Benchmarks are essential tools for evaluating the performance and stability of both processors and graphics cards and memory in traditional systems. However, the story is very different when we enter the territory of mobile phones and tablets. Manufacturers must find a balance between maximum speed, thermal thresholds, and battery life, causing the final number of a conventional benchmark to be unable to provide all the relevant information. Geekerwan on YouTube explains this very well and reminds us of the value of energy efficiency in mobile devices.
Let's imagine you've bought a relatively new phone like the Samsung Galaxy S22. In European territory, this phone comes with an Exynos 2200 SoC, while the rest of the world gets a Snapdragon 8 Gen 1. If we compare it to something like the OnePlus 7 Pro, the S22 has several advantages, starting with a much more refined architecture. In fact, if we run a couple of benchmarks, those advantages end up showing in the results.
However, when we move to running a specific application, it is likely that the phone with three years on its shoulders manages to defeat the latest generation model. Geekerwan on YouTube uses this same exercise taking Genshin Impact as a reference: The S22 wins benchmarks, but when it comes to gaming, the 7 Pro is faster. Why does that happen?
Extreme Efficiency and Benchmarks on Mobile Devices
Basically, Geekerwan (Yun Fei) reduces it to a question of efficiency. The benchmark only measures the maximum speed of a processor or graphics module under optimal conditions. This makes sense on a desktop computer (which can reach such conditions more often), but a mobile phone must obey parameters like the CPU governor, and DVFS (Dynamic Voltage and Frequency Scaling), which adjusts frequency and voltage 'on the fly' to reduce power consumption and temperature.
Looking at efficiency more closely, the user can also check whether manufacturers are telling the truth or not when promoting new chips. For example, Geekerwan indicates with his charts that the Snapdragon 888 is worse than the 865 in terms of efficiency, and the Snapdragon 8 Gen 1 is even worse than the 888. When measuring peak performance, the new SoCs clearly perform more, but their consumption ranges between 9 and 11 watts, the same as some laptops. Apparently, Qualcomm managed to solve this with the Snapdragon 8+ Gen 1, for now.
In Apple's case, its advances were not so big with the A12 and A13 compared to Qualcomm's offering, but the A14 annihilated everything in its path, and Geekerwan anticipates that the numbers for the A15 and A16 are even better. Samsung tells another story: Its Exynos are inferior to Qualcomm's chips, and the Exynos 2200 cannot even surpass the 2100 in efficiency. Google Tensor? Basically it doesn't compete, and MediaTek surprises with its Dimensity 8100, showing a curve similar to the A14.
Why is the situation so chaotic? Geekerwan gives us two reasons: ARM did not fulfill the performance promises in its new architectures, and Samsung's manufacturing is inferior to TSMC's. This is reflected in the graphics evaluation, which occupies the last part of the video. Not going too far, an Adreno 730 manufactured by Samsung is worse than an Adreno 730 from TSMC, and this affects the entire Android ecosystem. Apple's dominance is evident (especially when A15 comes into the equation), and MediaTek seeks to position itself as a kind of 'sleeper hit' with its Dimensity.
