Welcome, Guest
Username: Password: Remember me

TOPIC: iMac Pro vs 2017 iMac FCPX performance tests

iMac Pro vs 2017 iMac FCPX performance tests 09 Jan 2018 23:30 #93097

  • joema
  • joema's Avatar
  • OFFLINE
  • Platinum Boarder
  • Posts: 859
  • Thank you received: 185
  • Karma: 14
Here are some simple performance tests I did on the base 8-core Vega 56 iMac Pro vs my 2017 top-spec iMac 27. The iMP was on FCPX 10.4 and High Sierra 10.13.2, whereas my iMac is on 10.3.4 and Sierra 10.12.6. Each test run twice and below figures are the average. Each test pass used a freshly-created new library.

Test file: 6 min 9 sec UHD 4k H264 29.97 fps 100 mbps 4.4GB XAVC-S file from Sony A7RII. In all cases the file was imported with "leave files in place". The proxy files were deleted after the transcoding test. All other tests used the on-disk camera file not a transcoded or re-wrapped version.

Comments: Even though the top-spec 2017 iMac is a real speed demon on H264, the bare-bones "base" iMac Pro was generally faster except on the one H264 4k export test, and it was a little slower on proxy creation. The iMP was also faster at stabilization, even though I've seen several reports that it's slower.

It seems certain the iMac Pro is using some kind of hardware acceleration for H264 encode/decode. I don't know if it's a custom Xeon-W with Quick Sync or if FCPX is using AMD's UVD/VCE logic. It's very curious it was a slower than the iMac on 4k H264 export, whereas it was faster on 1080p.

Create proxy (all times min:sec)
iMac: 1:22
iMP: 1:48

Export to 1080p H264 Faster Encode:
iMac: 3:33
iMP: 2:26

Export to 1080p H264 Best Quality:
iMac: 6:59
iMP: 4:43

Export to 4k H264 Faster Encode:
iMac: 3:36
iMP: 4:30

Export to 4k ProRes 422:
iMac: 3:45
iMP: 2:28

Timeline render tests (via CTRL+R):

Aged film effect + default sharpen:
iMac: 5:31
iMP: 4:03

Stabilize 15 sec of 4k clip:
iMac, analysis phase: 0:45
iMP, analysis phases: 0:36
iMac, render phase: 0:05.6
iMP, render phase: 0:03.9
Last Edit: 09 Jan 2018 23:32 by joema.
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 11 Jan 2018 19:47 #93145

  • JarrodMFay
  • JarrodMFay's Avatar
  • NOW ONLINE
  • Senior Boarder
  • Posts: 62
  • Thank you received: 10
  • Karma: 0
Nice of you to post your tests for those looking to upgrade. I just upgraded from a 2010 Mac Pro to a 10 core iMac Pro and the speed increase is amazing. & the 5K screen (compared to my old 23" cinema display) is so nice during color correction. I love it!

---> Jarrod
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 12:40 #93172

  • FCPX.guru
  • FCPX.guru's Avatar
  • OFFLINE
  • Platinum Boarder
  • bbalser.com
  • Posts: 1614
  • Thank you received: 218
  • Karma: 18
There are also vast performance differences between an 8-core, Vega 56 iMac Pro and a 10-core, Vega 64 iMac Pro.
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 13:50 #93187

  • joema
  • joema's Avatar
  • OFFLINE
  • Platinum Boarder
  • Posts: 859
  • Thank you received: 185
  • Karma: 14
When comparing FCPX H264 performance of a 2017 top-spec iMac 27 to a base iMac Pro, it's important to remember the 2017 iMac is abnormally fast on that one workflow -- about 2x faster than a similarly configured 2015 iMac and also about 2x faster than a 12-core D700 Mac Pro -- I've tested all those.

I don't know why the 2017 iMac is so much faster than the 2015 on H264 in FCPX -- maybe the Kaby Lake Quick Sync was improved. However the points to remember are:

(1) Anyone coming from a Mac of any type (except the 2017 iMac) to even a base-level iMac Pro will see major performance gains on H264 workflows in FCPX.

(2) There are still gains vs the 2017 iMac just less evenly distributed. Although the 10-core Vega64 version is not gigantically faster than the 8-core on some synthetic benchmarks, what we care about is FCPX. There is some evidence the 10-core is considerably faster, although I haven't seen any highly controlled testing on FCPX of 8 core Vega56 vs 10-core Vega64 on H264.

(3) For ProRes workflows the difference will probably be even greater.

It's also possible Apple may wring further FCPX performance out of the iMac Pro in subsequent software updates. They are obviously doing something unique regarding H264 hardware acceleration else the Xeon-powered iMac Pro wouldn't be so fast on H264. The fact that 4k H264 encoding was oddly slower than the 2017 iMac (whereas 1080p was faster) implies there may be a driver bug or inefficiency in this initial version that might be improved in the future.
Last Edit: 12 Jan 2018 13:53 by joema.
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 14:37 #93194

  • FCPX.guru
  • FCPX.guru's Avatar
  • OFFLINE
  • Platinum Boarder
  • bbalser.com
  • Posts: 1614
  • Thank you received: 218
  • Karma: 18
I have the 10-core, Vega 64 iMac Pro, have run extensive tests on it, and have compared notes to someone I know with the 8-core. The 10-core, Vega 64 model does have some significant advantages. Even with H.264 the iMac Pro is significantly faster than a 2017 iMac that we compared tests with. I've not seen our iMac have any slow down with H.264 files at all.
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 16:56 #93199

  • dgwvideo
  • dgwvideo's Avatar
  • OFFLINE
  • Platinum Boarder
  • Posts: 1652
  • Thank you received: 178
  • Karma: 21
Ben - do you see an advantage to the 10 core over the slightly slower processor speed with the 14 core? The 14 core is not much more in cost.
Creating history....one edit at a time !
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 17:04 #93202

  • FCPX.guru
  • FCPX.guru's Avatar
  • OFFLINE
  • Platinum Boarder
  • bbalser.com
  • Posts: 1614
  • Thank you received: 218
  • Karma: 18
I've not tested a 14-core, but I've yet to see any app eat up all of my 10 cores totally yet. FCPX shows all cores slammed to the max when H.264 exporting, but only at the start, they settle down after that. I'm not convinced any app we in this industry use (aside from Maya, Modo, etc, which are a different plane of existence from us) up a full 14-cores .
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 17:11 #93204

  • VidGreg
  • VidGreg's Avatar
  • OFFLINE
  • Platinum Boarder
  • Posts: 868
  • Thank you received: 172
  • Karma: 25
For the curious
Barefeats has been running some comparisons between the Vega56 vs 64; 8 core vs 10.
barefeats.com/imacpro_vs_pt5.html
Hope this Helps, Greg
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 17:28 #93205

  • joema
  • joema's Avatar
  • OFFLINE
  • Platinum Boarder
  • Posts: 859
  • Thank you received: 185
  • Karma: 14
FCPX.guru wrote:
I have the 10-core, Vega 64 iMac Pro, have run extensive tests on it, and have compared notes to someone I know with the 8-core...Even with H.264 the iMac Pro is significantly faster than a 2017 iMac that we compared tests with. I've not seen our iMac have any slow down with H.264 files at all.

Ben, if you can point me to a 4k H264 test file you've tested, I'll download it and we can figure out the performance differences in H264 export between your 10-core Vega 64 iMP and my 2017 iMac. We could also compare this to the previous tests I did on the 8-core Vega 56 version.

Unfortunately H264 codecs have differing characteristics, so the test would be more meaningful using the same file. The file I used was from a Sony A7R2 but if you have a preferred file I can re-test that on both 8-core iMP and my 2017 iMac.

Or if we both have material from a common camera we could use that; it wouldn't have to be the same physical file. I have 4k 8-bit 4:2:0 29.97 H264 content from the Sony A6500, A7R2, A7R3, Panasonic DVX200, Panasonic GH4 and GH5, GoPro Hero5, DJI Phantom Pro 4, and DJI Inspire 2.

I'm mainly interested in performance for create proxy from 4k H264 and export from a non-rendered 4k H264 timeline to 1080p H264, both fast and best quality modes. But I can run anything you want.
The administrator has disabled public write access.

iMac Pro vs 2017 iMac FCPX performance tests 12 Jan 2018 20:35 #93214

  • FCPX.guru
  • FCPX.guru's Avatar
  • OFFLINE
  • Platinum Boarder
  • bbalser.com
  • Posts: 1614
  • Thank you received: 218
  • Karma: 18
I used sample files offered for download from RED. Drop me a line I'll give you the file names. ben b at b balser dot com. Combo of 4K, 5K, 8K in native R3D format. All tests and results are in the article.
www.fcp.co/final-cut-pro/articles/2029-new-imac-pro
Last Edit: 12 Jan 2018 20:35 by FCPX.guru.
The administrator has disabled public write access.