an hour even after calibration. I get the impression that manufacturers rely too heavily on cycle count rather than discharge rate and calculating the "rocks" (i.e., the percentage of design capacity lost to degradation). They basically have the monitoring count the number of times the battery has discharged the equivalent of a full design capacity, then give a health rating based on an idealized degradation curve.
My hunch is that this is more a technical concern rather than anything to do with RMA; that is, they're more concerned about how much it would cost to put more accurate monitoring componentry in every device than they are about honoring an implied warranty on a battery's design capacity. That kind of sounds cheesy when I put it that way, but it really comes down to the fact that lithium ion batteries are very sensitive to environmental conditions in a way that makes automated health monitoring difficult. They really don't like temperature extremes, and will sometimes have odd reactions to even minor fluctuations in storage temps. Couple that with the fact that accurate health grading requires patient user involvement and it makes sense why reliable health monitoring circuitry is a PitA that OEMs don't want to bother with.
There's a version of letting the battery drain fully that I would not recommend, specifically putting the device to sleep and leaving it to discharge. It has happened to me twice now and the result was a seemingly dead display, which gave me quite a scare.
Details and resolution steps
The device would work when docked, it would display everything and play games, but when undocked, there was a black screen. Fortunately I found resolution steps online—I needed to turn it off completely, then hold VolumeDn and the power button. The diode would blink, and when I came back to it an hour later, the screen was working.Glad I've accidentally been doing those for months now.
I've been accidentally doing that since I first owned mine.
Thank you very kindly to everyone in this thread. I've actually been having similar issues with my Nintendo Switch, where it dies around a reported 50% of battery life. It's not a hassle (as @Promonk mentioned) but it can take me by surprise sometimes, so I appreciate all the advice here and I'll try out some of these ideas and see if it helps my old Switch!
This will likely be up to manufacturer and how (or even if) they implemented such behavior. I'm glad Steam Deck behaves this way, it just confirms my thought that Valve really cares and made the device "the right way".
That said - it wouldn't apply to all battery devices, sadly. It may work with some phones, Switch, ROG Ally and others but don't be surprised if it doesn't.
Ahh, got it, I thought these were general procedures... thanks for the heads-up. I'll try to do some research to see if there's anything Switch-specific I should be doing. This thread already helped out a lot just in the terminology with terms like "cycling batteries". I was having a hell of a time trying to search up what was going on because I just kept finding people complaining about low battery life, lol
I realized I used incorrect technical term of "battery life" but I meant "battery health" - I"m not native English speaker and sometimes I do such kind of mistake.
Could I ask @cfabbro or @mycketforvirrad (see? i told you I will remember your nickname eventually!) if they could have a look at the title and based on their opinion either leave it or change it?
That's a technical distinction that most native speakers probably wouldn't recognize–similar to the distinction between "memory" and "storage" in computing–so I wouldn't sweat it if I were you. It was clear to me from the body of your post what you were talking about.
I have similar trouble with the Pico 4 (VR headset), which I'm told uses Xiaomi batteries. The problem is it automatically shuts itself down at 3%. There have been a couple of occasions in which it "fixed" itself and re-learned the battery capacity (the difference can be more than half an hour) but I'm not sure what did it.
I'm not familiar with the specifc device, but maybe it has a recovery mode, bootloader, or something similar you could boot into. Doubt it has auto shutdown at 3% in that mode, so you could use it to drain the battery completely and perform the calibration trick from above. Just an idea.
That's a really good idea. I have a terminal on it so I could probably reboot to recovery (it runs some kind of android).
A word of caution: running the battery down in firmware is a bad idea.
Part of the reason software shuts down at a percentage of charge is to protect the battery from what's called "over-discharge." Many battery types including lithium-ion actually require some amount of residual charge to remain in order to recharge, so that draining the battery completely dry can actually kill it. It requires specialized equipment to kick start a battery back to a charge where it's functional again, and in my experience, this is always attended by damage to capacity.
You're likely to do more harm than good by aggressively seeking precision in this regard. Yes, that 3% may very well be a complication that throws off the precision of charge estimates, but by focusing too heavily on the 3% you're ignoring the other 97% the device uses to calibrate, and that likely has a much bigger impact on the accuracy of the calibration.
Consumer devices will never have perfect charge remaining estimates because that's simply not a priority in their design. Devices that are designed to accurately estimate remaining battery tend to be single-purpose devices with discharge rates and storage conditions favorable to accurate measurement, such as medical equipment. In those cases battery cycle count and discharge rates are closely monitored because inaccuracy could very well pose life-and-death risks. OEMs of that sort of equipment can rely on the customers' technicians to maintain the equipment properly, and have elaborate warranty schemes devised to enforce proper maintenance regimes. The manufacturers of your VR headset have to account for the probability of some half-braindead doofus to store their gear under an aquarium or some other dumb thing, so they design for a broad range of situations. In other words, they focus on getting close enough in as many scenarios as possible, which inevitably means less precision.
My advice: do a two-cycle calibration using software shutdowns and see how that treats you. You can always do another if you feel it's not good enough, as four cycles isn't really that big of a deal in the lifetime of a Li-ion battery. Killing your battery in an attempt to squeeze a little more precision really isn't the outcome I think you're looking for.
Shutting down at 3% shouldn't be an issue for calibration really, since that's going to be a common feature in future use as well. You don't need or really want to completely discharge the battery, which can actually damage the capacity depending on the type of battery chemistry involved.
The main thing is to be sure to completely charge it from its shutoff percentage to full without turning it on, and to discharge it to its shutdown percentage without connecting it to power. Those are the conditions that calibrate it. You don't want usage or partial recharge to screw up the system's ability to calculate battery remaining.
To be clear: calibration won't change anything about the battery. It won't start suddenly holding a charge for much longer because you calibrated it as a technician would. All it will really do is allow whatever device it's powering to more accurately estimate the actual remaining charge. That might have modest effects in power management, but more likely it'll just be better at telling you what the actual percentage remaining is.
But that makes no practical difference if it's going to shut itself down when it estimates it's at 3%... Which would explain the few times the capacity seemed to increase so dramatically back to (close to) original.
It won't change anything in your case when it shuts down at 3% (which I would call good behavior or close enough), but it did change drmatically in my case when Steam Deck kept running for quite some time after reaching 3% (the number 3 is just a coincidence in both our cases).
That means the charging chip (or the one that keeps track of real capacity) on my Steam Deck was off by quite some degree and discharging fully and then charging fully back helped the chip to re-align itself and be more precise at telling remaining percentage and runtime.
That's part of why you do two discharge cycles when calibrating. If you do it right, both cycles contribute to the accuracy of the calibration.
I'd be leery of over-discharging the battery. It's possible to pretty much kill a Li-ion battery dead by running it completely dry. There's no point killing the battery in the process of trying to avoid buying another one.
Doesn't that kinda depend on what the 3% is based on and how the auto-shutdown is implemented? Lets assume that the device based the percentage on the fraction of estimated time remaining to estimated total runtime, and that the auto shutown were a software feature that is based purely on this percentage that's also shown to the user. Now if our flawed estimate is off in a way that would cause the device to vastly underestimate the time remaining and hence reach the 3% hurdle way to early (similar to how OP's Steam Deck kept running for a long time after reaching 3%), wouldn't that affect calibration since we're shutting down way to early in this scenario?
I suppose if the auto shutdown is instead implemented on a lower level based on different readings things will look different.
Is my logic somehow flawed?
The logic is sound. If it was software implemented and thelimut was purely set on percentage remaining that we get from the charging/battery chip (that has skewed way off), it could mean the device will shutdown with actually plenty battery left.
In my Steam Dec scenario that would mean that SD would shutdown at 3% leaving aother hour of gameplay software locked because the battery actually has the charge to run for longer time.
Thanks to good implementation, the outcome is comoletely opposite - it kept running thus draining the battery to real limit and then the charging/battery chip learned the capacity (when I fully charged it) and made new and more precise estimates.
I have since drained the battery once more this time with a bit more precise estimates, yet I kept playing for half an hour before it went from 3% to 1% in 10 minutes and the running another 20 minutes at 1% before finally shutting down. I suppose I will get even a bit more precise estimates after this. I'm not doing this exercise for the third time though, I'm using it normally for a few months before doibg this again.
Yeah, I noticed immediately from your story that the Steam Deck is better implemented. I know with perfect certainty that the Pico will shut itself down as soon as the estimate hits 3 no matter what because, of course, this device will be strapped to my face at the time (it pops up warnings at 20% and 10% too!)
Those 3% in your case may be simply to allow for clean shutdown while also leaving enought room for battery degradation in the future - ie. even with used battery those 3% will be enough for clean shutdown five years from now.
The thing is if there is a need for clean shutdown on VR headset... Does it even log/write something in some kind of memory? Does it run read only firmware? Can it become bricked by hard shutdown?
The 3% shutdown is to help preserve the battery from over-discharge as well. Li-ion batteries need a minimum charge to function by virtue of the chemistry involved, which is why I've been arguing against completely draining the battery during calibration. There's usually some circuitry that prevents the user from completely running the battery dry that's opaque to the user, but draining the battery via firmware (BIOS or equivalent, which is what you're recommending) likely isn't covered by that safeguard.
I'd HIGHLY recommend not finding workarounds for the software to drain the battery, as over-discharging a battery can render it inoperable without some method to jump start it again. Even if you have the specialized equipment necessary for a jump start, the capacity can be irretrievably damaged in the process. Trust me, I've had to dispose of literal buckets-full of laptop batteries for this very reason.
I'm not recommending discharging though BIOS. I just played until it stut itself off. It was software induced shutdown, not down to the battery management limit. I wouldn't recommend going for absolutely dead either. Even if the management learned this new capacity (down to the limit), it wouldn't make sense as it is the capacity you woj't ever reach with normal use.
I know quite some info about li-ion batteries as well and I completely agree with your point - don't discharge them until they are literally dead (or until battery management system cuts them off), it is unnecessary stress on the battery and won't give you any usable capacity over normal software limits.
Apologies. I must have confused you with another commenter.
No problem, no offense taken :-)
I may have written something that wasn't clear (enough) so you may have read and reacted to some of my comments actually.
Sidenote: I loe this community where we respect each other!
I found the comment I was thinking of and put a slightly more detailed response there.
I had forgotten that you said you were a hobbyist at the top of the post. Over-discharge is probably something that nearly every decently competent tech has encountered at some point, so it didn't really make sense for me to assume you were suggesting it.
I honestly don't know! It runs their own weird flavor of android. It has apps, which I don't use. I use the terminal to access the android settings app (not available from the UI). It has all the regular android trappings under the hood.
If it runs Android it would want to shut down cleanly. And to allow for that no matter what the battery condition is, they target 3%. This is my thought on that.
No, it's not flawed, but I think you might be overestimating the impact it would have. You're much better off taking the slight hit to accuracy to avoid killing your battery by over-discharging it.
No consumer device is ever going to have perfect battery remaining estimates no matter what you do. That just isn't a priority in their designs, and there are too many variables.