Why does AnandTech and other reviewers use benchmarks instead of real games?

[t1gran]

A couple of questions about smartphones' performance measurement:

1) Why does AnandTech and other reviewers use benchmarks instead of real games, when testing smartphones' performance? What's wrong with GameBench or similar apps?

2) I found in some review that Injustice 2 played on Galaxy S9+ (Exynos 9810) loads only 7% CPU and 21% GPU. According to other review this game on this model has a limit 30 fps (while 60 fps on Pixel 2 XL), poor graphics and lags. Does it mean game developers do not load processors in full? Why is it so?

 

[Andrei.]

Testing real world games is problematic in terms of tooling and repeatability. GameBench is a good effort but it's still inherently very flawed to get empirical data on something. GameBench also doesn't solve the iOS conundrum. As long as real mobile games don't have a built-in benchmarking mode with sufficient repeatability I don't have confidence in publishing numbers through different methods. There are different methods but most require rooting a device which at the time a review should be posted is unfeasible and again iOS is problematic.