Suggestions for next x264 benchmark

Discussion in 'Processors, Motherboards & Memory' started by turbojet, Oct 17, 2013.

  1. turbojet

    turbojet Newbie

    First of all thanks for providing hardware sites and users a standard x264 benchmark. I thought of some things that would make the benchmark simpler, more accurate and more reliable.

    1. Remove avisynth and dgindex dependency by feeding video directly to x264, this is faster and gives a more accurate difference of optimized x86 to x64 avisynth. This also allows osx and linux to use the benchmark without hassle.

    2. Remove cpu-z and admin rights dependency by using 'wmic cpu' to get cpu model and speed on windows, linux has 'cat /proc/cpuinfo', osx has 'sysctl -a | grep machdep.cpu' . Temp sensor isn't available without admin rights afaik but is it important if the benchmark completes?

    3. The 5.0.1 video penalizes turbo boost cpu's by throttling when encoding during the fades, there are many, which drop the cpu usage a lot, in some cases < 30%. Using a video that more represents typical usage without so many fades would make for a more accurate test.

    4. Choice.exe can be replaced with batch code for example:
    echo 1. 32 bit
    echo 2. 64 bit
    echo 3. both
    SET Choice=
    SET /P Choice=Which benchmark do you want to run?:
    IF NOT '%Choice%'=='' SET Choice=%Choice:~0,1% )
    IF /I '%Choice%'=='1' ( x264.exe ...)
    IF /I '%Choice%'=='2' ( x264_64.exe ...)
    IF /I '%Choice%'=='3' ( x264.exe ... && x264_64.exe ...)

    5. x264 resizing could help users decide which cpu they want to use. Some cpu's are faster at HD then SD for example. Maybe a choice of 720x, 1280x and 1920x. Is it too soon for 4K encodes?

    6. Looping the video in avisynth is a good idea, without it the three main os's can concatenate a short video to keep the download small and still get an accurate test.

    7. Clean up afterwards, is there any need for the video output? Data could be outputted to shell window instead of rtf, could even make an option to send results to techarp without dependencies if such a system was implemented.

    Basically all that would be in the download would be the scripts (windows, osx?, linux?), x264 binaries and the test video. I'm willing to help with windows and linux scripting if needed. Thoughts?
     
  2. Adrian Wong

    Adrian Wong Da Boss Staff Member

    Your help would be greatly appreciated. Pinging Dashken! :thumb:
     

Share This Page