- Mar 11, 2010
- Reaction score
Please note. All benchmarks were ran on BB 0.4 with the most memory I could keep available for the tests.
I mainly did this because of the threads that were popping up on which kernel to use. Granted each phone might be different but these are my results.
Just some benchmark results. Not much of a difference.
Quadrant Score. Each benchmark was ran 2 times.
1513 - Chevy ULV 1.2ghz
1563 - P3 LV 1.2ghz
1546 - P3 Normal Volt 1.25ghz
MFLOPS Scores. Each benchmark ran 2 Time. Thus far the first benchmark run was also the best results pf the two.. Surprisingly both P3 kernel ran the same.
17.406 - Chevy ULV 1.2ghz
16.812 - P3 LV 1.2ghz
16.812 - P3 Normal Volt 1.25ghz
So from my conclusion the best Kernel to run out of the three is P3 LV 1.2Ghz. No reason to run P3 1.25, same scores, however some may benchmark a little higher then others but i assure you it wont be by much and it WILL drain battery life a little faster.
Hopefully this helps some people out. If anyone wants me to some more Kernel testing post up.
The only issues is your conclusion is based on faulty results. As part of my research on Overclocking the Droid, I have found one thing to be an absolute fact. When using Linkpack to test a kernel, you must remove EVERY application possible and boot the phone with only the OS being the choke point if you are going to only run the test once or twice. Otherwise you must run the Linpack benchmark 10 to 20 times and look for the best MFLOPS. The reason for this is simple. Linkpack is a straight CPU benchmark. That means it can only test the clock rate of the CPU and nothing else. And since ALL kernels have to obey the same laws for running on the OMAP3430, the results will ALWAYS be the same if you remove your background apps and streamline the OS to the bare minimum needed to boot the phone. So a P3Droid, Chevy, Bekit, etc kernel that is clocked at 1.2Ghz will ALWAYS give the same result in linkpack if you setup the test properly.
And if you wish to test this theory, feel free to use the following formula under any Froyo 2.2 ROM and kernel combination with your standard load out of apps. Set Advanced under SetCPU to 32000, 20, 0, 0. Turn off all profiles. Set Min and Max to top frequency choice. Run the benchmark 20 times quickly, remember the top MFLOPS and then take MFLOPs / .01606 = CPU Frequency.
Your current results come out to:
17.406 - 1083.81 Mhz
16.812 - 1046.82 Mhz
16.812 - 1046.82 Mhz
That means your test results have a 10% error right now. Which is usually considered a large enough discrepancy to invalidate them as useful for making an informed choice.
No one at present has developed a benchmark that is truly useful to pick the best kernel for a particular Droid with a particular load out of apps. I have spent the past four months trying to figure out a way to do just that. And every time I find the flaw in the testing.
BTW, Quadrant has the same issue as Linkpack. The very fact that you can get different results based on running it 10 times, should prove that beyond question.