How to run a fair phone benchmark
By MobiBench5 min read
A fair phone benchmark is one where the only thing that changes between runs is the thing you want to measure. Start with a cool device, close background apps, keep settings the same, avoid charging and hot rooms, and run the test more than once before you trust or compare a result.
Why conditions matter
A benchmark measures how a device performs at that moment, under those conditions. A phone's performance depends on its temperature, its power state, what else is running and how it is configured. Change any of those and the result changes too, even on the same phone.
That doesn't make benchmarks unreliable. It means a score only makes sense together with the conditions it was measured in. The goal of a fair benchmark is to keep those conditions steady, so that differences in results come from the device, not from the room or the settings.
Before you start: a checklist
- Let the phone cool down. Start when the device is at room temperature and doesn't feel warm. If you have just been gaming, charging or using the camera, wait a while first.
- Close background apps. Apps running in the background compete for the CPU, the GPU and memory. Close what you don't need, and avoid starting downloads or updates.
- Unplug the charger. Charging produces heat inside the phone, and some devices manage performance differently while charging. Test on battery.
- Avoid hot environments. Don't test in direct sunlight, in a hot car or on a bed or sofa, which trap heat. A hard, flat surface in a room at normal temperature is ideal.
- Keep settings consistent. Use the same screen brightness, the same refresh rate setting and the same power or battery-saving mode for every run. Battery saver modes often limit performance.
- Charge the battery reasonably first. Some phones change their behaviour when the battery is low, so avoid testing when it is nearly empty.
- Keep the case consistent. A thick case can slow heat transfer. Either always test with it or always without it.
During the test
Once a test starts, leave the phone alone. Using other apps, answering messages or switching screens adds load and interrupts the measurement. Keep the phone on the same surface for every run, since holding it in your hand changes how heat leaves the device.
Load tests are meant to push the device hard, so it will get warm. MobiBench shows a warning about device load and heating before the CPU stress test begins. If the phone becomes uncomfortably hot, stop the test and let it cool down. A benchmark is never worth stressing a device that already feels too hot.
Run it more than once
A single result can be thrown off by something you didn't notice, such as a background sync or a warm starting temperature. Run each test at least two or three times, letting the phone cool down in between, and compare the results.
What consistent results look like
If repeated runs land close to each other, you can trust the figure. If one result is far from the others, treat it as an outlier and look for the cause before using it.
Back-to-back runs tell a different story
Running tests back to back, without cooling down, is not a fair comparison, but it is still useful. If scores fall from one run to the next, you are seeing heat build-up and throttling. That shows how the phone behaves during long sessions. Our guide to thermal throttling explains what is happening.
Comparing devices fairly
Comparing two phones is only meaningful when both were tested in the same way. Before you compare, check that:
- Both used the same app and the same test.
- Both started cool, at a similar room temperature.
- Both had similar settings: brightness, refresh rate and power mode.
- Each result is a typical value from several runs, not one lucky best.
Be careful with numbers shared online without context. You usually don't know the temperature, the settings or the software version behind them. Even two phones of the same model can differ slightly, because individual chips vary and software updates change behaviour.
How MobiBench helps
MobiBench is designed around measured results. The CPU stress test and the GPU test each run for about 60 seconds, long enough for heat to play its part, and scores are calculated from measured performance rather than looked up from the device name or brand. The GPU test reports the temperature and the peak temperature, so you can see how much heat a run produced.
Follow the checklist above, note your conditions, and your results will be much easier to trust and compare.
Frequently asked questions
How many times should I run a benchmark?
At least two or three times, letting the phone cool down between runs. Use a typical result rather than the single best one.
Should I run a benchmark while the phone is charging?
No. Charging adds heat inside the phone and some devices manage performance differently while charging, which makes results harder to compare.
Why is my score different from someone else's with the same phone?
Temperature, settings, background apps and software versions all affect results, and individual chips vary slightly. Unless the conditions match, small differences are expected.
Does battery saver mode affect benchmark results?
Usually, yes. Power-saving modes often limit performance, so keep the same mode for every run you want to compare.
See it on your own phone
MobiBench is a free Android app with a 60-second CPU stress test, a 3D GPU test and live system monitoring. Every test runs on your device.