posting benchmark scores is pointless, unless your just doing it for the fun of it, comparing with friends. just doing them to do them is pointless. also, they are a good way to see if changes that you made in your kernel affect the device in a good or bad way.yeah.. the benchmark should only be used under controlled environment.
Posting benchmark scores is really almost pointless.
Unlike desktop CPUs, our phones are PASSIVELY cooled, so we are at the mercy of the CPU and Ambient temperature (ie, if the CPU heats up, we have NO OTHER WAYS to cool it except force it to run at lower frequencies where as on desktop, with active cooling, the fans/coolant can go faster to carry away more heat keeping the CPUs at a constant operating temperature).
To make benchmark really meaning full, you will need to operate under the following conditions:
1. Run in airplane mode
2. disable all background operations
3. Note the ambient temperature of the ROOM or environment
4. Cool the phone to a known operating temperature
5. Run the benchmark and observe if thermal throttling kicked in or not and for how long
1. do not put it in airplane mode, do have a mobile data connection/wifi on. as many benchmarks need you to have a data connection.
2. right, kill whatever is working in your background.
3. i do benchmarks in warm and cool environments. some benchmarks prefer cooler temps, some prefer warmer temps. most benchmarks that stress the cpu will want cooler temps(except quadrant likes warmer temps), while the benchmarks that put more stress on the gpu then the cpu it wont matter.
4. generally, 20C-30C cpu temps are a good starting point. a little lower if you are benching the higher cpu speeds(above 1.7ghz).
5. i prefer to disable thermal throttling while benchmarking. with thermal throttle on, you just dont know what your device can actually do.
6. have a consistant setup for your benchmarking, keep it the same for every time you benchmark.
7. also, when benchmarking, lock in your cpu speed, dont let it scale. for example, i like to benchmark at 1836mhz, i set my high and low cpu speed to 1836. many of the benchmarks dont stress the device enough and the cpu will scale up and down. when it scales, you dont actually know what cpu speed its testing.
and this is absolutely right, be consistent and thorough..
repeat the same benchmark several times UNDER THE SAME EXACT conditions from steps 1~5
then take the average scores in order to really use the benchmark to judge performance...
Unless you work for Toms Hardware or Anantech, it's really time consuming and lot of preparation required to do a "simple" benchmark and make their results MEANINGFUL :angel:
Last edited: