There is a 9-fold difference in the thermal envelope of these CPUs, so that must explain the massive difference in performance, right? Not really.

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 hevc to h264