The Intel CPU in my 15” MacBook Pro is rated for 45W TDP (Thermal Design Power), whereas my phone’s ARM CPU is rated for a measly 5W. That is a 10-fold difference overall or 20-fold difference per core. The speed difference between my quad-core Intel i7 4850HQ 15” MacBook Pro and my 8-core Exynos 7885 Samsung phone was an astounding 1000%!

Nvidia NVENC hardware encoder quality test for H.264Īs I mentioned, I did the FFmpeg H.265 software encoding comparison between ARM and x86.Discussion on Doom9 Forum about x265 encoding speed on ARM processors.Relevant resources referred to in this article: If you want to contribute to the article, make a pull request on GitHub. If you want to read the comments or leave a comment, do so under the YouTube video. Video has additional tips and illustrations. You can find the video narration of this article on YouTube: The speed difference between desktop and mobile encoding was astounding! In the article, you will discover why there is such a difference along with what the future holds for H.265 software encoding on mobile devices. I have used Homebrew on my Mac to install the FFmpeg package, whereas on my Android, I have used Termux. Both on my phone and my computer, all the CPU cores were utilized above 95% during encoding. The command uses pure software encoding and is multi-threaded by default. I have used the following FFmpeg command to compress a 4K H.264 footage to H.265 footage with a CRF (Constant Rate Factor) of 22, both on my Android phone and on my Mac: I recently had a crazy idea of using my idle mobile devices as a video render cluster. How Fast is H.265/HEVC Software Encoding on ARM CPUs (iPhone/Android) With FFmpeg?

Ffmpeg x265