Samsung Tweaks Graphics in Exynos Galaxy S4 to Inflate Popular Benchmark Scores

[dgstorm]

For any of our members who have been around PC technology for a decade or more, you might remember long ago when the graphics card makers like NVIDIA and ATI used to artificially boost their graphics benchmark scores by including "optimization tweaks" for certain benchmarks. It was basically a sneaky way of manipulating the results of graphics results in the popular benchmarks of the time, but didn't really translate to real world performance numbers. The sole purpose of this was to make their benchmark numbers look better than the competition to entice consumers to their product.

For the most part this ended up "biting them on the behind" when it was later uncovered by savvy tech-guru sites like Anandtech. The practice was mostly dropped because it ended up resulting in bad PR. Sadly, it looks like history has repeated itself in the mobile industry, and never everyone has learned from the past. Samsung has been caught red-handed doing exactly this with their Exynos 5 Octa-core on the Galaxy S4. Once again, Anandtech was one of the tech sites which helped confirm/uncover the deception.

Apparently, what is occurring is that Samsung's Exynos 5 chip in the SGS4 can detect if one of the popular benchmarks (like AnuTuTu, Quadrant and GLBenchmark 2.5.1) is being run. It then automatically unlocks a higher clock speed in the PowerVR SGX 544MP3 graphics chip, boosting it from 480MHz to 533MHz. This results in as much as a 10% boost in graphics performance which will never actually translate into a faster experience for end users in their apps and games.

Even if you are huge fan of Samsung and the SGS4, we should still call them out on this one. It's just a stinky thing to do, and it needs to stop. In the mean-time, this reminds us that the artificial benchmarks should never be completely trusted. They can still be a decent tool to offer an overall impression of performance; however, ultimately, your experience on your device could vary, and the best "benchmark" you can use is the apps and games which you like to use.

Source: GalaxyS4Forums.net via Anandtech

 

[Asgard]

One of the reasons why I don't trust this benchmarks too much

The other one is lack of consistency, I've run the same benchmark 3 times in my Galaxy Nexus, and got a different result everytime like if I were testing 3 completely different phones.

 

[Dusty]

The best benchmark, in my opinion, is a simple video. Two phones bone stock fresh from the box. Run a few apps and perform the same tasks side by side... done. Most people will find that most newer phones are within a blink of each other at the most. How fast do you REALLY need to post to FB? The your phone is only as good as the biggest background-tasking app you install.

Benchmarks are basically for show, kind of like "number" or "article" racing with cars.