Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Samsung cheating in benchmarks.

WaterlilyWaterlily Member UncommonPosts: 3,105

Well, this is more cellphone related. But I thought it was interesting enough.

http://www.anandtech.com/show/7187/looking-at-cpugpu-benchmark-optimizations-galaxy-s-4

Samsung has a command that can check the speed of their CPU at any time as Anand shows.

So when Anand ran their benchmarks, some were running the CPU at 532...the maximum frenquency the chip can run at.

Ok, so I guess you could argue the cellphone just increases the CPU speed when needed.

 

Hm, not so, since it doesn't seem to be doing that for all benchmarks. So what's going on here.

 

So Anand poked around the TwDVFSApp.apk file with a hex editor. And low and behold, they're using detection code to detect certain benchmarks and do CPU specific optimisations. It does this for AnTuTu, BenchmarkPi and Quadrant benchmarks.

 

I'm writing this on a Samsung screen, so keep that in mind. But I'm not shocked that these companies go out of their way to hide and cheat when there are billions involved.

 

 

 

Comments

  • JaknifeJaknife Member Posts: 51

    Not doubting your figures at all, but there are 2 versions of the S4, with different main  CPU and GPU.

    There is the GT-I9500 and the GT-I9505

    One is faster than the other graphically, but has poorer battery life, and the other is slower but has a lot better battery performance.

    So just in case, it depends on what Samsung quoted for each of the 2 versions.

     

  • RidelynnRidelynn Member EpicPosts: 7,383

    No real news here. Graphics card manufacturers have been doing this for years with driver optimizations on PCs.

  • DarwaDarwa Member UncommonPosts: 2,181

    Look at the figures that Blizzard gives out.............

    Quix, go get 'em! ;)

  • QuizzicalQuizzical Member LegendaryPosts: 25,348
    This is why you should always use the same benchmark for both performance testing and power testing.  Use different benchmarks and hardware manufacturers will cheat by clocking the part higher for the former and lower for the latter.
Sign In or Register to comment.