====== Benchmark of XU4 with Active Cooler and XU4Q with Passive Cooler ======
This wiki page shows benchmark test results to analysis and compare their system performance\\
with ODROID-XU4 and ODROID-XU4Q that equips with a passive cooler.\\

{{:en:201704250351126731.jpg?423|}}

  * [[en:xu4_benchmark_xu4_xu4q&#benchmark_1_-_thermal_and_performance_using_sysbench| 1. Thermal and Frequency Performance Comparison]]
  * [[en:xu4_benchmark_xu4_xu4q&#benchmark_2_-_android_antutu | 2. AnTuTu Performance Comparison ]]
  * [[en:xu4_benchmark_xu4_xu4q&#benchmark_3_-_kernel_build_time| 3. Kernel Build Time Performance Comparison]]

We've tested three categories of benchmarking utilities and here is a brief result for each test cases.\\

**sysbench**\\
- to investigate and compare thermal/cpu frequency performance\\
- Sysbench running time increased by 16% and 6% for each cases of cpu max 2GHz and 1.8GHz\\
**AnTuTu**\\
- for examining a variety of categories performance\\
- The decreasing score value of XU4Q from XU4 is just 2% \\
**Kernel build**\\
- to check the hardest stress case\\
- About 5m 17s increase of the total execution time (20%)\\


Except the hardest case like kernel build, the differences are under just 10% or lower.\\

==== [ Benchmark (1) - Thermal and Frequency Performance Comparison ]====

=== Environment and Instructions ===

  * **sysbench** 100,000 iterations
<code>
# sysbench --test=cpu --cpu-max-prime=100000 --num-threads=8 run
</code>

  * setting the max cpu frequency of CPU4-7
<code>
# echo 1800000 > /sys/devices/system/cpu/cpu4/cpufreq/scaling_max_freq 
# cat /sys/devices/system/cpu/cpu4/cpufreq/scaling_max_freq 
1800000
</code>

  * All of 8 cores are ON

  * Script
<code>
#!/bin/sh

echo "Temp, FreqCPU0, FreqCPU4, FreqCPU5, Freq6CPU, FreqCPU7 "
while true :
do
     t=`cat /sys/devices/virtual/thermal/thermal_zone0/temp`
     t1=$(( $t/1000))

     f0=`cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq`
     f0_r=$(( $f0/100000))

     f4=`cat /sys/devices/system/cpu/cpu4/cpufreq/scaling_cur_freq`
     f4_r=$(( $f4/100000))

     f5=`cat /sys/devices/system/cpu/cpu5/cpufreq/scaling_cur_freq`
     f5_r=$(( $f5/100000))

     f6=`cat /sys/devices/system/cpu/cpu6/cpufreq/scaling_cur_freq`
     f6_r=$(( $f6/100000))

     f7=`cat /sys/devices/system/cpu/cpu7/cpufreq/scaling_cur_freq`
     f7_r=$(( $f7/100000))

     echo $t1, $f0_r, $f4_r, $f5_r, $f6_r, $f7_r

     sleep 2
</code>

=== Test Case 1 - Max CPU Frequency 2.0GHz case ===

Here is sysbench test result in case of max cpu frequency 2.0GHz.\\
  * Running Time : about 16% increase - XU4Q / XU4 = 420 / 362 = **116%**
  * Average CPU Frequency : about 22% loss - XU4Q / XU4 = 1.48 / 1.89 = **78%**
{{:en:1_xu4_max_2ghz.png?700}}{{:en:2_xu4q_max_2ghz.png?700}}

=== Test Case 2 - Max CPU Frequency 1.8GHz case ===

  * Running Time : about 6% increase - XU4Q / XU4 = 392  / 368 = **106%**
  * Average CPU Frequency : about 11% loss - XU4Q / XU4 = 1.61 / 1.8 = **89%**
{{:en:3_xu4_max_1_8ghz.png?700}}{{:en:4_xu4q_max_1_8ghz.png?700}}

=== Test Case 3 - Finding the threshold not to drop under max frequency ===

{{:en:5_xu4q_max_1_2ghz.png?700}}

==== [ Benchmark (2) - Android AnTuTu ] ====

Here is Android AnTuTu benchmark result for examining common and a variety of categories' score.

=== Environment ===
**
OS - LineageOS-14.1 Android 7.1.1 Nougat \\
Resolution - 1280x720 \\
The benchmark has been tested at twice.
**
=== Result ===
** ODROID-XU4 **\\
{{:en:screen-xu4_1.png?direct&600|}}\\
\\
** ODROID-XU4Q **\\
{{:en:screen-xu4q_1.png?direct&600|}}\\
\\
=== Detail score ===
^ Category         ^  XU4      ^  XU4Q     ^      % ^
| 3D               |  9823     |  9843.5   |   0.21 |
| Marooned         |  6337     |  6347     |   0.16 |
| Garden           |  3486     |  3496.5   |   0.30 |
| UX               |  21681    |  21443    |  -1.11 |
| Data secure      |  4976.5   |  4975     |  -0.03 |
| Data process     |  3517.5   |  3496.5   |  -0.60 |
| Strategy games   |  5677.5   |  5751.5   |   1.29 |
| Image process    |  4787.5   |  4505     |  -6.27 |
| I/O performance  |  2722     |  2715     |  -0.26 |
| CPU              |  23933.5  |  22774    |  -5.09 |
| Mathematics      |  5245     |  5248.5   |   0.07 |
| Common use       |  7798     |  7621     |  -2.32 |
| Multi-Core       |  10890.5  |  9904.5   |  -9.96 |
| RAM              |  5791     |  5543     |  -4.47 |
| Result           |  61228.5  |  59603.5  |  -2.73 |

\\
{{:en:xu4vsxu4q_antutu.png?direct&700|}}


{{:en:xu4vxxu4q_percentage.png?direct&600|}}



==== [ Benchmark (3) - Kernel Build Time ] ====

To check the hardest stress case, we did put kernel build test on XU4 and XU4Q.\\
The test instruction is as following.

<code>
$ make odroidxu3_defconfig
$ time make -j8
</code>

Based on the result, in this case, the kernel build time on XU4Q gets **around 20%** increase from XU4's one.

^  time   ^  XU4  ^  XU4Q  ^
|  real  |   25m 1.43s   |  30m 18.36s    |
|  user  |   149m 58.72s   |  184m 44.36s   | 
|  sys   |   22m 54.96s    |   27m 21.55s   |