Device benchmarks may not mean much in the real world but they are a useful tool to compare the raw performance of different devices. But a new report suggests that they may be even less representative of what we can expect in terms of real-life performance than originally believed. AnandTech decided to take a closer look at a claim that Samsung was limiting the GPU clock speed on the Samsung Galaxy S4 to 480MHz except in the case of certain benchmarks where it would be increased to 533MHz.
AnandTech was able to confirm the claim and found that the adjustment improved performance by some 11%. It also found that the same subterfuge is used for processor benchmarks for both the international and US variant of the Galaxy S4. For those benchmarks which trigger an adjustment, such as GLBenchmark 2.5.1, the Exynos-powered Galaxy S4 activated the ARM Cortex A15 cluster running at 1.2GHz. In other suites, such as GFXBench 2.7, the Galaxy S4 runs on its Cortex A7 cluster running at 500MHz
Digging further, AnandTech found evidence of a piece of code called “BenchmarkBooster” that detects when certain apps are run and makes the adjustments above. Among the apps identified are AnTuTu, GLBenchmark 2.5.1 and Quadrant while GFXBench 2.7 and Epic Citadel are not listed.
In the end, the credibility of benchmarks is now in question. Samsung is unlikely to be the only manufacturer trying to manipulate results to get an advantage over its competition. If so, results may be artificially inflated (benchmark doping?) for certain devices, leading to erroneous comparisons.
Be sure to read AnandTech‘s report for a more detailed explanation. Then let us know what you think of Samsung’s actions. Will they deter you from buying a Galaxy S4?
Source : AnandTech