FFmpeg h265 to WebM VP9 encoding comparison part 6

In this part 6 see FFmpeg comparison for encoding speed, fps, size and bitrate from an h265 (hevc) MP4 video file to VP9 WebM with both -deadline realtime and -deadline good with -crf values 15-30 and -cpu-used 0-5.

Part 1 was h264 -deadline good.

Part 2 was h264 -deadline realtime.

Part 3 was h264 with -deadline good compared to -deadline realtime.

Part 4 was h265 -deadline good.

Part 5 was h265 -deadline realtime.

Introduction

To see the encoding options for VP9 WebM with FFmpeg see my post here. The testing is using the constant quality option (crf) and the differences with different deadlines along with the cpu used parameter, these effect speed and the outcome quality.

The server

The server used for testing had dual Intel Xeon E5-2687W v4 CPU’s set @ 3.50GHz, meaning 24 cores and 48 threads in total. 251 GB of usable ram and NVMe drives. Thanks to exception0x876 at wishosting for the opportunity to use this.

The operating system was Ubuntu 18.04 and the FFMpeg version was 3.4.8-0ubuntu0.2.

The input

The input media is from jell.yfish.us with the H.265 testing file here. At 4k resolution (3840×2160) 29.97 fps, 30 seconds long, 250Mbps bitrate and file size of 914MB. This file comes in MKV format but was copied (-c copy) to MP4 for the testing.

The commands

The basis of the testing commands for this part is a combination of the two different -deadline values, obviously the -cpu-used and -crf values changed throughout. -row-mt 1 and -threads 48 stayed consistent through the whole testing process.

ffmpeg -i h265.mp4 -c:v libvpx-vp9 -row-mt 1 -threads 48 -cpu-used X -deadline X -crf X -b:v 0 outputXYZ.webm

WebM file size

WebM bitrate

Encoding time