[Joe Marler]

[Scott Witthaus] “After 2 hours of lecture and demo…they were ALL editing at the end of the 3rd hour… Barely a question asked… much easier learning path than with FCP7…So it seems ease of use and speed don’t seem to be an issue unless one is trying to “unlearn” another software or forcing a work pattern on X that should not be done…

You’ve had good results and we’ve heard of many “new student” situations like that. However I think there’s a valid point that some things on FCPX are simply less intuitive than a track-oriented editor. E.g, doing a J or L cut “the FCPX way”, vs detaching the audio.

With a track-oriented editor the A and V tracks are just right there in front of you and there is no concealment or special deductive reasoning to figure it out. With FCPX I’m not sure I could have ever figured the “right way” of doing a J/L cut it out if I didn’t see a tutorial video. Apple page on doing split edits: https://support.apple.com/kb/PH12635?locale=en_US

The new way made sense once I learned to do it, but (to me) it wasn’t an intuitive, discoverable process.

One of the most extreme examples is stabilizing or applying optical smoothing to a multicam clip in FCPX vs Premiere. In Premiere it is intuitive and logical, whereas in FCPX it (ironically) requires an arcane sequence of steps.

[Joe Marler]

I am working on a documentary with 2.2 terabytes in one event, consisting of 5,600 clips and 117 hours of material. Content is all on a 16TB Thunderbay 4 in RAID-5. On my top-spec 2015 iMac 27 it works fairly well but FCPX could use more optimization on thumbnail generation. Unlike Lightroom, you can’t tell it to go ahead and pre-generate all the thumbnails. It tries to do this “just in time” and can get bogged down.

I found that having spotlight indexing *disabled* on the media drive makes it worse, and enabling spotlight makes it better. That is opposite to the normal advice.

If you are getting close to 10% free space that can definitely degrade performance — assuming that I/O is the issue. Hard drives start putting data on the outer cylinder where rotational speed (hence transfer performance) is the fastest, then fill up toward the hub. It’s true when you go below 10% HDD I/O performance can degrade drastically but long before that you are probably gradually losing performance. But your problem might not be I/O.

What part is slow? Skimming or going FF/Rev in the timeline or rendering the TL or applying effects or what? If you don’t have the viewer set to “Better Performance” you could try that.

If you are editing H264 that can be CPU intensive even on a higher-end Mac, especially if compute-intensive effects are applied such as stabilization. Even if it’s ProRes the effects must still be calculated, so gradually as you work on a project and apply more effects it can slow down.

If you are not using proxy you could try this on one project by just transcoding to proxy (not optimized) the used media indicated by view->show used media and then filtering (CMD-F) on used media, selecting only those, right-clicking and transcode to proxy, then set the viewer to proxy and examine performance.

Also if you have specific editing actions which are slow, examine whether those are consistently slow or only after FCPX has been running for a while. If you restart FCPX and they are fast it could be a memory management problem within FCPX.

[Joe Marler]

[Bill Davis] ” It reminds me how lucky I am to be able to work on smaller scale projects in a mostly laptop driven environment. The speeds my 2015 rMBp achieves with the stock Apple configuration have been truly amazing!…I suspect that when the industry re-conceptualizes the desktop systems in the same way they have with the “big sales driver” laptops – perhaps with ganged external connected GPUs and bigger pipes – it’s mind boggling to imagine how much performance will be within reach to so many creative workers. “

In the latest episode of Final Cut Pro X Radio, Mark Spencer said his main editing machine is a 2012 MacBook Pro: https://itunes.apple.com/us/podcast/final-cut-pro-radio/id1061415673?mt=2

That said, editing H264 4k will bog down many systems which seemed very fast on 1080p. I have the highest-spec 2015 iMac 27 available connected to multiple Thunderbolt arrays, and even FCPX can be sluggish on H264 4k, esp. multicam. It’s like the old days where you have to transcode everything before editing.

Unfortunately I/O bandwidth and GPUs offer limited help since many of the code paths are CPU bound. Since 4k is 4x the data, to obtain the same responsiveness as 1080p would generally require 4x the CPU power which is just not available in the i7 family. I/O and GPU can be scaled more and where they benefit that’s great.

An 8-core iMac with a new-generation GPU would help some, or an updated nMP with a 22-core E5-2699 v4 and equally updated GPUs. However that 22-core CPU is $4,000, so any workstation using it won’t be cheap.

[Joe Marler]

[Gabe Strong] “For what it’s worth, I get 25 seconds on the Bruce X test, but I have no idea if that is any
good compared to the new Macs or not. It sure seems to edit fast in FCP X to me, but I will admit I’m not
sure what ‘good’ is.”

My 2015 top-spec iMac 27 does about 17.1 sec on the BruceX test. I’m not sure how revealing that is — it’s a narrow GPU-intensive benchmark. My 2013 top-spec iMac 27 was twice as slow on BruceX but in actual FCPX editing it felt about the same. There may be a perf. penalty when doing bit block transfer operations on the ultra-high-res retina screen.

FCPX is pretty fast on a late-model iMac 27, but it still can’t smoothly edit multicam 4K H264 without transcoding to proxy. In some cases I transcode even for single cam. It is faster than Premiere on the same hardware but what’s the difference if you have to transcode on both Premiere and FCPX for smooth 4K editing? With Premiere that is now a built-in feature, and while it seems to transcode slower than FCPX, once it’s done the performance is lightning fast.

In general I’d recommend the nMP for production work, despite the aging design. It is a dilemma for people contemplating a purchase in the very near future. Hopefully within a few months we’ll have more info on possible iMac and nMP updates.

[Joe Marler]

[Oliver Peters] “Any concerns with the draw on the internal power supply of the tower? I presume you would need to put the boot drive into one of the optical slots if you populated the tower with 4 of these drives?

Also, have you run any AJA Sys Tests or BMD Tests on these for throughput speed?”

The spec sheet for the HGST 8TB drive says 7.4 watts operating power consumption. That is much less than a 2TB or 4TB WD Caviar Black, and only slightly more than a 4TB WD Red. You’d have to compare this to the spec of your current drives.

Re capacity limits, that would be determined by the version of OS X and possibly even the firmware. I think from an OS standpoint 32TB is doable, however there might be issues with the firmware and disk controller. You’d have to investigate that. Some info is here: https://www.everymac.com/systems/apple/mac_pro/faq/mac-pro-how-to-upgrade-hard-drives-what-type-supported.html

The new HDDs are 6 gbps SATA 3 so I don’t know the backward compatibility with the stock Mac Pro 5,1 disk controller.

I’m using two HGST 8TB drives in an two-drive OWC Mercury Elite Pro Dual as RAID-0: https://eshop.macsales.com/shop/Thunderbolt/External-Drive/OWC/Elite-Dual-RAID

That is Thunderbolt so of course it wouldn’t work with an older Mac Pro but from a pure HDD standpoint the drives are pretty good. In the above chassis I get 306 MB/sec read and 382 MB/sec write on Black Magic, and 390.8 MB/sec read and 329.9 MB/sec write on Quick Bench’s extended test, with “allow cache effects” disabled. I get substantially faster performance with 4x 4TB HGST drives in a Thunderbay 4 using SoftRAID for RAID-5, about 487 MB/sec write and 532 MB/sec read on Black Magic.

I’m not experienced with the specifics up upgrading older Mac Pros with newer HDDs, I just wanted to mention it as a possibility. It might not be worth the investigative time and effort. That would lead to either a nMP or a top-spec iMac 27, which both support Thunderbolt. For professional use I would tilt toward the nMP. I know lots of people (myself included) do professional 4k editing on a top-spec iMac 27 but I think the extreme pixel count of the retina screen slows things down a bit, even if using proxy. The iMac is very fast at importing and exporting H264 on FCPX due to Quick Sync, but that is only one part of a large workflow. If you have to transcode to proxy, or are using ProRes or REDCODE then much of that Quick Sync advantage goes out the window.

[Joe Marler]

[Oliver Peters] “One of the concerns I have in this upgrade is that a lot of the desirable local storage is shifting to Thunderbolt. So If I want to add a fast local RAID, it almost pushes me into a new Mac Pro as opposed to updating the tower any further. “

However I think you can update the tower to a max of 32TB using 8TB 3.5″ drives. I am using two of these HGST 8TB drives: https://amzn.com/B00NP6AOCK. They are very energy efficient and don’t output much heat for this class of HDD, yet performance is high. It also has a five-year warranty.

The competing drive in this category is the Seagate 8TB Enterprise: https://amzn.com/B016AG0ITU

If you later get a refreshed nMP you could redeploy these drives in a RAID box like those from OWC. I also use a 16TB Thunderbay 4, and it (along with SoftRAID) work very well: https://eshop.macsales.com/item/OWC/TB2IVT16.0S/. In theory I could put 4 x 8TB drives in it for 32TB total.

[Joe Marler]

[Oliver Peters] “update the primary machine from an older 12-core MP tower to a new machine. What is the consensus out there for extensive 4K editing? Either some configuration of the nMP or a decked out iMac. Thoughts? Experiences? Configurations?”

I have about 64 terabytes of Thunderbolt RAID storage on my 2015 top-spec iMac 27 and it mostly handles H264 4k in FCPX, although I always use proxy for multicam material. I also use Premiere CC 2015.3 for some things, which is somewhat more CPU-intensive and slower than FCPX when scrubbing a 4K H264 timeline.

Of course ProRes and REDCODE would be different but I’d still be concerned about using even a top-spec iMac 27 for your application since you have other options.

I would first suggest possibly upgrading your 12-core Mac Pro tower. Some GPU options are here: https://www.macvidcards.com/

My second suggestion would be get whatever nMP you can afford, despite it being late in the design cycle.

My third suggestion would be a top-spec 2015 iMac 27. Any of these choices can do the job, just some better than others.

If you use Premiere I would definitely not use the iMac. Premiere makes heavier demands on the CPU and GPU for a given amount of work, and having more CPU cores and a more powerful GPU would be real assets, which are not options with the iMac. That said, as of 2015.3 Premiere supports proxy and in my tests it really speeds things up, although it took about twice as long to generate as FCPX.

From an editing standpoint, the data path (hence screen output performance) should be independent of output screen resolution. You are reading and transforming data in memory based on the project resolution. IOW scrubbing 4K on a MacBook Air should not be faster than a retina iMac 27.

However ultimately the data must be written to the display, and for a 5k display that’s about 2x the pixels to move as a 4k display, and 4x the pixels of a 2013 and earlier iMac 27. The retina iMac 27 works pretty well editing on 4k video but IMO it needs higher performance display hardware, whether that’s the GPU, the Timing Controller (TCON), the busses or all of those.

Fast, fluid 4K editing is just really difficult on almost any hardware or editing software. We are unfortunately at the tail end of the nMP design cycle so people who need updated machines are in a difficult situation. Lots of people use the retina iMac 27 for 4k editing and it does work, but I would tend to recommend the nMP or an upgraded Mac Pro tower.

[Joe Marler]

I don’t have a quick answer but I notice you are trying to create a four-drive RAID-1 (mirror) set, where the drives are not the same size. I don’t have any experience with that, and I cannot find any AppleRAID documentation about how mirror sets would work with different-size size drives, or how more than two physical drives are configured in RAID-0. IOW are all four drives mirrors of each other? Most AppleRAID RAID-0 examples show using two same-size drives, and there is no further info given about how different sizes or more than two drives in the RAID-1 set would be configured.

I strongly recommend using SoftRAID or SoftRAID Lite for this, not AppleRAID. It avoids using the terminal commands and also is higher performance. In my limited testing on several two-drive Thunderbolt RAID-0 arrays, benchmarks were about 15% faster using SoftRAID vs AppleRAID: https://www.softraid.com/

The other problem with AppleRAID after El Capitan is there’s no GUI method to check the disk subsystem if you have an abrupt shutdown. You may only created a new RAID volume infrequently but uncontrolled shutdowns are more frequent, especially if you have a bunch of machines. Apple removed that from the GUI, so the only way is doing it from terminal. I also don’t think AppleRAID has any kind of “callback” notification of a drive failure. With RAID-1 you could have a failure (or several) and never know it until the last one fails. SoftRAID provides this.

Also SoftRAID has excellent customer support and is very fast to fix things. My only relationship is as a user who long used AppleRAID and got tired of issuing terminal commands due to El Capitan lobotomizing Disk Utility.

[Joe Marler]

As Brett said, try multicam, even though you have one audio and one video source.

Also make sure each group of clips from each device labeled in the FCPX metadata (either camera angle or camera name). Below is the procedure:

1. Import all clips
2. In Event Browser at top left select List View (it’s the button above the red X at left)
3. In Event Browser, click on clip name to sort by name. All the clips from each camera or device should be grouped together.
4. Select a group of clips from one camera/recorder (Click on 1st, go to last and Shift-click).
5. In Inspector at upper right, click on Info button. If Inspector pane is not on, turn it on with Window>Show Inspector
6. At bottom of Inspector Pane, click the drop-down box and select “General”. This shows more details.
7. In the Inspector Pane, enter a camera name. When you press Enter it will add that camera name to all the clips you have selected in a single step.
8. Repeat step 4 and 7 for each camera/recorder.
9. In Event Browser, select all clips from all the cameras you added camera name info to.
10. Right-click on the selected group and select “New Multicam Clip”, then set checkbox “use audio for synchronization”

[Joe Marler]

I also had the same problem. In my case Spotlight indexing was disabled on the media drive. When I re-enabled this and allowed time for the indexes to rebuild, FCPX browser thumbnail generation became much faster on large events.

Disabling Spotlight on media drives is a common recommendation and maybe a good idea in some cases. However for very large events with thousands of clips, it appears Spotlight somehow accelerates thumbnail generation. I think it’s ultimately a bug in FCPX since thumbnail generation would not just slow down, but would stop completely unless Spotlight was enabled.

There is still some initial lag on huge events (7000+ clips), and Activity Monitor indicates very high I/O rate (15,000 I/Os per sec on my 16TB RAID) but after they are generated it is very fast, even after selecting different thumbnail heights. If I select longer thumbnails (not just single frame) those must generate as I scroll down, but afterwards it remains fast — even after rebooting.

Unfortunately there is no way to tell FCPX to proactively generate them — you have to scroll down. The UI is somehow sending messages to the thumbnail generation thread. This is probably an optimization to avoid a lengthy thumbnail/preview generation period as with LightRoom, etc.

There may be multiple causes of this behavior. I first tried deleting the thumbnail files in the Library, which did not help. In my case it was unique to thumbnail mode in the Event Browser — list mode never had any problem, and thumbnails always appeared in the timeline.

I have 7,000+ clips (many 4k H264) and over 4TB in one event on a top-spec 2015 iMac 27 and 16TB Thunderbay 4, and it works very well. It is amazingly fast after the thumbnails are generated.