Synthetic tests do not reflect the actual performance of the smartphone, especially if we are talking about flagships, but that doesn't stop manufacturers from trying by unfair means to improve the results in the benchmarks. In 2013 cheating caught Samsung, and then it turned out that the machinations involved in almost everything. When the truth was revealed, large companies refused to such tricks. Chinese manufacturers, apparently, heard nothing about it and continue to use questionable methods in the pursuit of "parrots". Meizu and OnePlus got in among the first.
Working with the utility Snapdragon Performance Visualizer that tracks the speed of the processor when running different applications, users of the XDA Developers forum found that when running some benchmarks "flagship killer" OnePlus 3T does not reset the frequency of the chip to a standard of 0.31 GHz, keeping it at the level of 0.98 GHz for "small" cores and 1.29 GHz for more powerful. Because the smartphone is not loaded, it seemed suspicious.
The fee for the maximum result will be heating of the case
To get to the cause of this strange behavior, enthusiasts turned to the developers of Geekbench. They promised to help and created a version of the app that the smartphone will not be able to identify as a benchmark. On the first run the results were about the same, but during the following tests the standard Geekbench began to pull ahead. To 17 run difference in multi-core test reached 5%, and the temperature of the body of the smartphone is with overclocked CPU was on average three degrees higher.
OnePlus 3T overstates the frequency of nuclei in AnTuTu, Androbench, Quadrant, Vellamo and GFXBench. The company said that this mechanism is intended primarily for games, and in future updates OxygenOS smartphone will no longer activate it in the benchmarks.
What c Meizu?
The most powerful core proved to be the most lazy
Second test subject was the flagship of the beginning of 2016 Meizu Pro 6 desacuerdos processor Helio X25. It was originally shown the highest results in the benchmark — for unknown reasons, the two most powerful cores are almost never included, even at maximum load. With the upgrade to Flyme OS 18.104.22.168 G unit getting hotter and began to pick at Kraken, WebXPRT 2015 and JetStream 2-2.5 times more points. The Geekbench test showed that count on the increase in the speed of operation of applications is not necessary — "big" cores did not show themselves in disguised benchmark, but in the standard version of the result has improved by almost 80% for multithreaded mode. It turns out that most Helio X25 behaves like an ordinary and, frankly, not the most powerful OCTA core chip.
Optimization is more important
There is no doubt that after controlling for other Chinese smartphones, the list of violators will expand. The producers should pay more attention to the optimization of the system and not overclocking hardware in the benchmarks. For example, OnePlus 3 worth from the very first, to teach how to use 6 GB of RAM, so it does not kill most of the background processes to conserve power. Google Pixel barely squeezed into the top ten most powerful Android flagships of 2016, although he has not ceased to be one of the fastest devices on the market. Here and believe after that, synthetic tests...
Source: XDA Developers
For those who want to know more:
- A slight miscalculation: a service to store code Networks accidentally deleted almost all the data
- Bioengineers have created artificial cells that have "passed the Turing test"
- The Italians have developed an Autonomous robot-the porter
- 15 "untranslatable" emotions that you never knew
- Beneath the island of Mauritius traces of the "lost continent"