Close ad

Server & Hosting AnandTech.com caught Samsung cheating on Galaxy S 4 benchmarks:

We should see roughly an 11% performance increase in GLBenchmark 2.5.1 over GFXBench 2.7.0, and we'll eventually see a bit more. The reason for this difference? GLBenchmark 2.5.1 seems to be one of the benchmarks that is allowed to take advantage of higher GPU frequency/voltage settings.
[...]
At the moment, it seems that only certain benchmarks are allowed to use higher GPU frequencies. AnTuTu, GLBenchark 2.5.1 and Quadrant have fixed CPU frequencies and a GPU clock of 532 MHz, while GFXBench 2.7 and Epic Citadel do not. Upon further investigation, I came across an application that changes the behavior of DVFS and allows this change of frequencies. Opening the file in a hex editor and searching for strings inside, I discovered hard-coded code containing profiles/exceptions for specific applications. The string "BenchmarkBooster" speaks for itself.

So Samsung set the GPU to overclock when running certain benchmarks and the phone did better in the test. At the same time, overclocking is only available for benchmarks, not for games and applications. What to expect from a company that paid students to write fake critical reviews of competing phones?

However, it is surprising that at the time of optimization for CPU and GPU benchmarks of phones or tablets, anyone can still give. For example, the iPhone usually didn't have the highest processor speed, the most RAM, or the best test results, but it was smoother and faster than its competition thanks to software optimization. In the Android world, it's obviously still a matter of who has a higher CPU clock or better benchmark results, while software optimization comes second. Overclocking the GPU is obviously easier.

.