I tested Kdenlive render performance on my quadcore system

My system:

Code: Select all

Arch Linux 4.1.13-1-lts x86_64
Mainboard  FOXCONN A7GM-S 2.0
CPU        AMD Phenom(tm) 9550 Quad-Core Processor, 4 x 2200 MHz, 64-bit
RAM        7.8G

ffmpeg 1:2.8.3-1
libx264 2:148.20150725-1
mlt 0.9.8-3
x264 2:148.20150725-1
kdenlive v15.11.90-55-g8580c88 * master

Results:

Code: Select all

AMD Phenom X4 9550 (4x 2.2GHz, 64bit), 8GB RAM
+----------+-------------+--------------+-----------------+
| threads= | Average (s) | Faster than  | Faster than     |
|          |             | 1 thread (%) | n-1 threads (%) |
+----------+-------------+--------------+-----------------+
|        1 | 362.54      |              |                 |
|        2 | 230.01      | 36.56%       | 36.56%          |
|        3 | 218.33      | 39.78%       | 5.08%           |
|        4 | 216.78      | 40.20%       | 0.71%           |
|        5 | 215.74      | 40.49%       | 0.48%           |
|        6 | 215.79      | 40.48%       | -0.02%          |
|        7 | 215.76      | 40.49%       | 0.01%           |
|        8 | 215.55      | 40.55%       | 0.10%           |
|        9 | 215.83      | 40.47%       | -0.13%          |
|       10 | 216.32      | 40.33%       | -0.23%          |
|       11 | 214.34      | 40.88%       | 0.92%           |
|       12 | 215.70      | 40.50%       | -0.64%          |
+----------+-------------+--------------+-----------------+

original data

Output video: http://youtube.com/watch?v=bp_s7Zem93k

Method:
I used time -f "%e" to measure the render duration. I did each render 3 times and calculated the average.

The parameters other than threads were:

Code: Select all

pid:10815 in=0 out=2541 $MELT atsc_1080p_25 avformat - properties=x264-medium preset=veryfast crf=20 ab=160k real_time=-1

Project:


Conclusion:
Before the test, I claimed that using more than 4 threads improves render performance significantly (because I could feel how my computer becomes slower). This test seems to show that it does not. The test has weaknesses however.

Weaknesses:
1. This was tested while I was away from the computer but I often render while using the computer for other tasks. Perhaps the test is not representative of performance during such use.
2. The project is relatively simple and short but not completely simple (just a cut job for example is even simpler). Perhaps more complex or less complex projects would have significantly more or less multi-thread benefits.

Benchmark your computer:
http://archive.org/details/kdenlive-benchmarks (300MB zip with 1 sh file, 1 mlt file, resource files, freely licensed)

I recommend you run the .sh script it in a terminal after rebooting, turning off unnecessary software and leave the computer be for a few hours.

To provide your system and software specs, copy paste the following into a terminal (I hope it works):

Code: Select all

cpu_speed=`sudo dmidecode --type Processor | grep Max\ Speed | sed -e 's/.*Max Speed: //'` ;\
echo ;\
uname -a ;\
cat /proc/cpuinfo | grep model\ name | tail -n 1 | sed 's/.*: //' ;\
cpu_arch=32-bit ;\
if [ `grep -o -w 'lm' /proc/cpuinfo | sort -u` == "lm" ]; then cpu_arch=64-bit; fi ;\
echo $cpu_arch ;\
echo `nproc` cores a $cpu_speed ;\
free -h | head -n 2 | tail -n 1 | sed -e 's/G.*/G/' -e 's/.*   //' ;\
kdenlive --version ;\
melt --version |head -n 1 ;\
ffmpeg |& head -n 1