The precision of UL benchmarks is usually better than 3% when testing under ideal conditions. This means that running the benchmark repeatedly on a consistently performing system in a well-controlled environment will produce scores that fall within a 3% range.
A score may occasionally fall outside the margin of error since there are factors in modern, multitasking operating systems that cannot be controlled completely. There are also devices that simply do not offer consistent performance due to their design. In these cases, you should run the benchmark multiple times, and take an average or a mode of the results.
Score variance on mobile devices
Some high-performance mobile devices use aggressive thermal throttling to control the temperatures of the processors inside the device. As the device heats up it intentionally slows the processor to avoid overheating, which in turn leads to lower performance scores on successive runs. Measuring the effects of thermal throttling is another interesting way of using UL benchmarks to compare the performance of different devices.
To reduce this effect, we recommended waiting 15 minutes between runs to allow the device to cool down. You should also unplug the device during testing as the battery charging circuitry also generates heat. Try running the test in a cool environment. Don't run the benchmark in direct sunlight or near sources of heat.
Best practice benchmarking
For more tips on getting the best results from your device, please see our How to Benchmark page.