Samsung appears to be throttling 10k Android apps on Galaxy devices | Hacker News Reader

Samsung appears to be throttling 10k Android apps on Galaxy devices

osnews.com

91 pointsmarcodiego45 comments

OnePlus did this too[0]. They effectively skipped slowing down all apps which could be used for benchmarking. This was discovered by someone and it made major news headlines.

[0]: https://www.androidauthority.com/oneplus-9-benchmarks-164005...

Your link doesn't match what you are saying.

Your link talked about a phone throttling popular apps which were marked as battery hogs to help conserve battery life.

Depends how you look at it. What can't be denied is that they removed the performance limitations for benchmarking apps and this is the real issue - you don't get that kind of performance in real life usage.

It sounded like different apps were throttled different amounts which means that you need to measure the performance of different apps separately. If you wanted to throttle a benchmark, what level of throttling should it get?

>In testing, I had encountered something which really perplexed me, and caught my attention; seemingly inexplicable slow browser benchmark figures which were not in line with any other Snapdragon 888 device in the market, getting only a fraction of the scores and performance of other devices.

We have detected that OnePlus is blacklisting popular applications away from the its fastest cores, causing slow down in typical workloads such as web browsing. We have confirmed that (a) benchmarks or (b) unknown apps get full performance; most of the top popular non-benchmark apps get notably reduced performance. This is perhaps to improve battery life at the expense of performance, but it does mean that the regular benchmark results are somewhat useless for user experience.

https://www.anandtech.com/show/16794/oneplus-9-performance-e...

There is a difference between benchmark apps being meaningless and cheating in benchmark apps.

For example one HN user told a story of a magazine who benchmarked different compilers including this user's. This user's compiler recognized the benchmark code was dead code and just eliminated it where other compilers didn't. This magazine called him out for cheating and it hurt sales for his compiler. He didn't mean to cheat the benchmark. It's just that the benchmark being used in an outdated methodology for evaluating performance.

In this case, Samsung is not only cheating on battery life testing, but also cheating on performance benchmarks.

>The only sensible rationale for such a decision is to improve a device’s power efficiency and battery life.

https://www.anandtech.com/show/16794/oneplus-9-performance-e...

>John Poole, the lead developer of Geekbench, was able to reproduce the wild performance changes based on whether the S22 thought it was running a benchmark or a game. Poole changed Geekbench's package name to that of Genshin Impact, a popular game, and saw benchmark scores plummet. The Snapdragon Galaxy S22 dropped its single-core score 46 percent

https://arstechnica.com/gadgets/2022/03/samsung-caught-throt...

I don't feel this is cheating. Sure, I agree there could have been more transparency, but I don't think this amounts to cheating.

If Samsung had used this performance limiting feature during testing, and then released phones that didn't have it still citing the better battery numbers, then yes that would be cheating.

That's not what's happening here.

Is Intel cheating too because their processors can allocate tasks to more performant or efficient cores?

Is Intel cheating because they can exceed the TDP ratings when thermal and power constraints allow it?

When Volkswagen and other automakers applied different software to cars that were being emissions tested and ones that are driving on the road, that is cheating.

This is different imo.

Slowing down all well known programs except benchmarks can't possibly be anything else but cheating.

I tried a OnePlus once (OnePlus 7T in 2020). I thought at that time OnePlus was supposed to be free of bloatware, a pure Android experience.

However it did some annoying thing with any app that it considers to be a game. Whenever opening a game it'd pop up some message how it "optimized" the game (whatever that meant), and it also had some unremovable game center app (forgot exact name).

I didn't want those popup messages nor that game center thing, but there was no way to disable it, there was only some complicated way of disabling it individually for every specific game which it could also forget after updates.

So if OnePlus was supposed to be bloatware free, nope, not at all, comes with annoying things that can't be disabled.

Nexus and Pixel are the only bloatware-free Android phones I've seen.

I wouldn't call a feature bloatware. I understand your grievance but to me bloatware it's third party apps, not components of the OS.

Its definetly bloatware, regardless of who put it in

By that logic, Apple is the worst offender of the bunch.

Well, I mean, yes.

True. BatteryLifeWare?

I always thought only vendors can put in bloatware - it comes non-optional per default. Who apart from the software vendor could be responsible?

I think it agree. Bloatware is always installed by the vendor but is usually third-party software.

I think the key definition for me is if it is part of the "system" it isn't bloatware. Bloatware is something that is unnecessary added on. Especially if it could be added in by a third party, for example pre-installed Facebook apps.

Of course the system isn't a well defined concept. In the example about I think the "optimizer" feels like system whereas the game center feels like bloatware.

I don't know if it's the SoC or Android itself that's gotten so bad lately, but phones these days are exhibiting similar battery life to phones from years ago while having enormously larger batteries.

I think we're seeing the same phenomenon we've seen with computer memory: Hardware has become larger and faster by orders of magnitude as compared to the early 1990s - yet actual user-facing programs have not become faster, I actually know people who prefer Word 6.0 to modern Word if given the choice.

More power does not translate to faster experience, it translates to less efficient software. More battery life translates to doing more wasteful things in the app, optimisation is not considered "worth it" anymore.

Related, even if not directly referencing battery life or Android: https://news.ycombinator.com/item?id=30442529

>I actually know people who prefer Word 6.0 to modern Word if given the choice.

In the same vein, I use Photoshop CS2 because it works in Wine, the licenses were distributed for free by Adobe themselves, and specially because it has most of the features I would want from PS.

I'd imagine it's hardware growing to meet the capabilities of the battery...

While batteries keep growing, we keep increasing screen sizes, pixel density, refresh rates... I believe 5G also consumes more power.

On my last phone, I had a few apps that were power hogs, but I'd say that's more poor design by their programmers.

> On my last phone, I had a few apps that were power hogs, but I'd say that's more poor design by their programmers.

The two misbehaving apps I use at the moment are Google Fit / Google Earth. Even after I finish my workout it drains the battery, as indicated in Settings -> Battery. I have to force stop them manually. If Google can't write a behaving app, after all this time, isn't that indication of some fundamental flaw in the OS ?

Both of these applications provide tons of data to Google; they have a clear incentive to keep them running in the background.

A significant factor in this is screen capabilities. Look at the battery life test results comparing phones with ~1080p screens and ~720p ones, and those with simply larger screens at the same resolution, and you will see definite patterns. There are other significant factors like the overall screen tech used, so it is far from entirely clear-cut, but the fact the device size and resolutions have increased along with battery capacity goes quite a way to explaining why the extra capacity isn't being well reflected in overall run-time between charges.

We also tend to use our phones a lot more than we used to, so they spend more time in a given day in active modes rather than standby ones, and many people have more background apps checking for updates than they used to, so there are extra short periods of CPU wake time, and radio activity, when the device would otherwise be in a lower power standby state.

SoCs being more powerful will have an effect too, but they are also getting more efficient especially when not in full use, so I doubt that this is significant compared to the factors I mentioned above.

device size and resolutions have increased along with battery capacity

Is that actually true? Five years ago it seemed like every phone had a QHD screen with some 4K options. Now most phones are 1080p, or are set to run at 1080p despite having higher resolution screens. The best phone display I've ever used was the 2560x1440 beauty on my Note 4 and I'd kill to get another phone with a QHD 16:9 OLED panel.

> Is that actually true?

It seemed to be the last couple of times I was researching new phones. Though there is a significant caveat there that I don't buy top-of-the-range flagship devices as the hit when I break them is too high, perhaps the difference does not carry to the really fancy devices with all mod cons because they have more expensive but more efficient screens? The other caveat being that I'm going by 3rd party tests: this isn't something I've verified with my own study.

As mentioned above, overall tech could play a role in some larger and/or higher res screens doing better than they might compared to other models wrt battery life: IIRC Samsung use AMOLED screens which are generally more power efficient than the screens that might be used in my price range?

Not my experience; I have a Samsung Galaxy S20 FE. The battery regularly lasts 3 to 4 days with light usage. Something that was unheard of a few years ago. Of course, you can compare this to the pre smartphone era, where batteries often lasted a week, or longer, but that's not really a fair comparison.

Do newer Samsung phones still have the ultra power saving mode? It used to convert the screen to greyscale, throttle CPU, disable all background services and apps, and useable only for phone calls, sms, and basic web browsing. It could last over a week.

> It could last over a week

Install LineageOS but skip "GAPPS" (Google apps), don't install nosy background-active spyware like Facebook and its ilk and you should be able to easily go beyond 1 week on a device with a 4500mAH battery with "light usage". in full colour, with normal functionality. Standby time of up to a month (I got 21 days on a Motorola Defy with a 1700mAh battery), "normal" charge interval of twice per week with normal use. I would not buy a device which needs daily charging.

Yes they still do, without the greyscale as standard nowadays though. Its pretty great, you can "promote" 1-2 normal apps to work in that mode too. I used it heavily on trips where I only needed the camera and gmaps anyway - my trusty S10 lasted for days thanks to it.

My current phone is also an S20 FE (SD865). Comparing it to my previous phone, a Oneplus 6 (SD845), both get almost identical amounts of screen-on time while browsing the web at 150 nits, despite the S20 FE having a 4500mAh battery and the OP6 a 3300mAh battery. The SD835/845 era was a high water mark for performance per watt and so far the situation is only getting worse. The 888 and 8 gen 1 are both hot power hogs and the exynos 2200 is a garbage fire.

My original Galaxy S in 2010 would last 2-3 days pretty regularly...

Are you still using it as your daily phone? If yes, which custom rom/android version are you running? and is it still usable?

I was using it as my only phone with the stock OS up until mid 2018 when it started having what looked like hardware issues - like, sometimes the phone wouldn't ring but I got the missed call notification seconds later. It still had around 1 - 1.5 days of battery life at the time, with the original battery.

From my understanding is part of the issue is the quality of batteries is decreasing.

Too much demand so phone makers have to use shittier batteries.

Anecdotally, I have an S10 and in the two years I've owned it, it's gone from finishing the day with 30% left, to having to charge it 1-2 times a day. I've never had an android phone reach this level, my V30 would still get to the end of the day at 0-10% after two years.

Did something change in how you use it? Or do you use fast charging a lot?

Also purely anecdotally ofc but my 3y old S10 still finishes with 10-20% left at night, and I'm using it heavily for browsing during the day (FF+Ublock). I do have some power saving settings enabled though, and disabled fast charging (since it charges over night anyway).

I assume screen is a massive drain; starting with Apple's "Retina" push, and all manufacturers following, my tiny phone now has resolution and refresh WAY higher than anything I deem functionally useful (though I do have a friend or two that will incessantly talk to me about how "smooooth" scrolling or whatever is; don't care!).

As well, while on my original HTC G1 I optimized the thing to an inch of its existence, with modern phone there is SO much happening in background, I don't even know where to start so I give up and have literally bought couple of 10-packs of charger wires, and have wireless charging pods in several rooms.

I miss the days of my Palm Treo, when I could take it on a vacation and not bother with charger :P

What are the effects of these performance limits, aside from putatively increasing battery life?

If they did this in an above board way, this would be the default power profile in the OS/HW. Now they're charging users for high CPU performance, but only benchmark apps get to use it. It's a bait and switch where they try to fool people into thinking there's high performance at low power usage.

Are you sure its benchmarks which are the only apps which can use it?

The idea of making a blacklist of apps which take up 100% of a CPU despite not needing it to function sounds like a good idea to me. Some games might do that by endlessly spinning rather than sleeping or something until the next time they need to process state/draw a frame etc.

Benchmarks are obvious candidates not to put on such a list because when you use them taking up 100% performance is reasonable.

So depending on the apps on the list I could see it being deceptive, or just curating a list of apps which weren't developed with power usage in mind.

After looking into it I don't get the outrage about it - it's a bog standard setting in the power settings. Here on my german phone it's called "Verarbeitungsgeschwindigkeit" (Processing Speed) under advanced power options. You can choose between optimized (standard), high and maximum. So sure, you can't disable the "Game Optimizing Service" app, and its set to optimized/throttled by default, but you can easily just set it to maximum (ie. disable throttling).

I'd argue for 90% of customers slight throttling and better battery life is a good thing.

Depends where you sit on the issue - when Apple used algorithms to throttle cpu spikes to prevent devices with older batteries from shutting down it led to a significant outcry and numerous lawsuits.

The samsung profiles fit squarely in vanity: high benchmarking scores, but slower real world performance.

I did start to notice sluggishness since a week or two ago. Maybe they thought they could target S21 users to get them to buy an S22? I actually saved and shelled out for a S21 Ultra 5G thinking it would tolerate the bloat from modern apps for at least 3 years.

Also posted here yesterday https://news.ycombinator.com/item?id=30542249

I'm fine with this, battery life is outstanding lately. My major frustration with Android is Googles automated approach to everything. For that reason alone I'm considering moving back to ios after more than a decade.

Reply on news.ycombinator.com