30
Jul
image

Update: Samsung has posted an official response to yesterday's benchmark kerfuffle, explaining that the maximum frequency for the S4 is actually 533MHz, but that it is actually scaled down for "certain gaming apps that may cause an overload". The maximum frequency, according to the statement, is also attainable in "apps that are usually used in full-screen mode" like the gallery, S Browser, etc. This may not fully explain the explicit mention of certain benchmark apps in TwDVFSApp, but it is at least nice to see an official response to the situation.

Here's the full statement:

Under ordinary conditions, the GALAXY S4 has been designed to allow a maximum GPU frequency of 533MHz. However, the maximum GPU frequency is lowered to 480MHz for certain gaming apps that may cause an overload, when they are used for a prolonged period of time in full-screen mode. Meanwhile, a maximum GPU frequency of 533MHz is applicable for running apps that are usually used in full-screen mode, such as the S Browser, Gallery, Camera, Video Player, and certain benchmarking apps, which also demand substantial performance.

The maximum GPU frequencies for the GALAXY S4 have been varied to provide optimal user experience for our customers, and were not intended to improve certain benchmark results.

Original Post: According to a report from Anandtech earlier today, Samsung might be selectively causing hardware in the Galaxy S4 to over-perform in certain benchmark apps, skewing results in its favor while frequencies achieved during the tests aren't available elsewhere.

Essentially, Anandtech explains, the Galaxy S4's GPU normally clocks in at 480MHz. The official figure isn't explicitly advertised by Samsung, but a variety of tests seemed to prove the frequency as normal, at least until it came to benchmarks.

Anandtech reports that upon launching certain apps, GPU frequency jumps to 532MHz (an ~11% boost), and the device's CPU governor is fixed regardless of what the app is actually doing. The GPU findings are specifically related to the Exynos variant, while CPU performance is affected in both Exynos and Qualcomm-powered versions of the S4.

It turns out, according to the post, that an app (TwDVFSApp.apk) is responsible for changing DVFS (Dynamic Voltage and Frequency Scaling) settings such that certain apps (whose package names are actually coded into TwDVFSApp) can "boost" performance. The apps include benchmark standards like AnTuTu, Quadrant, GLBenchmark 2.5.1, and others.

As the Verge rightly notes in its coverage, benchmarks have never been a completely accurate mirror for real-world user experience, but as Anandtech explains, Samsung's optimizations could set an unsavory precedent, encouraging SoC and device makers to focus on "gaming the benchmarks rather than improving user experience."

At the time of writing, Samsung has declined to comment on the matter. We'll let you know if that changes.

For a fully detailed explanation, hit the source link below.

Source: Anandtech via The Verge

Liam Spradlin
Liam loves Android, design, user experience, and travel. He doesn't love ill-proportioned letter forms, advertisements made entirely of stock photography, and writing biographical snippets.

  • Di Lu

    You should clearify that the GPU clock boosting happens only on the Exynos Octa version of GS4.

    • João Pedro Santos

      " The GPU findings are specifically related to the Exynos variant, while CPU performance is affected in both Exynos and Qualcomm-powered versions of the S4. "

  • http://www.androidpolice.com/ David Ruddock

    If true, this is pretty sad. I'm not saying Samsung management necessarily sanctioned it, but still - gaming benchmarks? How many more phones is that going to sell? A couple thousand? Not a smart move.

    • João Pedro Santos

      This really disappointed me too.. a lot!

    • http://androidforums.com/members/spartoi.html Spartoi

      Probably a lot tbh. Many "nerds" still care about benchmarks for some reason.

    • drvadlamudi

      A very cheeky move just so that they can show higher benchmark scores over otherphones

    • Cobra30

      But who cares about benchmarks? The average user? Nope. And I don't see Sammy advertising the benchmarks, so I don't see it as a way to sell more phones.

      • le_lutin

        A lot of the big review sites include benchmarks. I'd say this would influence their overall opinion of the phone.

      • Matti

        I don't know how Samsung markets the phone in different regions, but where I live and the surrounding countries, they tend to have posters or subway ads that feature a picture of the phone, a list of secs and a small bar chart/graph comparing it to other models. I think that's the reason. The average user might not know much, but those graphs with the "boosted" benchmarks probably go some ways in conning people.

        • Matthew Fry

          Are the units labeled on the graphs? That's a common marketing gimmick. If anyone complains you say it's just artwork with no statement as to its actual performance against competing products.

    • ssj4Gogeta

      They're not the first ones doing that. You should look at the PC GPU industry.
      Qualcomm and others probably also do that, they just use software optimizations targeting benchmarks instead of boosting clocks.

    • bungadudu

      This is not a new thing and also Samsung isn't the only one doing it. HTC does this also but nobody seems to care...

      Have a post from Richard Cerqueira - https://www.google.com/search?safe=images&redir_esc=&q=%20I%20wonder%20if%20they%20will%20include%20an%20even%20better%20version%20of%20%60libhtc-opt2.so'%20to%20'improve'%20the%20benchmark%&norc=1 (yeah, it was quickly edited)

      But on xda - http://forum.xda-developers.com/showpost.php?p=36033515&postcount=744

      Wasn't expecting AP to fall to this intoxication and start basing Samsung while not interested in knowing the whole truth.

      • mobilemann

        it's also on every other tech site in the world, and was brought about to the spotlight by anandtech, who have 10x the credibility of any android site.

        so, no.

        • bungadudu

          So if the others are just a bunch of ignorants who do nothing but rehashing various posts, AP should also do the same?
          AP was! known exactly for the vice-versa.

          • mobilemann

            Kidding? I'm saying I believe anandtech over you, clearly

  • ercl

    Hm.. Samsung has many dirty tricks under its sleeves eh..

  • URABUS0924

    Did we time warp back to 2001? Samsung didn't learn anything from the ATI quake/quack debacle?

    • Amish Crusader

      I am not condoning this practice, however Samsung is just doing what I would do anyway: push the phone/tablet at it's hardest to benchmark (switch to noop/deadline, raise the min/max) and then kick it down to conserve battery when actually using it.

      I remember being shocked and outraged when I first heard about this kinda thing (quake3.exe renamed to quack3.exe and watching the ATi benchmark deflate), but now I am just a little more leery about performance claims unless I see something in person (or perhaps on a youtube video failing that).

  • cy_n_ic

    Android police! Start writing citations and issuing bench(mark) warrants!

  • jnt

    *takes cash from Samsung* This makes absolutely perfect sense for Samsung to do this and I love my GS4 more than any other phone in the world, ever, for all of time, in the world, no... the universe...

  • faceless128

    pretty sad to cheat like that, especially for only an 11% boost. they should have been like Intel and skipped operations and stuff. just overclocking? that's barely trying...

  • Neutral

    That's typical Korean mentality. Just a bunch of copycats. Their Useless Interface UI sucks and forced closed and crashed like nobody business. Cheap piece of plastic SHITE.

    • Kay

      You might be Japanese or Taiwanese.

  • Cobra30

    I don't know. It's still benchmarking what the S4 can do, even if it's showing settings other than default. The average person doesn't care about benchmarking. Most folks that do, most likely know how to overclock the cpu to get the same results for normal use anyways. It would be different if it were somehow sending false data to the benchmarking apps, as it is it shows what it can do if needed. Not a big deal imo.

  • Rob Byrnes

    So this means that Intels Turbo Boost feature is also "cheating". OK, I know that Turbo Boost is not under the control of the OS like this boost is, but still.... I'd love to see some enterprising hacker manage to mod this file to overclock on any game :)

    • simp1istic

      Turbo boost is actually usable outside of benchmarks

    • Shitiz Garg

      Intel's Turbo Boost works for any application that pushes the CPU. It's not limited to benchmarks

  • ElvisMarmaduke

    So does this also affect custom kernels and ROMs? Because my benches rock everyone on any TW base

  • simp1istic

    I find it interesting that anyone can honestly defend or be ok with this practice.

  • nagi

    Hah.... same thing happened decades before with Nvidia and ATi. Drivers were selectively disabling quality features to get every tiny bit of speed out of the cards. Cheating, that is what this is. Shame on you, Samsung.

  • blowntoaster

    I'd still like to see official comment from Samsung. does the GPU boost work only in Benchmarks as the Verge or whoever found this tested, or does it boost in other activities as well? This is what I hate about the Android following...quick to bash even though the facts are not all there...
    boosts or not, The I9500 is still a mighty beast in and top dog in the Android realm.
    So you can scream "cheat", S-lag nonsense etc all you want. As they say, "Jealousy makes you nasty...".

    • Weston

      It is only in benchmarks. And I am in no way jealous of the S4. I hate Samsung. The s4 sucks. It doesn't hold a candle to the One.

      • blowntoaster

        thanks for making my point...Jealousy...I'm not saying the One is bad. it's excellent, just like the Xperia Z and S4 is...all good phones. But each has it's good and bad points. but when one of them gets more attention and praise, the others' supporters get jealous, and then nasty...
        accept that the S4 is a great phone, and can hold a dozen candles to the One...(and the Xperia Z too)....

        • mobilemann

          It is truely funny.

          "no, my s600 based 2gb of ram smartphone is better, urgh ~ look at that thing, it ways 13 grams more!?!!"

  • HebeGuess

    I won't fully blame Samsung for this because HTC doing something similar, though their approach was not blatantly compared Samsung's. As last time I checked I only suspect HTC only change governor but not setting min freq. My : http://forum.xda-developers.com/showpost.php?p=36033515&postcount=744

  • jz100

    Every company does this. You are being ignorant if you don’t realize this. Every company has done this for many many years, that is fact, and Anandtech knows it. They just chose to trash Samsung for whatever reason.

    There is nothing wrong with companies doing this either. It’s the exact same thing that you see when nvidia or AMD put out driver updates stating improved fps in specific games or benchmark programs like 3dmark.

    • mobilemann

      well, apple didn't respond when people reported that their phone was dual 1ghz (later found to be dual 1.3) All while samsung specifically has said to the media their GPU was running at 533mhz, which is only available to the phone running benchmark apps.

      so sorry, no.

      if you want to come out with a claim like that, at least have it be a company you can back it up with some sort of proof.

  • jz100

    I see a lot of ignorant newbies posting here who obviously know nothing about how this works. What Samsung did was good, not bad. I guess you would rather have less fps in demanding applications.

    • sillmacka

      You only get the boost (533 MHz) in benchmarks. ALL other apps run at 480 MHz. Perhaps you should refrain from calling people ignorant and actually read Anandtechs article?

    • A

      Read it again, jz100.

  • Egnimatic_Foolishness

    Reading these comments, I'm actually at a loss, as to how people think that Samsung is doing a good thing. If your reading comprehension isn't great, Samsung has it set up, so that when people (phone reviewers and nerds) run benchmark tests on the phone, the phone will ramp up more than what it's normally capable of, in order to game the system so that the benchmarks are artificially inflated.
    Which fools people that care into thinking that the phone is more powerful than it actually is. It doesn't happen during day to day use or boost performance when the end user may actually need it, but only when somebody is running a benchmark to compare performance between phones.

    I'm not saying that Samsung should be tarred and feathered, but in no way does this benefit the consumer. It's designed specifically to mislead anybody that runs benchmarking on these devices.

    • bungadudu

      You know HTC does the same thing, right?

      • Egnimatic_Foolishness

        And that has what to do with this? Rigging the system to inflate your performance is wrong regardless of who's doing it.

        • Jeffrey Smith

          Of course wrong is wrong, Samsung or HTC or any others.
          But fair is fair, and we shouldn't single out one manufacturer as being somehow unique as a cheater if others are pulling the same junk.

    • Anthony Walker

      Realistically, how much do you think Samsung is netting by doing this?

      • Egnimatic_Foolishness

        From average consumers, little profit whatsoever. To people that live on sites like this and anandtech, it's probably a decent amount of cred if not profit. I don't know how significant the difference is (I've never really been much of a specs guy) but if someone uses these values to state (in their review) that the s4 is more beastly than it is, then Samsung has gained profit from impressing the nerd community.

        • Aaron Stevenson

          Fortunately, its the nerd community that can actually take advantage by installing custom ROMS/kernels that allow over clocking of the CPU/GPU. They CAN make it so that the S4 runs these frequencies during day to day usage.

  • EH101

    If you're relying on benchmarks to make your phone purchase decisions when essentially all top tier devices offer similar performance, you're doing it wrong.

    Basically, this changes nothing. It's still a top tier phone that is equal to any other phone on the market, other than in exterior materials of course(which is equally ridiculous to complain about).

  • Dan

    I love this. I think it's great that Samsung is trashing the ridiculous results from the various benchmarking software. Benchmarks don't provide an actual measure of user experience, and I'd love to see more manufacturers actively trying to disrupt the results.

    If you want to know how well a phone or system works, use it. Don't just fire up your lemming software and read a number.

  • jOn Garrett

    who here is stupid enough to buy a phone based on benchmarks? so what if speed is off my a few milliseconds. in your everyday real world use do you really notice or even care if an app loads a fraction of a second faster?

  • McLean Riley

    I understand why it's being reported, but why do we care? Many rom devs have shown that they are just random number generators at best. If the device is smooth playing 3d games who cares what it benches?? I think it's actually a good play by Samsung, gets free advertising to the people who are comparing their phones on benchmarks. They know it's smooth now that they have updated it a few times so might as well cheat the system a little.

  • BigMixxx

    Now you KNOW you can overclock the gpu/cpu...

  • Sean Lumly

    This is why we need fps counters in games. To date, nothing beats real-world benchmarking.

    • Nomaan
      • Sean Lumly

        Excellent! Let me revise my statement:

        This is why blogs need to start using fps counters with real games... :)

        While this will give a very good indication of real-world performance, it doesn't tell the whole story about the performance of the GPU -- some devices have a terrible resolution handicap (eg. Nexus 10 or iPad4). For these tests, we look to off-screen rendering, a domain that seems exclusively tied to benchmarks.

  • Rovex

    I dont think most of you understand how a phone works.

    A PC tries to maximise FPS at all times, its not power limited and unless chocked with dust shouldn't be thermally limited at stock speeds. Phones however are thermally and power limited, as a result they arent trying to give you maximum FPS all the time, they are trying to give you the minimum acceptable FPS they can to preserve power and reduce heat.

    A benchmark is all about raw performance. What Samsung have done is simply forced maximum stock clock during a benchmark to show what the phone can actually do.

    Ok so why not 533 in games? Unnecessary thats why. Its power management determines that 480 is enough, and given the power of the GPU in the Octa, it will be. You will actually gain nothing by it running at 533, and actually lose battery life.

    As Samsung have said some other apps do run the GPU at 533, these apps are media apps that need the transcoding power of the GPU. Games simply dont.

    So why the script? Samsungs own apps will be hard coded to request maximum clock, but the benchmarks will not, the white list is there to ensure they run at the maximum potential ofthe phone.

    Is it cheating? No. Benchmarks are a poor indicator of gaming power anyway and benchmarks are not games. Games do not need max FPS, the whole point of a benchmark is to deliver maximum FPS.

  • Kevin Kerr

    Sure, Samsung. Attempt to affirm your innocence by arguing the converse. Then I must ask, why is the script of this specific behavior called "BenchmarkBooster?" Your utter insidiousness has been exposed by AnandTech.

Quantcast