I had time to do some export testing from Premiere Pro CC and Final Cut Pro X 10.1. Definite proof that second GPU is being used, and worth it!
I started with the same RED UltraHD timeline I had for the real time performance tests for Final Cut Pro X 10.1 and Premiere Pro CC, except that I copied everything and pasted it twice again to the end of the project/sequence.
As a test I exported that timeline to 4K ProRes 422 and to 1080P HD (scaled from the UltraHD for Apple Devices). ProRes is an all I-frame codec and should be faster to encode that a typical long GOP H.264 encode. The source timeline is 23.976, aka 24P.
In both Premiere Pro CC and Final Cut Pro X the export is handled in the background, so work can continue on the edit. I chose not to continue to work as that would have created a variability to the encode times as resources are split between playback and the encode. All encode tests were done with no other apps running and limited interactivity with the computer.
Premiere Pro CC
When you export from Premiere Pro the actual encode is handled by Adobe Media Encoder. The first export you do, there is a 31 second (on the test Mac Pro with SSD) to start up AME. There is no corresponding delay when exporting from Final Cut Pro X. These times do not include AME’s startup time, but in fair comparison with Final Cut Pro X, it should be added.
Export to UltraHD ProRes 422: Â Â 7’34”
Export to H.264: Â Â 4’11”
Clearly there is some optimized code for the export to H.264 for Apple Devices. I also attribute the relatively slow encode to ProRes 422 to code that Adobe has little opportunity to optimize (or even move to the GPU).
What got interesting is when I disabled the second GPU for Premiere Pro and AME. There was little change to the export to UltraHD ProRes 422, confirming that the code is obviously not optimized for Adobe’s Mercury Engine (OpenCL) being used on the Mac Pro.
The export to H.264 for Apple devices, with only one GPU in use, took 13’38”, confirming without doubt that Adobe are a) doing a great implementation of OpenCL and b) definitely using the second GPU when they have the chance to optimize the code themselves.
For comparison, I did a similar export to 1080P MXF Opt1A DNX 170 1080P, which took 2’20” with one or two GPUs enabled. (Again likely code Adobe can’t optimize themselves.)
Final Cut Pro X
The Final Cut Pro X Project timeline was a few seconds longer than the equivalent in Premiere Pro. I started with an unrendered Timeline and background rendering off for the most direct comparison with Premiere Pro I could.
Final Cut Pro X to UltraHD ProRes 422 from Ultra HD timeline – 3’35”
Final Cut Pro X to HD H.264 for Apple Devices – 3’43”
The H.264 encode is only slightly faster than the same encode from Adobe and I would suggest that also confirms that Final Cut Pro X is taking advantage of the second GPU on that encode. What is interesting is that the ProRes 422 result also suggests some additional optimization has been done on the ProRes encoding in the Apple universe. That’s an advantage when you control your own codec I guess.
I next turned on background rendering and let it render the timeline while I did something else. This rendering would normally happen during idle time ready for export. In my experience I’ve never had to render for playback on the two relatively modern laptops I’ve used Final Cut Pro X on, but apparently there is some use for smoothing playback on less powerful systems.
Total render time for the timeline 2’11”. Doing the two export tests again from the rendered timeline:
Final Cut Pro X to UltraHD master file from Ultra HD timeline – 54″. So with background render and this export the total time was 3’05, which – surprisingly – is less than the direct export with rendering at the time of export.
Final Cut Pro X to HD H.264 for Apple Devices – 1’54” or with background rendering included 4’05”, but of course that’s not the way you’d do it. Background render happens during idle time, so that effectively a master file is ready for export when you go to export, something that differs from the way Adobe does it.
Bottom line: background rendering is your friend when it comes time to export from Final Cut Pro X, and it doesn’t matter which app you use, you’ll benefit from the second GPU where you have one (including non-Mac Pro dual GPU configurations).
30 replies on “Export times: Mac Pro, Final Cut Pro X and Premiere Pro CC.”
Thanks for finally testing some exports. I think in your tests the QuickTime codec is a large variable in the testing methodology. I would prefer to eliminate that variable to test the dual GPU and would suggest this test:
1:00:00 Red Epic 5k file @ 23.976 straight export
Export to H.264 in a .mp4 file type and not an .mov wrapper
Test single and then dual gpu
the debayering and lack of QuickTime variable should make a very interesting result
Neither app exports to Red native.
The H.264 tests were all mp4 not mov.
Tested single and dual GPU with Premiere.
All done.
How long is the sequence you are exporting?
A minute 25 seconds in Premiere, 1’33 in FCP X.
I’d be curious what the export times would be from a MBP or iMac from the 2 apps as well. There’s been a lot of people suggesting that the new Mac Pro isn’t worth it compared to a top-of-the-line iMac. I’m fairly certain due to your test of using only one GPU on the Premiere export that the times would be much better, making it worth the investment for someone exporting video many times in a day.
Any dual GPU mac will be worth it over the single GPU.
I just received my new Mac Pro with 8 cores and dual D700s, 32GB of RAM and a 512GB Drive. I am editing all footage off of a Pegasus R6 Thunderbolt RAID. The drive speed isn’t a factor either way.
So…. some of my VERY preliminary testing with new Mac Pro shows that Premiere CC (latest version) using the plugin I use most (film convert -FC) needs much more optimization for the dual GPUs – and CPU utilization for that matter.
I know FC only uses one GPU presently from the developer. Overall, using two typical projects as an example, I’m only seeing 25-50% speed up in export speed over my late 2012 27″ maxed out iMac. (Sending to the AME queue is slower by about 10%, but the delta is the same over the iMac regardless). That’s significant of course but not the at over 100% I expected to be seeing at the least given the MacPro config of 8/D700. I’ve noticed as well that Premiere Pro CC barely hits the CPU and GPUs. I have yet, in my very limited testing, see it max out like I did on the iMac no matter what I throw at it, and I’m talking CPU only.
Of course that’s just testing two similar projects; and it depends on what one is doing. Some stuff is much, much faster like Red Giant’s Denoiser II or Warp Stabilizer. Those are approximately 3-4x faster rendering out for me. I used to avoid them for speed reasons a lot of the time but now they are fast enough to rely on quickly. Other stuff like DxO prime noise removal on RAW stills is much faster too, as is Photoshop CC: some effects like blur, sharpening, resize there are nearly instant now even on giga pixel files.
I expect Premiere Pro CC to improve over time (there’s still much more optimization to be had.) And of course FCPX is much faster on the new machine. but I very much dislike its whole editing paradigm. For me: the timeline is just horrid on it; simple things like replacing a word in someone’s dialogue can often turn into a multi click, multistep process that is nearly instant in Premiere and most every other NLE. Just to trying to see your whole timeline on a laptop is a chore, to see what your edits and sound modifications are in detail are problematic, trying to keep things in sync can become unwieldy due to they magnetic timeline and so on. I’m dumfounded that you can’t even zoom your timeline window to full screen, or move panels around for more efficient workflows! I tried, but the speed increase on export is more than cancelled out by the speed reduction while working with it, at least for me. Maybe I am just old and can’t relearn the paradigm, but I’m not the only working editor to think this.
I’m sure Adobe (and Apple) will improve over time. They have to to stay competitive. In the meantime I’ll take my 45%. Your mileage may vary.
Thanks for the report on Premiere Pro’s performance. Wish you hadn’t bothered to do a gratuitous slur against FCP X, which BTW, is used on way more high end file and TV than Premiere Pro is right now. See Richard Taylor’s master list. It’s fine that you don’t particularly like it, but just get over it.
I didn’t think it was possible to “slur” software. That sounds personal. I just want something that works, whether it’s made by Adobe, Apple, or some guy down the street in his basement. I really dislike people taking personal feelings on comments about software. Please trust me Philip when I say that there was nothing personal about my comments at all, they’re just regarding actual workflow on actual projects, and time lost with workarounds on tools that don’t work as good as they could. That includes Premiere, which, for all the good things I like about it, has to be one of the most crashy pieces of software I have used in a long time.
All I was saying is that FCPX doesn’t follow editing paradigms I’ve been used to for many many years. I don’t think one should have to hunt for multiple steps to do something as simple as, say, adding a audio transition. It’s two clicks in most every other software I’ve used, but when I looked up how to do it in Final Cut Pro X on Larry’s site, it was multiple steps and dragging and moving little buttons and drawing curves. I just don’t have time for that. Maybe I’m missing something.
Anyway, I did some more testing by loading up similar media files in FCP X and just scrubbing through using the film convert plug-in. It was actually slower, in best quality, then Premiere was under full. I was shocked. The export however was reversed, it was moderately faster under FCP X. Clearly everybody has more tweaking to do today software to take advantage of the new machine.
The discussion of FCP X’s “suitability” was out of context and in my opinion (and it is my blog remember) inappropriate.
That FCP X doesn’t follow old, outdated paradigms is a good thing. A very, very good thing, as we move into a future where those paradigms (film and tape) no longer exist. New users have no idea why things are set up the way they are in legacy style apps.
Audio transitions are really easy if you aren’t locked into old paradigms.
Look Premiere Pro is a great app. Adobe have done a good job, and it’s a good choice for those who don’t particularly care to adapt to the future.
Feel free to not read or comment in the future.
And including a third party non-optimized filter into the mix was a really stupid way to do any testing. Really, really stupid to then think that had any relevance to the performance of the host application (in either case).
Sheesh.
Philip: I’m really sorry you’re so upset, I’m just trying to have a discussion on editing paradigms and software. I apologize for whatever I did to upset you, no harm was meant and none of my concerns were remotely directed at you. Perhaps it was not in line with the post about FCPX and Premiere export speeds, and for that I apologize. I have thanked you for your tests… twice, and on many other occasions for your help (and recommended your tools and blogs to others), so I really don’t understand why you are so seemingly mad at me. There’s just no reason for anger in what I see as a productive discussion about tools.
I will add this: I use that plugin all the time. So it’s not “really, really stupid” to test with what I use constantly. You of all people understand the interplay of plugins and host software performance, so your comment just makes no sense to me.
I will respectfully disagree on audio crossfades, which was near the top of the list of wished for changes by FCPX users that you sent me to.
“there was nothing personal about my comments at all”
Huh?? How was that not PURELY personal?? “I dislike it”… can’t get more personal than that. Hilarious. 😀
And I completely agree with Philip. An “addendum” like that was and is uncalled for and superfluous in this context, and nothing but yet *another* laborious “I hate it because it won’t think like ME i.e. I don’t get it!” statement which is of no use to anyone, but merely perpetuates the tired prejudices that have been floating around for nearly 3 years now. And yes, the ignorance on the matter becomes infuriating after that period of time. It is in fact up to YOU to think like IT to actually use it efficiently, sorry. Otherwise why even bother? If you don’t care to ridden yourself of stale old habits and muscle-memory for the sake of a faster and far more efficient workflow, fine. Your prerogative. But then at least do us – the people that use X daily on a professional basis very very successfully and productively – the favor and don’t be so painfully solipsistic and blame *the software* for your lack of expertise and understanding of how things work and/or meant to work.
BTW the statement “trying to keep things in sync can become unwieldy due to they magnetic timeline” by itself makes it forehead-slappingly obvious to me (and I’m sure any and everyone that has in fact *understood* the use and concept behind the MT) that you in fact don’t get it and are missing some seriously basic understanding of X. That’s too bad. Since if anything, then that is EXACTLY what the MT does *exponentially* better than ANY other NLE paradigm out there and what it was even conceived for in the first place: to maintain sync!! Wow.
Oh, and dialog replacement a “multi click, multistep process”??! Dude… learn the software BEFORE you make a claim or venture an “expert” opinion, which is what you apparently were brazen enough to think that was. Seriously.
Put an opinion like that out there, expect it to be challenged. Make it a tired and three year old opinion based on nothing other than misconception and/or misunderstanding of the most basic concepts, expect that challenge to be an “impatient” one.
Phillip – Thanks for the testing!! –
Which Mac Pro model did you do the testing on?
How many processors and Ram and hard drives. . .?
Paul if you’re having trouble with the timeline in FCP X and can’t learn the new paradigm just drag a black clip down to the base track and make it the length of the project before you begin editing!
The timeline will then behave just like Premiere and FCP 7 if there’s a big long clip on the base track . . .
12 core, dual D700, 32 GB RAM, all off the internal Flash storage.
DanO that’s a great tip thanks.
For me that’s the least useful, nonsensical suggestion possible. Also unfortunately one that is given happily and far too often. One for the “brain-lazy”.
Don’t get it or plain don’t want to? Why bother? Stick with what you have. I see no value in spreading and perpetuating bad habits and counter-productive workflows as band-aids for the NLE/paradigm bigoted. Just sayin’.
I did think the discussion of FCP X was out of context in the first reply. Up until then, I thought it was an excellent post. And I may have over-reacted a bit, so I too apologize.
Moving on, I understand what you’re saying about testing with a third party plug-in you use all the time, however it is very unfair to assign the results to the host app when it is the plug-in’s performance that you’re testing, not so much the host app.
Doing a cross dissolve for audio with a transition is more difficult in FCP X than in FCP 7, however I think it’s a really bad way to do an audio transition and I never use it. I prefer to choose how and where my audio cross fades and that’s where the fade handles in FCP X work very well for me.
I do get very tired of people who’s personal preference is to use a more traditionally styled NLE and then somehow argue that anything different is “worse”. For you that works, for me I’ve been dying for an alternative way before FCP X was released. From Sep 2009 (nearly a year before FCP X was released) http://www.philiphodgetts.com/?p=2298
My NLE history goes back to 1994 with Media 100, then I was a very, very early (as in beta tester) adopter of FCP 1 in 1999. I co-authored a book on Premiere 5 that was never finished because I preferred FCP, and my coauthor preferred Premiere 4!
While I have a lot of admiration for the work done on Premiere Pro CC (and have many friends at Adobe) I do strongly believe that it was way past time for a paradigm shift, (as the article above suggests) and I guess only Apple could do it.
I also think Media Composer is an excellent choice if you’re in a studio situation for movies or television, but I think its influence is waning.
For me, it’s nice to have a choice that’s NOT traditionally based. I have not met anyone – in person or online – who finds they work much faster in FCP X *once they make the adjustment*. And that’s a barrier for many.
Today may not have been my finest day, and thanks for sticking with it Paul. Cheers.
Philip, thank you for the apology, I appreciate it and I do understand when you’re coming from. I have to admit that your passion also prompted me to take a second look at FCP X. I figure, I have this brand-new computer with all this power, and the only application for editing the currently takes advantage of all that power is FCP X. So there is an advantage to me if I can do everything I do and CC in FCPX. That’s the prompt point The way, and the subject of your original blog close, getting more export speed. So I was led to this discussion by a technical concern, and learned what is and what is not fast with the new Mac Pro that I have. Then I started thinking about everything else.
I’m watching a presentation from Larry Jordan about whether FCP X is ready for professionals. He starts off with this big long speech about it doesn’t matter the software so much is it matters if the tools we use help us do a better job. He said a bunch of other things, but my point was I understand where he’s coming from. If people are willing to open their minds to other paradigms they might find that they can serve their clients better. But I think that’s what you’re saying here, you strongly believe this.
When FCP X was first released, everyone hated it including myself, but I did write a blog post, that is now lost, that said maybe it’s about thinking about things in a different way to enable us to be better in the future. Things like keywords and meta-data instead of bins for example. These things could present a lot of speed up and workflow if we can just forget the old paradigms and start a new.
Larry also talks about fear. We all fear that a brand-new tool we don’t quite understand will not enable us to get the work that we need to get done efficiently. I have projects that are very important, they directly help professionals in the security profession save lives. I literally cannot afford to do the best job possible, because in some of the work I do if I screwup and I’m not on time, and I can’t tell the story most effectively, it’s a big deal.
So that’s where my fear comes from. Now I’m about to start another major project, and I’m at a point where I could seriously look at learning FCP X again. When I went over all the improvements since launch in a list, I was actually quite shocked at all the stuff that they added it back in or improved. It’s actually quite stunning how much improvement there was Ben, with no additional fee for the next versions.
I went and considered my typical workflow for say, a documentary, on the surface anyway, I don’t see anything that could stop me from creating the same thing in FCP X. While gaming the speed, of the new Mac Pro, and perhaps all kinds of other improvements from things like metadata that you champion. I can also you can use tools such as your own to get data in and out of the program from other sources such as aftereffects, audition, etc.
I can’t say I’m going to be successful, or that I won’t one into some roadblocks, but I will try. The inability to drop on auto transitions for example bothers me, but I see that someone wrote a plugin that lets you add a audio crossfade with what looks like a basically blank video clip in one step.
I’ve got some training, I’m going to sit down and look at everything again, and see if you write. Perhaps it’s time to think a little differently.
PS please excuse some of the weird mistakes, I dictated all that from my iPhone while running around
I think it did a pretty good job of the dictation.
So… a couple weeks or so later and I am astounded- dumbfounded really at how good FCPX 10.1 is compared to 10.0. It’s like every single one of my concerns was addressed.
I have been relearning it and after a week of intensive learning and studying, I am getting my head around the paradigm. And my test projects are cutting together much more fluidly. I think I am staring to “get it”.
I made a checklist of issues from 10.0 and -every single one- was overcome since if you include 10.1. (Media management especially). There’s an answer for virtually everything now that was missing. (ok audio crossfade filters are still a workaround but there’s a filter for it from Alex4D that works in most cases and like you said, doing it manually isn’t horrible.)
It’s very fast too. Very very fast. Oh my it’s fast.
Export times are faster as well, except using AME I see H.264 can be faster. But the ProRes is much faster coming out of FCPX (on the MP it’s almost instant for under 5 min pieces).
Philip, I can understand where you were coming from with your frustration now. It was that passion that ultimately made me look twice and I am happy I did.
Premiere remains a fine tool and I will use it for some things (and some jobs demand it). It seems more fluid editing un-transcoded with RedRAW too, if only that I can preview at “best performance” at a seemingly higher res. (FCPX is still fluid enough there and ProRes 4K flies).
That said, I don’t so far see any issue to moving to FCPX for my day to day stuff. I’ll let you know after my next big project; I’m going to jump in and edit a whole 20 min show with it and see. The Multicam alone was a revelation, and the key wording is instinctive.
I guess I have to say to people like me who wrote FCPX off on launch, should look seriously at 10.1 again. It’s a huge improvement, and there’s a rich ecosystem now out there for support.
PS latest version of Film Convert equalized the performance between both platforms… and they don’t even use FxPlug 3 yet.
I think it is *extremely* telling that your fist and oh-so-critical post about X was when you in fact didn’t even ACTUALLY KNOW X in its most current incarnation!
Maybe THAT will explain the level of “anger” that you received in response (I’m only now scrolling this far, sorry). Since I think it’s fair to say you (factually and by your own admission) *didn’t know what you were talking about*, and it showed. And that, like I said, just plain gets really REALLY old and tiring, which can lead to defensiveness and tempers flaring at times.
Thanks for running and sharing the results from your test. I found it fascinating since I’ve been working with Adobe CC on a Dell T5600 running Win 7 for a few years, and have been very curious about this. It points to what I’ve long suspected, that Adobe CC for Mac is a completely different product than Adobe CC for Windows. I have one friend, called “Mad Dog”, because of his cross-platform experiments with both music and computer tech, who, as a test, partitioned his Mac to give all the resources to Windows, with just the bare minimum to the Mac OS, then tested the Win 7 version of Adobe CC. He claims it not only outperformed the Mac version, but also outperforms the Dell. His tests showed that the best performance was Adobe CC running on Win 7 on a Mac. I found that very interesting and thought I’d share it. I’d love to see a lab recreate that test. Again, thanks, Philip. These comparisons are very helpful.
That is curious. Thanks for sharing.
Since you don’t actually specify it, Philip, did you export the H.264 with the “Faster Encode” or “Better Quality” setting??
Because, in case you didn’t know, only the “Better Quality” setting truly benefits from the GPU rendering. Meaning that if you export with “Faster Encode”, your export times won’t actually be noticeably different on a nMP than from say a maxed out iMac. With “Better Quality” on the other hand it will be *exponentially* faster, believe it or not. 😉
So really, a comparison with the “Better Quality” setting would have been a true measure and more than likely have shown Premiere to be that much (a lot?) slower even. Part of the thinking behind it being, that “true pros” would only use the better quality, therefore they benefit the most.
(sorry for being so late to the game here. 😀 )
I cannot remember which setting I used. In fact I wasn’t aware of those settings in the Share panel at all. Better Performance/Better Quality timeline playback quality doesn’t affect export which is always at Better Quality.
Philip
Simply output for Apple Devices, check the Settings tab, and you’ll see “Faster Encode” (single-pass) and “Better Quality” (two-pass VBR). By default both Apple Devices presets are set to “Faster Encode”, so if you didn’t change anything manually, that’s what it will have been. 😉
That’s of course assuming that that’s what you used… 🙂