Android smartphone Benchmark Charts; CPU, GPU, memory & storage performance, and touch latency

Here’s my benchmark results for each SoC and device.

List of Benchmark Results

AnTuTu benchmark is a popular Android benchmark app, but often falls victim to benchmark boosts, plus it is now excluded from the Play Store.

To minimize the impact of benchmark boosting, I will add my own measurements on multiple applications in turn.

In addition, the individual articles also include Genshin game performance data measured by WeTest PerfDog (some are in Japanese).

Benchmarked devices (model name numbers match the * numbers in the graph below)

* Basically, measurements are taken at the highest resolution and refresh rate that can be set on the device.

About Benchmark Boost

In particular, Chinese smartphones sometimes use so-called “benchmark boosting,” in which the CPU clock is fixed to maximum only during benchmarking to make the numbers look better.

Apart from benchmark boosting, there is also a behavior called “throttling” that limits CPU/GPU operation of apps by identifying them by package name, even though the benchmark results are normal.

I do not consider “throttling” to be a major problem if there is a workaround for it, since it is normal for heat generation to increase as performance improves, but I consider “benchmark boosting” to be a malicious behavior that makes benchmarking meaningless and prevents fair comparison with other devices.

Why can’t 3DMark benchmarks measure true 3D performance and thermal throttling? – AndroPlus

Many people, including smartphone manufacturers, use benchmarking applications as an indicator, but in fact, the Play Store distributed version may not be…

AndroPlus

CPU Performance

CPU performance is measured by Geekbench 6.

Single Core

Multi core

Below is the results of Geekbench 5.

Single Core

Multi core

Graphics Performance

GPU graphics performance is measured by Wild Life, a cross-platform test of 3DMark.

Performance for everyday use

Comfort in everyday operations is measured with PCMark Work 3.0; the Work 3.0 benchmark checks how well it handles common tasks such as web browsing, video editing, document and data manipulation, and photo editing.

It measures overall performance, including GPU and storage performance as well as CPU.

It needs to be at least 7000 or higher for comfortable everyday use.

Power efficiency

The average consumption (mW) divided by the average frame rate when playing Genshin at the highest quality and 60 FPS (measured with WeTest PerfDog) setting is used as an indicator of how efficient the device is.

Less is better, and the lower the power consumption per FPS, the better the device can maintain a smooth display with less battery consumption.

It shows not only the performance of the SoC itself, but also how well it is optimized in software.

Even a model that loses to a high-end SoC in terms of SoC stand-alone performance may be able to deliver stable performance in actual game play.

Here are the numbers for World of Tanks Blitz at 120 FPS and highest quality setting.

Chinese smartphones are often optimized only for Chinese games such as Genshin, but if the power consumption seems to be low even for WoT from Europe, it means that the performance is likely to be good in general, not just optimized for Genshin only.

Note:

• AI Frame Rate Stabilizer ON…51
• Ultra-steady frames ON…46, 49, 54
• Performance mode ON…53, 54, 63
  (Performance mode is required to unthrottle)

Storage and memory performance

Storage and memory performance is measured by CPDT Benchmark.

Sequential write

Sequential read

Random write

Random read

Memory copy

Tap and screen latency

I measured tap latency (delay in response to touch) and screen drawing latency with the WALT Latency Timer (developed by Google).

I used the result of about 30 taps and measured the screen drawing latency with maximum brightness at the center of the screen.

The total of the latency between touch to kernel, kernel to Java and screen drawing latency at ACTION_DOWN is the “latency taken until the screen display changes after touch” and the less the better.

Apparently, devices with MediaTek SoC are prone to delays in screen rendering.

Drag latency

I also measured latency when dragging with the WALT Latency Timer.

Screen brightness

I measured screen brightness with a Vici LX-1336B luminance meter.

White images and white HDR videos were used for the measurements, and the brightness was measured near the center of the screen with automatic brightness adjustment turned on.

While the nominal values often show off peak values in a big way, the actual values are often not that high.

It is said that it is difficult to see indoors with 400-500 nits and outdoors with no more than 1200 nits, but the number of smartphone models that can produce 1200 nits is still limited.

Conclusion

Even if the same standards are used, newer models tend to be more improved and optimized, resulting in better performance.

The middle-range models in terms of classification can have performance that exceeds that of high-end models from a few years ago, and it is good that we live in an era when high-performance smartphones are available at low cost.