Why do users say FCP X is fast? Some new numbers.

[Tom Sefton]

The most recent version has got miles faster exporting R3D to ProRes and h264, but as yet it still can’t compete with RedCineX Pro from our new Mac Pros. It’s still pretty sluggish when editing Dragon footage at 6K, but overall the improvements have been fantastic. Hopefully future versions will get even more smooth!

Co-owner at Pollen Studio
http://www.pollenstudio.co.uk

[Walter Soyka]

[Bill Davis] “A LOT of the new speed is clearly due to X being optimized for the new R9 GPU in the new Retina MacBook Pro, but I honestly didn’t expect THIS much of a boost. It bodes well for what all computer manufacturers can accomplish when they truly optimize tasks like intensive video editing processes to the latest advances in hardware as those appear.”

See also this thread, started by the original tester:
https://forums.creativecow.net/thread/335/82723

I think it’s important to note that as designed, the primary factor in this test is actually H.264 compression, not the rendering pipeline. I’d bet Apple is using Intel Quick Sync [link] to accelerate H.264 compression, and I’d bet Adobe is not.

Quick Sync is a hardware feature of recent Intel Core consumer-class processors, but not on of Xeon workstation/server processors. Look at the original data:
https://docs.google.com/spreadsheets/d/1j8fTm7y6AbkDipHhFBKknTqQ45q7RxKmV4wq59_xa_s/edit#gid=0

Note in the test data how a MBP running FCPX shames a nMP running FCPX by a factor of almost 6x! On the nMP, where Quick Sync is not a variable and the test is a slightly purer pipeline comparison, Pr actually outperforms FCPX by about 20%!

I’m not saying that knowing the speed of H.264 compression isn’t important or that Apple’s use of Quick Sync is not relevant; it is. But I do think this test as currently designed does not support your hypothesis that the results are driven by the new GPU.

I would, however, agree with your conclusion (assuming you’re talking about software developers moreso than computer manufacturers, as there is only one company in our industry where that Venn diagram would overlap).

Walter Soyka
Designer & Mad Scientist at Keen Live [link]
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
@keenlive   |   RenderBreak [blog]   |   Profile [LinkedIn]

[Steve Connor]

So is the timeline and edit performance significantly increased? I honestly don’t care in the slightest about output renders as that’s such a small percentage of my project time I don’t mind a slightly longer wait. What I do care about is the timeline and editing in FCPX having less lag.

[Bill Davis]

Walter,

As always I appreciate your analysis and know that in a world where someone is trying to discern what system is optimized to perform all manner of high-stress rendering and mission critical workflows, the factors you mention are clearly relevant.

I also suspect that there are readers here who simple want to EDIT real world stuff, plenty of which is generated by legions of new generation MXF, AVCHD, ProRes and yes, h-264 spewing cameras – and for them, this kind of “real world” test is pretty relevant. The fact that an easily purchased laptop running equally accessible software can “play with the big boys”in the I NEED TO GET MY WORK DONE FAST! game is news enough.

Not perhaps at the total top of the game, I admit. If you’re shooting as the above poster is, 6k Red Dragon files — then what you can even do that on a commercial laptop is surely interesting, but, i suspect, not exactly central to their decision making processes.

For the editor who just needs to get their day to day work done quickly and efficiently (like me!) it’s STILL a revelation that my new laptop provided such a HUGE boost in productivity. Way beyond what I expected when I bought it to replace my aging old machine.

As always, your detailed analysis of the underlying technology is welcome and personally interesting. As always, I’m really glad you’re here.

Know someone who teaches video editing in elementary school, high school or college? Tell them to check out http://www.StartEditingNow.com – video editing curriculum complete with licensed practice content.

[Bret Williams]

Ditto. If you have a large library (job) with lots of projects (sequences) and versions and each has compounds, titles, color correction, effects, templates, and all the other stuff that encompasses a usual edit, then the app really is a dog. Lag, beach balls, you name it. I tend to restart the app every half an hour to clear out whatever the gunk is that’s gumming it up.

But yeah, 3 months later it saves me 5 minutes on my export. 🙂

[Bill Davis]

I used to see beachballs and the need to re-start occasionally too.
Haven’t once since the new hardware. But maybe it’s an “order of complexity” thing. Be interesting to see whether whatever new MacPros are in the future make a commensurate difference to what I’m seeing in older laptop tech vs the latest laptop tech.

It’s also likely the super fast SSD is making a perception difference for me. Seems like waiting for anything is now a thing of the past. Note I currently keep smaller projects on the SSD and do larger projects via USB 3 externals because that’s what I have currently. With dual Thunderbolt 2 ports on the MBPro, I’m sure I’ll be seeing a throughput boost there as I migrate my storage devices to the new TB buss.

It’s been known for a long time that X does seem to LOVE the latest and fastest hardware you can afford.

Now if they just keep the prices on external SSD’s falling I’ll be happy.

Know someone who teaches video editing in elementary school, high school or college? Tell them to check out http://www.StartEditingNow.com – video editing curriculum complete with licensed practice content.

[Oliver Peters]

I decided to run a similar test on my laptop, but with a slightly different sequence set-up. My machine was bought early this year, so it’s essentially the new hardware, except with the Nvidia card.

15″ Retina MBP 16GB RAM 1TB SSD (MBP 11,3)
Iris + GT750M GPU
2.8 GHz i7

BMD Speed Test with a 5GB file wack is around 940 write / 860 read.

Mac 10.10.4
FCP X 10.2.1
PProCC 2015

All media and exports to the same internal SDD. My source clips were 1080p/23.98 ProResHQ in a 1080p/23.98 timeline. TRT 5 min. 5 sec. All clips were scaled up by 20% and had native color correction. Several titles with shadows and several 5-layer PIPs with scale and rotation on layer 2-5. Base layer clip in the PIP had a Gaussian blur. This was duplicated in both apps.

FCPX export to 1080p ProRes 1:48
PProCC with Open CL export to ProRes 5:41
PProCC with Open CL export to MXF 6:20
PProCC with Open CL export to ProRes via AME 6:05
PProCC with CUDA export to ProRes 3:45
PProCC with CUDA export to MXF 4:40
PProCC with CUDA export to ProRes via AME 5:55

The MXF and AME exports were done to test if these might be faster in an Adobe workflow since you aren’t going through QuickTime and because AME is usually a faster renderer. The later doesn’t appear to be true with a single file. Clearly FCPX is a faster render/exporter and that has nothing to do with H.264 optimization in this case.

As far as performance with X, I bought this machine for a large on-site corporate conference gig. I was cutting down 1-2 hour multicam recordings of presentations. 4 streams of ProRes from a single USB 3.0 drive. The operation was very fluid. I will attest to the fact that comparison to any older machine, like my 2009 MacPro tower is night and day, even though the tower now has an SSD. Large FCP X projects or long timelines and long source clips simply bog down the machine in the worst way when you are trying to rock & roll in the edit. You absolutely can’t get good performance if you turn on waveforms or filmstrips in the timeline. The MBP shines in that regard.

Premiere Pro also works quite well, but CC2015 is a bit touchier in things like screen refreshes and seems to now suffer from the same quit/restart issues that have plagued FCP X. I work on two machines with Nvidia cards (counting this laptop and a client’s) and I must say that overall the Mac works better with the AMDs. I have a Sapphire 7950 in the tower. The Nvidias just seem glitchier.

– Oliver

Oliver Peters Post Production Services, LLC
Orlando, FL
http://www.oliverpeters.com

[Mathieu Ghekiere]

Hi Oliver,

when I saw the original test-video that Bill linked too, I was glad to see it but I also was in doubt because he didn’t share full details on codecs and settings used. (Even though I’m an FCPX user switched from FCP7, I want my results fair and transparant and clear, because I think Adobe is trying to do a lot of good stuff with Premiere too, even though I don’t see myself going back to tracks anymore, but who knows).
So it’s nice to see your results too, which pretty much confirm that FCPX is a very fast renderer (which I also notice on my pretty new machines – a 2014 Macbook Air and a base model 5K Retina iMac), but with full settings and codecs revealed to have a more informed view.
Thanks and regards,

[Neil Goodman]

[Steve Connor] “So is the timeline and edit performance significantly increased? I honestly don’t care in the slightest about output renders as that’s such a small percentage of my project time I don’t mind a slightly longer wait. What I do care about is the timeline and editing in FCPX having less lag.”

this.

[Charlie Austin]

[Neil Goodman] “this.” … is not an issue if you have a newer machine. (see Olivers test above) Not surprising, as Apple is in the business of selling computers. 🙂

————————————————————-

~ My FCPX Babbling blog ~
~”It is a poor craftsman who blames his tools.”~
~”The function you just attempted is not yet implemented”~