iMac 5K vs Mac Pro, Video Transcode 4K to 1080P

IntegrityChecker(tm) v1.2.3 64-bit, diglloydTools 2.2.7b

iMac 5K 4.0 GHz, 32GB, 1TB SSD, AMD Radeon R9 M295X 4GB GDDR5 -versus-
2013 Mac Pro 3.3 GHz 8-core, 64GB, 1TB SSD, D700

Transcoding 4K video to 1080p is a computation intensive process. The 2013 Mac Pro Xeon CPU lacks the specialized video instructions that the CPU in the iMac 5K and 2013/2014 MacBook Pro have. Thus in spite of its dual GPUs, the 2013 Mac Pro delivers a pathetic performance for this simple task. More elaborate video operations might not behave similarly; it gets very specific. And see comments below.

See also Transcoding Panasonic GH4 UltraHD 4K Video to 1080p: Surprising Speed on 2013 Mac Pro with Final Cut Pro and QuickTime.

A simple transcode of 4K video to 1080p using Quicktime Player (Export).

Transcode 4K video clip to 1080p (OS X QuickTime Player)
iMac 5K 4 GHz Radeon M295X vs 2013 Mac Pro 3.3 GHz 8-core D700

The above is obviously not a professional video / high-end test. But for many users, even the concept of transcoding is new, and simplicity/speed rule. It is thus relevant to the vast majority, but not to video professionals.

Dez writes with a pro perspective:

The reason for the "pathetic" results "in spite of the dual GPUs", is that Quicktime Player doesn't leverage those GPUs, but it does however, tap into the Quick Sync core in the i3, i5, i7 series chips. Thus the iMac transcode was aided not only by the extra clock speed, but also by the special h.264 encoding core built into the CPU.

Transcoding video with Quicktime player is a poor test of performance for many reasons, the largest being that, unlike most pro video apps (Avid Media Composer aside), it does not leverage GPU acceleration at all, and it's implementation of multi-threading is notoriously poor. A better "real-world" test is to use something like Davinci Resolve, Compressor or Adobe Media Encoder, all of which use GPU acceleration for encoding files. Additionally, Quick Sync only works with h.264, so if you need to transcode to a different codec, you're stuck with vanilla CPU number crunching.

Another reason to avoid the Quick Sync is that the quality of the files are poor when compared to straight CPU (unaccelerated) crunching with a good encoder like Sorenson Squeeze or GPU accelerated encoding via something like Resolve or Media Encoder.

Really, the best test if you want to appeal to pro users is to transcode from 4k RED Camera or Arri Alexa raw files into an intermediate codec such as Apple ProRes 4444 or Avid DNxHD 345x using an application like Davinci Reslove.

If you're not trying to appeal to a pro user, then do the same test to ProRes 422 HQ using Final Cut Pro X or Compressor, which is a common workflow for the DSLR and Go Pro crowd.

MPG: the best test of performance is one relevant to the user and usage scenario.