FLB r10b vs r11 benchmarks

[Guest gameSTICKER]

flibblesan recently released version R11 of FLB-Froyo, and R11 introduced couple of kernel improvements:

BFQ: http://algo.ing.unimo.it/people/paolo/disk...description.php

BFS: http://en.wikipedia.org/wiki/Brain_Fuck_Scheduler

Smartass: http://www.ziggy471.com/2010/11/07/smartass-governor-info/

This benchmark is intended to test performance of R10b vs R11 and smartass vs ondemand CPU governors. These tests might not be perfect, but they give a glimpse of performance improvements. Maybe later, senior members can come up with better version of these tests :(

So I clean installed FLB r11 and did some benchmarking. All benchmarks were done with clean installs after wipe. The ROM was then left idle for 15 mins and restarted once. I only installed and run 'Usage Timelines' in background to check CPU usage. Phone was set on 'Airplane Mode'. Screen was set to 'ON' all the time. THIS BENCHMARK was used to perform tests as this one measures memory IO and disk IO performance too.

Bottom line, with R11 BFQ and BFS brings significant performance improvements. SmartAss on the other hand is meant to give more battery and not performance, and it plays well. Maybe some battery tests can come later. Please let me know what do you think :D

[Guest Nick Rhodes]

Hi,

This is a good start.

Does your benchmark represent best case, worst case, average or what scenario?

I suggest you do a number of test runs and take an average, and show, upper, lower q1, and q3 figures to get an idea of consistency.

I think that your test on row 26 could be erroneous for example, repeating that test would give us an idea.

Was the phone left to idle again between step of the benchmark ?

As I think if you don't your test is flawed, as opacity benchmark is the only one run from idle, where the rest the cpu will be running at more than idle as each test starts immediately.

What do those benchmarks mean to real world usage of the phone ?

For example taking CPU scores - how often will the average user max out their CPU to be able to see a difference in peak performance ?

Need to put these figures into context and significance and do some adhoc tests - e.g. does crazy birds run better ?

Battery tests would be very useful.

Cheers, Nick

[Guest Simon O]

Thank you for doing this!

Correct about smartass scheduler. It gives more battery as whenever the screen is off it drops the CPU speed down to the minimum. Obviously if you only use your phone for calls, texting and a little bit of facebook you'll value battery life more than pure performance.

[Guest Frankish]

So does the smartass have any effect with screen on? My latest cm7 install seems slower in games but i thought it might be the smartass gov....off to change and test back to ondemand!

[Guest Rotmann]

Great testing. I have found on CM7 that smartass really slows the system down, ondemand is much better with me. So yes, smartass really relies on battery economy. If you don't use fancy 3D effects or don't play, smartass is the way to go.

Edited March 22, 2011 by Rotmann