You keep asking me what kernel is best and I keep telling you to try them, well thats still the answer but you may want to check this out. Thank The folks at Evoreviews and DONATE. Its worth running over to their page also to donate or check other reviews or to download many many kernels. I absolutely take no credit for anything shown below...I should also say thanks to chogardjr the link was in his sig...
Linpack
The linpack benchmark is a measure of a system's floating point computer power. It measures how fast a computer can solve a system of linear equations like this Ax = b. Basically it's a big math problem that takes millions of floating point operations to solve. These millions of floating point operations are what is recorded as the final score in MFLOPS or millions of Floating Point Operations Per Second. Because this program is taken from a java written program and developed into an android program, it isn't a raw processor test like so many people value it as. It is more a reflection
of the state of the Dalvik Virtual Machine where the java code is executed. So software updates to the dalvik vm can greatly improve or destroy your Linpack scores on the exact same hardware. For example, the jump from Android 1.5 to 2.0 and to 2.2, upgrades to the dalvik vm have caused linpack scores to climb with each version. Linpack scores were taken with setcpu set to the maximum frequency in performance mode when available on each kernel. Five consequetive runs were done and the highest 3 scores were taken and averaged on each kernel. This way there was no doubt that any background process was interfering with the score or that the cpu needed to warm up, or any other nonsense. Of course higher
numbers in this test are better.
FPS2d
This is a test that a lot of people dont really understand it seems. Everyone brags "oh i get 60fps on fps2d on my super cool motorola pos" Well what they fail to understand is that 50fps with 1 frame of deviation is going to look loads better than 60fps with 15dev. It's a matter of stable fps so your eyes aren't able to tell a difference while doing things. Some of these kernels got 54fps with 1-2ish deviation and that should be looked at as much better than 56fps with 4-5ish deviation. Then again.... some people can't tell a difference between 30 and 60 fps lol. What I did for you was split this into two graphs. One showing the total fps each kernel got. And another showing the deviation. Again setcpu was set to 1152min and max on the "performance" setting when available, and five runs were done and the highest three scores are taken and averaged to make the score you see on the graph.
Quadrant
This is one of my favorites. It pulls a lot of different tests together all into one benchmark and shows you just what your phone is made of. FP ops are tested, data transfer, gpu, etc. and it gives you one combined composite score. This as long as there are no crazy drops in FPS2d can be looked at as a way to determine sheer speed of a certain kernel. Again no this test setcpu was set to 1152min and max on the "performance" setting when available, and five runs of Quad are done and the highest three scores taken and averaged into the score you see on the graph.
Screen on, Full frequency Power draw
The amount of power being consumed by the phone while the frequency of the processor is at it's max is what this test is all about. This should give you a baseline of what high cpu usage programs are going to be pulling while you run each kernel. Even though when they actually load it down instead of it just sitting at 1152mhz it will be considerably more current, this was an easy way to make it fair and see how each kernel will perform under EVEN load. For each value Setcpu was set to 1152mhz min and max, on "powersave" mode where available. The phone screen was shut off for five minutes, turned on (always set on full brightness, i refuse to look at my phone any other way unless in a pitch black room with a killer migraine) and left idle for five more minutes and the lowest consistent mA rating reported from CurrentWidget was recorded and graphed.
Screen on, Minimum Frequency Power Draw
This was testing the amount of power drawn while the phone is turned on, idleing with the screen on at the kernels minimum frequency. Some kernels had a minimum frequency of 128mhz, some were 244-256mhz, the ones that drop to 128 didn't seem to show any improvment in amperage draw as you can see. Maybe if the kernel makes can shave some mV's off by dropping to 128 it would help, but it appears to be running the same voltage, so no improvement is obtained by it. This test of course was done with setcpu set to min and max at 128 or 244-256 depending on kernel. Always in the "Power Save" mode when available. The phone's screen was shut off for five minutes to allow it to become completely idle. Turned on for five minutes to make sure nothing was going on, then the lowest consistent amperage over that second five minute period is what was recorded and graphed, as this is what your phone would be running at most of the time when it isn't doing
anything with the screen on.
Screen Off, Minimum Freq, Idle Power Draw
This was testing what your phone will be doing when it has nothing better to do lol. SetCPU was clocked to 128-256mhz min and max, depending on kernel again. The phone was put to sleep for 10 minutes and the minimum average current draw was slapped onto this here graph. Wohooo doggies! This is a very important number. My phone sits there eating my battery without doing a damn thing most of the day. This is a great indicator of which kernel is going to have amazing battery life, and which ones are going to eat you alive.
Linpack
The linpack benchmark is a measure of a system's floating point computer power. It measures how fast a computer can solve a system of linear equations like this Ax = b. Basically it's a big math problem that takes millions of floating point operations to solve. These millions of floating point operations are what is recorded as the final score in MFLOPS or millions of Floating Point Operations Per Second. Because this program is taken from a java written program and developed into an android program, it isn't a raw processor test like so many people value it as. It is more a reflection
of the state of the Dalvik Virtual Machine where the java code is executed. So software updates to the dalvik vm can greatly improve or destroy your Linpack scores on the exact same hardware. For example, the jump from Android 1.5 to 2.0 and to 2.2, upgrades to the dalvik vm have caused linpack scores to climb with each version. Linpack scores were taken with setcpu set to the maximum frequency in performance mode when available on each kernel. Five consequetive runs were done and the highest 3 scores were taken and averaged on each kernel. This way there was no doubt that any background process was interfering with the score or that the cpu needed to warm up, or any other nonsense. Of course higher
numbers in this test are better.
FPS2d
This is a test that a lot of people dont really understand it seems. Everyone brags "oh i get 60fps on fps2d on my super cool motorola pos" Well what they fail to understand is that 50fps with 1 frame of deviation is going to look loads better than 60fps with 15dev. It's a matter of stable fps so your eyes aren't able to tell a difference while doing things. Some of these kernels got 54fps with 1-2ish deviation and that should be looked at as much better than 56fps with 4-5ish deviation. Then again.... some people can't tell a difference between 30 and 60 fps lol. What I did for you was split this into two graphs. One showing the total fps each kernel got. And another showing the deviation. Again setcpu was set to 1152min and max on the "performance" setting when available, and five runs were done and the highest three scores are taken and averaged to make the score you see on the graph.
Quadrant
This is one of my favorites. It pulls a lot of different tests together all into one benchmark and shows you just what your phone is made of. FP ops are tested, data transfer, gpu, etc. and it gives you one combined composite score. This as long as there are no crazy drops in FPS2d can be looked at as a way to determine sheer speed of a certain kernel. Again no this test setcpu was set to 1152min and max on the "performance" setting when available, and five runs of Quad are done and the highest three scores taken and averaged into the score you see on the graph.
Screen on, Full frequency Power draw
The amount of power being consumed by the phone while the frequency of the processor is at it's max is what this test is all about. This should give you a baseline of what high cpu usage programs are going to be pulling while you run each kernel. Even though when they actually load it down instead of it just sitting at 1152mhz it will be considerably more current, this was an easy way to make it fair and see how each kernel will perform under EVEN load. For each value Setcpu was set to 1152mhz min and max, on "powersave" mode where available. The phone screen was shut off for five minutes, turned on (always set on full brightness, i refuse to look at my phone any other way unless in a pitch black room with a killer migraine) and left idle for five more minutes and the lowest consistent mA rating reported from CurrentWidget was recorded and graphed.
Screen on, Minimum Frequency Power Draw
This was testing the amount of power drawn while the phone is turned on, idleing with the screen on at the kernels minimum frequency. Some kernels had a minimum frequency of 128mhz, some were 244-256mhz, the ones that drop to 128 didn't seem to show any improvment in amperage draw as you can see. Maybe if the kernel makes can shave some mV's off by dropping to 128 it would help, but it appears to be running the same voltage, so no improvement is obtained by it. This test of course was done with setcpu set to min and max at 128 or 244-256 depending on kernel. Always in the "Power Save" mode when available. The phone's screen was shut off for five minutes to allow it to become completely idle. Turned on for five minutes to make sure nothing was going on, then the lowest consistent amperage over that second five minute period is what was recorded and graphed, as this is what your phone would be running at most of the time when it isn't doing
anything with the screen on.
Screen Off, Minimum Freq, Idle Power Draw
This was testing what your phone will be doing when it has nothing better to do lol. SetCPU was clocked to 128-256mhz min and max, depending on kernel again. The phone was put to sleep for 10 minutes and the minimum average current draw was slapped onto this here graph. Wohooo doggies! This is a very important number. My phone sits there eating my battery without doing a damn thing most of the day. This is a great indicator of which kernel is going to have amazing battery life, and which ones are going to eat you alive.
