I got NVidia NVENC encoding working, and an optiplex with a lower end Quadro video card beat an 4 core Xeon.<\/p>\n\n\n\n
Trick is you need to compile your own<\/p>\n\n\n\n
Here are the details<\/p>\n\n\n\n
Here are the results of different versions of ffmpeg transcoding from a 10 minute 26\nsecond video from a recent football game. \nThe file is 1.2GB large and has a frame of 1920×1080 with a rate of\n29.97FPS<\/p>\n\n\n\n
The parameters used in ffmpeg are as follows<\/p>\n\n\n\n
AUDIO\u00a0 |\u00a0 -acodec libmp3lame -ab 160k
VIDEO\u00a0 |\u00a0 -vcodec libx264 -vprofile high -preset slow -b:v 2500000 -vf scale=-1:720 -threads 0 -f mp4<\/pre>\n\n\n\nFor the most part they did not change, when I used the\nhardware encoding I did have to change the video slightly to<\/p>\n\n\n\n
VIDEO | -vcodec h264_nvenc<\/strong>\n-vprofile high -preset slow -b:v 2500000 -vf scale=-1:720 -threads 0 -f mp4<\/pre>\n\n\n\nTo use the hardware encoding<\/p>\n\n\n\n
After transcoding files were down to 1280×720 using an h.264\ncodec and were 210MB in size.<\/p>\n\n\n\n
The computer was a much older Dell Optiplex 760, Intel Core2\nDuo CPU E8400 @ 3.00GHz, 8GB of RAM and a NVIDIA Corporation GK107 [NVS 510]\n(rev a1) video card which is based on NVidia Kepler technology which does\nhardware h.264 encoding and decoding
\n<\/strong><\/p>\n\n\n\nPrecompiled RPMS<\/h1>\n\n\n\n
I have been using these binaries for a while now, downloaded\nfrom http:\/\/li.nux.ro\/repos.html<\/a> and they were the slowest.<\/p>\n\n\n\n
Time: 31min\n55sec<\/strong><\/p>\n\n\n\n
Precompiled Binaries<\/h1>\n\n\n\n
Was recently introduced to this, a simple statically compiled\nbinary from https:\/\/johnvansickle.com\/ffmpeg\/<\/a><\/p>\n\n\n\n
Extremely easy to install and use<\/p>\n\n\n\n
Time: 26min\n42sec<\/strong><\/p>\n\n\n\n
Compiled from source<\/h1>\n\n\n\n
Here are stats from when I compiled from source the latest\nversion. Kind of tricky and there were\nspeed bumps. This run was based on software\nencoding, statistically the times are similar to the precompiled binaries<\/p>\n\n\n\n
Time: 27min\n07 seconds<\/strong><\/p>\n\n\n\n
Compiled using hardware encoding<\/h1>\n\n\n\n
This is the best performance \u00a0of all, using nvenc encoding, it took a 5th<\/sup> of the time WOW<\/p>\n\n\n\n
Time: 5min 24sec<\/strong><\/p>\n\n\n\n
Precompiled RPMs on workstation<\/h1>\n\n\n\n
For yucks, I tried this on a workstation class HP Z series\ncomputer. <\/p>\n\n\n\n
CPU is Intel(R) Xeon(R) CPU E31230 @ 3.20GHz and 12GB of RAM<\/p>\n\n\n\n
Time: 11min 59sec<\/strong><\/p>\n\n\n\n
I was so impressed with what this tiny Quadro card could do I went out and got a beefier Quadro K4000 which could only go on the workstation. But the same file was done under 3 minutes (2 min 55 sec)<\/p>\n\n\n\n
Now I am going to give a large 24GB file a try, which on a system I had up last year took nearly 10 hours (9 hours 50 minutes 52 secs) which took on the Optiplex under 2 hours (1 hour 58 minutes 15 secs).<\/p>\n\n\n\n
Maybe this will be done before the night is over and I can post an update.<\/p>\n\n\n\n
Weight: 279.8<\/p>\n","protected":false},"excerpt":{"rendered":"
I got NVidia NVENC encoding working, and an optiplex with a lower end Quadro video card beat an 4 core Xeon. Trick is you need to compile your own Here are the details Here are the results of different versions … Continue reading