Compare Export Speeds and GPU Use for Apple Final Cut Pro, Adobe Premiere Pro and DaVinci Resolve on an M1 MacBook Pro

Last week, I compared how Apple Final Cut Pro, Adobe Premiere Pro and DaVinci Resolve use CPU cores when exporting files. (Link)

This week, I took this one step further. I compared the export speed of each of these three, along with how they used CPU and GPU systems, as measured by Activity Monitor. The results surprised me.

NOTE: This article illustrates how many CPU cores Apple Final Cut Pro uses during various stages of editing: Import, editing, rendering and export.

MY COMPUTER

These tests were run on an 16″ MacBook Pro with an M1 Pro and 32 GB of RAM. As you’ll see from the charts below, no application maxed out the system.

THE PROJECT

For the project, I took a recent webinar, which was edited in Final Cut, exported an XML file and converted it using X2CC for use in both Premiere and Resolve.

• Project duration: 43:01
• Frame size: 1600 x 900 pixels
• Video codec: ProRes 4444

All my webinars are recorded in ProRes 4444. This one included picture-in-picture and graphics. It did not include titles, effects or color grading, because those don’t transfer between programs. The finished audio was a stereo pair created in Adobe Audition.

For the export, each program converted the ProRes 4444 file to ProRes 422, which required rendering. While ProRes is hardware-accelerated on Apple silicon systems, there is no way to tell, using Activity Monitor, whether any application used hardware acceleration in converting these files. My assumption is that they did. No files were rendered prior to export.

In all cases, the source files were played from a 4-drive RAID which supports a maximum read speed of 350 MB/second. Only Premiere came close to reading at this speed. The exported file was stored to the MacBook desktop, which supports a maximum of 5,000 MB/second write speeds. No software exported more than 100 MB/second.

EXECUTIVE SUMMARY

Both Premiere and Resolve were more than twice as fast as FCP. No application maxed out the system.

Premiere did an excellent job of reproducing the original project from Final Cut’s XML export. Resolve had problems. It did not support FCP still frames, accurate scaling, accurate cropping or muted audio levels from the XML file, though it did support picture-in-picture and the graphics well enough to run this test.

CAUTIONS

Comparisons like this are inherently tricky, because export speed is dependent upon many variables:

• Whether the project is rendered (faster) or unrendered (slower)
• Whether you export the project codec (faster) or convert it (slower)
• The frame size and frame rate of the project and whether you are altering it during export (slower)
• Whether the codec is hardware accelerated (faster)
• The storage bandwidth for both read and write
• How much movement there is between frames (more movement is slower)

THE DETAILS – FINAL CUT PRO

(GPU activity during export in Final Cut Pro.)

Final Cut took almost 25 minutes to export this 43 minute file.

This chart shows GPU activity during the export. While Activity Monitor does not show how many GPU cores were active, we can see that total GPU activity was about 50%.

NOTE: The spikes were caused by BusyCal accessing the web in the background.

(Red indicates CPU activity originated by the operating system. Green indicates CPU activity originated by user-controlled software.)

For CPU performance, Final Cut leaned heavily on the two efficiency cores and only one performance core. I found this surprising, as Apple keeps touting how it is optimizing Final Cut to take advantage of Apple silicon processors.

NOTE: FCP averaged 90 MB/second read and 30 MB/second write to storage during the export.

THE DETAILS – PREMIERE PRO

(GPU activity during export in Premiere Pro.)

Premiere took slightly more than 11 minutes to export this 43 minute file — 2.2X faster than FCP.

Premiere used more GPU horsepower than FCP, though certainly not all that was available.

(Red indicates CPU activity originated by the operating system. Green indicates CPU activity originated by user-controlled software.)

The big difference was that Premiere used all the CPUs to export this file, with very heavy use of the first five performance cores.

NOTE: Premiere averaged 250 MB/second read and 70 MB/second write to storage during the export. At the start of export, Premiere accessed the RAID at 350 MB/second for many seconds. No other software reached that speed.

THE DETAILS – DAVINCI RESOLVE.

(GPU activity during export in DaVinci Resolve.)

Resolve took slightly less than 10 minutes to export this 43 minute file — 2.5X faster than FCP. I’m not sure why the big spike in GPU activity at the start of export in Resolve. However, like Premiere and FCP, Resolve did not use all the GPU capability the M1 has available.

(Red indicates CPU activity originated by the operating system. Green indicates CPU activity originated by user-controlled software.)

While Resolve used all the CPU cores, the two efficiency cores were used the most and the last three performance cores were not very busy at all. That being said, Resolve finished more than a minute faster than Premiere.

NOTE: Resolve averaged 220 MB/second read and 70 MB/second write to storage during the export.

SUMMARY

All these speed tests need to be taken with a large grain of salt, because there are so many variables in render and export speed, especially when you move projects between software. However, there are three observations that stand out to me:

• Only Premiere used all ten CPU cores. FCP and Resolve only used some of them.
• No application used all the CPU power available to it.
• No application used all the GPU power available to it.

Keep this in mind as you consider how to configure your next computer system.

---

Bookmark the permalink.