Samsung has been caught manipulating benchmark apps for Android to get a higher score than they should. The phone in question is the international version of Galaxy S 4, which comes with Samsung’s Exynos 5 Octa chipset. This SoC integrates four ARM Cortex A15 cores (1.6GHz) and four ARM Cortex A7 cores (1.2GHz) in a big.LITTLE configuration. It’s certainly quite a technical accomplishment but unfortunately the achievement has been marred by one company’s shady tactics to gain advantage over competition.
The test was performed by Brian Klug of Anandtech and it showed how Samsung has pre-set its Galaxy S 4 smartphones to get a higher benchmark score then they should. The GPU on the chipset is clocked at 480MHz when running any game or demanding app. However the test showed the GPU clock jumping to a 532MHz when running a few specific benchmark apps. Not only this but the CPU exhibited a similar behaviour. Running GLBenchmark 2.5.1 causes a switch to the ARM Cortex A15 cluster clocked at 1.2GHz, even when the benchmark itself is not running and the phone is sitting idle. Try the same with a benchmark app that isn’t part of the ‘cheat sheet’ and the SoC switches over to the Cortex A7s running at 500MHz.
Anandtech found similar behaviour on the Qualcomm Snapdragon 600 variant of Galaxy S 4. Launching these benchmark apps forced the CPU to jump to its maximum frequency of 1.9GHz, even if the benchmark itself isn’t actively running. Further investigation found the culprit. Samsung has hardcoded profiles that get activated as soon as a specified benchmark app launches. This includes Quadrant standard, advanced, and professional, linpack (free, not paid), Benchmark Pi, and AnTuTu.
Samsung hasn’t yet responded to these claims.