,

Tip #646: When Does Video Compression Use the GPU?

… for Codecs & Media

Tip #646: When Does Video Compression Use the GPU?

Larry Jordan – LarryJordan.com

Most editing codecs use the GPU, most web codecs do not.

Topic $TipTopic

One of the never-ending debates is how to configure the “best” computer. While this question is unanswerable in general, when it comes to video compression, here’s what you need to know.

CPUs, in general, provide linear calculations – one calculation after the other.

GPUs, in contrast, provide parallel calculations – multiple calculations occurring at the same time.

  • GOP-based codecs benefit most from linear – CPU – calculations due to the structure of the GOP. If you need to create H.264 materials, the faster the CPU, the faster compression will complete.
  • I-frame codecs, on the other hand, benefit from the GPU because different frames can be calculated at the same time, then stitched together for the final movie.

This ability of I-frame codecs to use the GPU to accelerate render and export speeds is one of the reasons they are recommended for editing.


Please rate the helpfulness of this tip.

Click on a star to rate it!

1 reply
  1. Loren S. Miller
    Loren S. Miller says:

    Both ProRes and H.264/5 codecs give GPU’s a workout during live playback, which valuable when time doesn’t allow transcoding to ProRes or other i-frame codec. Two or three streams of 4K ProRes 422 (HQ) playback are possible on the dual GPU’s of a 2013 Mac Pro, served by a peppy RAID box like OWC’s Thunderbay!

    And when support is suddenly broken by a new release of Adobe Premiere Pro, (oh, let’s call it V 14.1 April 2020?), you really notice the choppy playback, frozen frames, etc. Fortunately Adobe makes it easy to backstop an install.

    Another way to check how your GPU’s are or are not working with your material is to boot up Activity monitor, and open the window displays for CPU and GPU activity. There you can see whether the GPU’s are supported and how much.

    Reply

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply to Loren S. Miller Cancel reply

Your email address will not be published. Required fields are marked *