FCP X and CPU Cores.....

[LethalWolfe]

About the only clear cut indication is comparing export times with GPU enabled vs disabled on the same machine and material. Unfortunately there is no easy way to do that on a Mac, short of pulling the GPU card on a Mac Pro tower or Hackintosh. That would be an interesting test; I wonder if anyone has ever done it?

I didn't do an extensive test, but I used Adobe Media Encoder to do three exports of a 5min chunk of a project I'm working on. In AME (and PPro) you can choose whether or not you want GPU acceleration on or off. My machine is a 2009, 2.66GHz 8-core MP (4,1) with an Nvidia 660 Ti GPU (2gig). The exports pulled from the original media (no preview/render files were used). For the export settings i used the H.264, Vimeo 1080p24 preset.

1. CPU only - 7:37
2. OpenCL - 6:57
3. CUDA - 6:50

The intricacies of programing and what goes on under the hood is not in my wheel house so I present this info w/o comment.

You can download a free trial of PPro (AME comes bundled with PPro, it can't be downloaded by itself AFAIK) if you want to do more exhaustive comparisons on your own.

 

[linuxcooldude]

Well I double checked my settings with Apples Compressor and it was set for MPEG 4-Basic. So I put same video file in H.264 single pass. CPU load jumped to 30% instead of 13% but encoded faster from 6:45 to 5:37. Length of file plays 6:17. Still not bad.

 

[joema2]

...In PPro....you can choose whether or not you want GPU acceleration on or off. My machine is a 2009, 2.66GHz 8-core MP (4,1) with an Nvidia 660 Ti GPU (2gig). The exports pulled from the original media (no preview/render files were used). For the export settings i used the H.264, Vimeo 1080p24 preset.

1. CPU only - 7:37
2. OpenCL - 6:57
3. CUDA - 6:50

OK, I have tested export time on the same machine (my 2015 top-spec iMac 27) using both Premiere CC 2015.4 and FCPX 10.2.3, both running on OS X 10.11.6.

For Premiere CC I used software rendering, GPU accelerated OpenCL and GPU accelerated Metal. In all cases I pre-rendered the timeline before exporting since we are interested in export & encode time, not render time for effects. The issue is whether GPU acceleration can help export and encode performance.

I don't see any evidence that GPU acceleration helps this, which is in line with my previous explanation. FCPX is vastly faster than Premiere CC at exporting to H264 but this is not due to the GPU but likely due to FCPX using Quick Sync.

Test results:

Media: 5 min H264 1080p/29.97 video from Canon 5D Mark III

Premiere CC export parameters: H264, Preset: Match source - high bitrate

Export time using GPU OpenCL rendering: 3:50
Export time using software rendering: 3:51
Export time using Metal rendering: 3:51
Export time using Metal rendering, preset: HD 1080p 29.97: 6:02

FCPX export time, same clip, fast encode: 1:05
FCPX export time, same clip, better quality: 2:16

There was no perceptible quality difference between any of these, both from comparing frame grabs and comparing video motion and color of the exported material.

FCPX is about 3.5x faster than Premiere CC when exporting to H264 at similar quality levels on the same hardware. This is regardless of whether Premiere is using software rendering, GPU-accelerated OpenCL rendering or GPU-accelerated Metal rendering. Likely the main reason for this is FCPX using Quick Sync, which is unfortunately unavailable on Xeon.

Comparing playing time to export time, FCPX exports 1080p to H264 at about 5x real time playback, whereas Premiere CC exports at about 1.3x real time -- whether software or GPU-assisted rendering is used.

This indicates (at least on this hardware) the GPU is not helping H264 export or encode.

Hardware: 2015 iMac 27, 4Ghz i7-6700K, 32GB, 1TB SSD, M395X, 16TB Thunderbay 4