Samsung is throttling the performance of over 10,000 apps

AppleInsider

Samsung is the latest smartphone vendor to be caught using a list of apps that the operating system throttles automatically -- and benchmarking apps are not on the list.

Some of the affected devices in the Galaxy S21 lineup. Credit: Samsung

As reported by Android Authority, Korean Twitter users have compiled a list of 10,000 apps that are marked as subject to "performance limits" imposed by Samsung's Game Optimizing Service. Yet, despite being ostensibly related to games, the list of affected apps also includes Microsoft Office apps, Netflix, Google Keep, TikTok, among others.

This list also included certain Samsung apps, such as Samsung Cloud, and even the default phone dialer preloaded on Samsung phones.

Notably, benchmark apps such as GeekBench 5 and Antutu have not been throttled.

In a test by a Korean YouTuber, the creator renamed an otherwise-unaffected app to that that is listed as subject to performance limits. The benchmark registered a much lower performance score for the renamed app.

In his test, an unmodified app of Wild Life Extreme scored 2618, while the app with a spoofed package name that matched an app on the list to be optimized scored 56% worse, at 1141 points.

According to Twitter user GaryeonHan, Samsung employees believe that the intentional throttling is "bad behavior." They drew comparisons to the Volkswagen emissions scandal, where the German automaker had intentionally programed emissions controls to activate only upon emissions testing. The Twitter user has also claimed that Samsung's Vice Chairman Lee Jae-yong has ordered an investigation into the issue.

There are two possibilities for the list. First, the company could use the list to keep battery drain down on some of the world's most popular apps, artificially inflating battery runtime tests performed by users and themselves for marketing purposes. Second, Samsung could also be attempting to game the benchmarking tests commonly performed on tests and not delivering that speed to all apps evenly.

The performance limits don't appear to be installed on phones in the Galaxy S22 line, and the Galaxy S20 FE and Galaxy S10e devices, Android Authority has observed. Rival smartphone manufacturer OnePlus had previously been caught "optimizing" apps that ultimately subjected them to poorer performance on certain OnePlus phones.

Read on AppleInsider

sflocal

The way iKnockoff companies like Scamscum get away with this is disgusting.  If this were Apple, guaranteed that the the botomfeeding lawyers, media, and iHaters would be screaming for Tim Cook's head, and clickbait YouTubers would fill the Internet with countless videos on how Apple is a "greedy" company and ripping off users.

anonconformist

It will be interesting to see if they claim it’s for battery optimization and the reason why.

Truth of the matter is many applications are inefficient in how they’re designed to run, and burn through battery for no real good effect.  As much as that’s true all too often, users should make such choices as to whether they get better battery life versus speed, and not let it be decided other than by themselves.

JustSomeGuy1

If you look beyond the obvious "they're cheating" response, this opens up some fascinating questions.

*Why* are they doing this? The answer is a lot less obvious than you'd think. Benchmarking apps will use as much CPU as they can get their hands on, and so will obviously be affected by being on the list, but most of those 10,000 apps should not be. After all, MS Office, Netflix, Tiktok - generally speaking, none of them require full CPU utilization. They're just not doing that much work.

So why are they listed? DVFS should be enough to manage the CPU for all these apps! And even though the DVFS implementation on various Androids has sometimes been poor, it should still be better than the ridiculously blunt instrument this throttle is.

In fact, this throttle is likely bad for battery life in most cases. Race-to-idle is demonstrably the best way to go, in almost all cases. The only exception would be if the CPU were set to run so far off the optimal voltage that running at full speed was dramatically more expensive than executing the same number of cycles at lower speeds, which shouldn't be the case, except possibly for a couple of the recent Qualcomm chips (and I'm skeptical even for those). Even then the DVFS implementation would have to be totally broken.

The two major obvious exceptions to this general rule are benchmarks and graphics-intensive games. On PCs/Macs, you'd also include video editing and a few other things, but these are phones, so that's mostly not a thing. And those aren't going to be on the throttle list or the phone will show up as slow when tested. So... what's the throttle for? Can they have broken DVFS that badly?

This is really really weird.

Edited to add: I could imagine using this kind of tool to deal with very badly written apps that suck up all available CPU for no reason - say, an app with a busy idle loop. But there can't be that many of those.

rezwits

Response to ABOVE ^^

BATTERY PERFORMANCE, period!

iyfcalvin

sflocal said:

The way iKnockoff companies like Scamscum get away with this is disgusting.  If this were Apple, guaranteed that the the botomfeeding lawyers, media, and iHaters would be screaming for Tim Cook's head, and clickbait YouTubers would fill the Internet with countless videos on how Apple is a "greedy" company and ripping off users.

If Apple did that, there’d be huge class action lawsuits lined up on Cook!  😂

Anilu_777

rezwits said:

Response to ABOVE ^^

BATTERY PERFORMANCE, period!

But if Apple throttles for battery performance the world loses its collective mind and class action lawsuits are thrown around like confetti!

anonconformist

JustSomeGuy1 said:

If you look beyond the obvious "they're cheating" response, this opens up some fascinating questions.

*Why* are they doing this? The answer is a lot less obvious than you'd think. Benchmarking apps will use as much CPU as they can get their hands on, and so will obviously be affected by being on the list, but most of those 10,000 apps should not be. After all, MS Office, Netflix, Tiktok - generally speaking, none of them require full CPU utilization. They're just not doing that much work.

So why are they listed? DVFS should be enough to manage the CPU for all these apps! And even though the DVFS implementation on various Androids has sometimes been poor, it should still be better than the ridiculously blunt instrument this throttle is.

In fact, this throttle is likely bad for battery life in most cases. Race-to-idle is demonstrably the best way to go, in almost all cases. The only exception would be if the CPU were set to run so far off the optimal voltage that running at full speed was dramatically more expensive than executing the same number of cycles at lower speeds, which shouldn't be the case, except possibly for a couple of the recent Qualcomm chips (and I'm skeptical even for those). Even then the DVFS implementation would have to be totally broken.

The two major obvious exceptions to this general rule are benchmarks and graphics-intensive games. On PCs/Macs, you'd also include video editing and a few other things, but these are phones, so that's mostly not a thing. And those aren't going to be on the throttle list or the phone will show up as slow when tested. So... what's the throttle for? Can they have broken DVFS that badly?

This is really really weird.

Edited to add: I could imagine using this kind of tool to deal with very badly written apps that suck up all available CPU for no reason - say, an app with a busy idle loop. But there can't be that many of those.

My current role for over 7 years now is developer support for a major OS.

The number of stupid things a large number of developers do that I’ve seen suggests you have unwarranted confidence in the general developer population overall.  I would also note that which platform it is on is just an implementation detail.

JustSomeGuy1

anonconformist said:

JustSomeGuy1 said:

If you look beyond the obvious "they're cheating" response, this opens up some fascinating questions.

*Why* are they doing this? The answer is a lot less obvious than you'd think. Benchmarking apps will use as much CPU as they can get their hands on, and so will obviously be affected by being on the list, but most of those 10,000 apps should not be. After all, MS Office, Netflix, Tiktok - generally speaking, none of them require full CPU utilization. They're just not doing that much work.

So why are they listed? DVFS should be enough to manage the CPU for all these apps! And even though the DVFS implementation on various Androids has sometimes been poor, it should still be better than the ridiculously blunt instrument this throttle is.

In fact, this throttle is likely bad for battery life in most cases. Race-to-idle is demonstrably the best way to go, in almost all cases. The only exception would be if the CPU were set to run so far off the optimal voltage that running at full speed was dramatically more expensive than executing the same number of cycles at lower speeds, which shouldn't be the case, except possibly for a couple of the recent Qualcomm chips (and I'm skeptical even for those). Even then the DVFS implementation would have to be totally broken.

The two major obvious exceptions to this general rule are benchmarks and graphics-intensive games. On PCs/Macs, you'd also include video editing and a few other things, but these are phones, so that's mostly not a thing. And those aren't going to be on the throttle list or the phone will show up as slow when tested. So... what's the throttle for? Can they have broken DVFS that badly?

This is really really weird.

Edited to add: I could imagine using this kind of tool to deal with very badly written apps that suck up all available CPU for no reason - say, an app with a busy idle loop. But there can't be that many of those.

My current role for over 7 years now is developer support for a major OS.

The number of stupid things a large number of developers do that I’ve seen suggests you have unwarranted confidence in the general developer population overall.  I would also note that which platform it is on is just an implementation detail.

I too have seen a ton of developer stupidity over the years, but I think in this case you're overly pessimistic, in that this sort of thing causes significant customer blowback: it's easy to tell that an app is chewing on your battery, and people complain loud and long when that happens.

rezwits said:

Response to ABOVE ^^

BATTERY PERFORMANCE, period!

Your answer is nonsensical. As I pointed out, this strategy is likely to harm battery performance overall due to the advantages of race-to-idle.

mjtomlin

Is it a question of throttling down other apps, or a question of not properly throttling benchmark apps so that they can score higher? They were caught doing this once before, as were other Android based phone makers.

verne arase

I'm not sure the is a bad thing for end-users - it may simply be Samsung’s attempt to alleviate dispatching flaws within Android.

I remember thinking a long time ago that the appearance of WaitNextEvent loops could be terribly wasteful if not done right - and I don't know enough about Android to know if it’s done right.

daniPhone

iyfcalvin said:

sflocal said:

The way iKnockoff companies like Scamscum get away with this is disgusting.  If this were Apple, guaranteed that the the botomfeeding lawyers, media, and iHaters would be screaming for Tim Cook's head, and clickbait YouTubers would fill the Internet with countless videos on how Apple is a "greedy" company and ripping off users.

If Apple did that, there’d be huge class action lawsuits lined up on Cook!  😂

I remember a few years ago Apple throttling entire devices with a degraded battery..

hammeroftruth

daniPhone said:

iyfcalvin said:

sflocal said:

The way iKnockoff companies like Scamscum get away with this is disgusting.  If this were Apple, guaranteed that the the botomfeeding lawyers, media, and iHaters would be screaming for Tim Cook's head, and clickbait YouTubers would fill the Internet with countless videos on how Apple is a "greedy" company and ripping off users.

If Apple did that, there’d be huge class action lawsuits lined up on Cook!  😂

I remember a few years ago Apple throttling entire devices with a degraded battery..

They did and there were many lawsuits. Apple shot itself in the foot by offering the battery replacement program that ended up not getting the effect they wanted. It ended up costing them a lot of money because they underestimated how many people would replace their battery, even new batteries were replaced on devices if customers asked, and they didn’t reach their estimate of how many people would end up upgrading to a new device. Which in turn hurt their image more when customers got upset that the new battery didn’t speed up their device. 

I got torn a new ahole here, when I posted that iPhones only have a 3-5 year relative lifespan and a battery on an older device won’t make it as fast as it was when it was new because more than likely the OS has been updated and that update takes a toll on older hardware. 

Nobody believed me when I said the main reason why the throttling was introduced was to prevent the device from shutting off unexpectedly which is like pulling the cord on your desktop computer while it’s on. 🤷🏽‍♂️
 

Apple did fuck up by not explaining it properly and didn’t even mention it when it rolled out in a software update, and didn’t give customers an option to turn it off and let their device shut down. 

MplsP

JustSomeGuy1 said:

If you look beyond the obvious "they're cheating" response, this opens up some fascinating questions.

*Why* are they doing this? The answer is a lot less obvious than you'd think. Benchmarking apps will use as much CPU as they can get their hands on, and so will obviously be affected by being on the list, but most of those 10,000 apps should not be. After all, MS Office, Netflix, Tiktok - generally speaking, none of them require full CPU utilization. They're just not doing that much work.

anyone who's used any microsoft app knows how they unnecessarily suck resources. 

hmlongco

Rather telling that they throttled major applications and games.... but not benchmarking apps.

THAT'S the part that's going to come back to haunt them....

OutdoorAppDeveloper

Some form of app throttling may actually be a good idea. Why? Well because developers write their apps to work as well as possible on the devices they target the app for. They don't care how much power the apps use. They just want their apps to look as cool as possible when the user opens them for the first time. Later phones with much faster processors don't need to run those apps at full speed to make them run like they did on the devices they were designed for but the app does not know that. The app just uses the processor as much as it can.

Example: A simple bar graph. Each bar in the graph is going to animate sliding smoothly to the right when the user scrolls the graph because that looks cool. You could do it with a Metal shader so it runs efficiently on the GPU but that's hard. Instead you just draw a rectangle for each bar on the CPU. You do that over and over again so that the bar animates. Frame rate? Who cares? Just draw the bar as fast as you can. So what if the iPhone feels warm in your hand while your app is running? Just so long as it looks pretty that's all that matters.

Understand now? That's why Samsung is throttling apps. Users don't know how or why apps work they way they do. They just see that their shiny new Samsung phone gets less battery life than the previous model with the same battery but a less powerful processor.

JustSomeGuy1

MplsP said:

JustSomeGuy1 said:

If you look beyond the obvious "they're cheating" response, this opens up some fascinating questions.

*Why* are they doing this? The answer is a lot less obvious than you'd think. Benchmarking apps will use as much CPU as they can get their hands on, and so will obviously be affected by being on the list, but most of those 10,000 apps should not be. After all, MS Office, Netflix, Tiktok - generally speaking, none of them require full CPU utilization. They're just not doing that much work.

anyone who's used any microsoft app knows how they unnecessarily suck resources. 

True in general, probably false in this context. The issue is CPU utilization specifically, and in particular unbounded (or at least continuous) utilization. The bar graph example posted above by "OutdoorAppDeveloper" is a reasonable example. MS phone apps are unlikely to be doing that as it would be noticed.

There's still a really interesting story here that we haven't heard. Or, sure, it could be Samsung devs being insane. But chances are it's not that simple.