Joe Marler
22
Apple Final Cut Pro X Coins
39
COW Coins
Forum Replies Created
-
[John Perez] ” I would HIGHLY recommend going for the R9 M395X 4GB GPU and replacing the Fusion drive with a 512GB SSD (or 1TB if you can afford it)….The M395X 4GB GPU is actually a generation newer GPU than the M390 and will really help accelerate apps like FCPX that utilies the GPU for rendering and effects etc…”
I would definitely not get the 1TB Fusion Drive version, even if most of his media is external. I have both 2013 iMac 27 with 3TB Fusion Drive and 2015 iMac 27 with 1TB SSD, and for many video FCPX editing tasks there isn’t much difference, none of which is attributable to the I/O subsystem. The 2 and 3TB Fusion Drives have 128GB SSD which is much larger than the SSD on the 1TB drive. On a few things the 4Ghz CPU in the 2015 is faster and the M395X is mixed but modestly faster than the GTX-780m.
It doesn’t take much video to exhaust the internal iMac 27 storage, no matter what that is. Thus if the media is eventually destined to be on external storage, then getting a 512GB SSD is a good option since it does have some modest real-world speed advantage on the 2 and 3TB Fusion Drive in some workflows, and is probably more reliable. But the real-world I/O performance difference is not nearly as much as benchmarks would indicate.
Even the M395X is not lightning-fast so I’d definitely recommend this, since you don’t want anything slower. That said FCPX is very efficient and I can edit H264 1080p content on my 2013 MacBook Air without using proxy.
The big factor is H264 4k. It is very demanding on any hardware and any software. If the OP needs or ever plans on this, he needs the fastest possible machine. It is mostly a CPU limitation and secondarily GPU on effects, not I/O. However if proxy or optimized media is used to speed up the CPU/GPU part, then it can quickly become an I/O limitation since the transcoded media is much larger.
Here are some benchmarks I ran. The M395 tests were contributed by another user:
https://joema.smugmug.com/Computers/IMac27BenchmarksVsOthers/n-hR5wtF/i-3mL6T5r/A
-
[Eric Santiago] “…AVCHD files in FCPX…clips are average 10 min in length…Beach balling constantly on both nMac and 5K iMac….
I manage to (took way too long) to convert them all to Optimize/Proxy….I haven’t sat down and looked at it yet but what are most doing with their workflow when they receive this format?”Is this 1080p or 4k? I’ve edited lots of 1080p AVCHD content on a 2013 and 2015 top-spec iMac 27 with no problem, not even using proxy mode. These were from a variety of consumer camcorders such as the Canon HF-G30.
If it is 4k or for some reason the codec is more difficult, using proxy is the best approach. Generating optimized media may not be necessary. It consumes more time and space to generate — using only proxy is often sufficient. Of course you must remember to set the viewer to proxy, then set it back to optimized/original before the final export.
I sometimes use EditReady to re-wrap the AVCHD content before importing, but I don’t think it’s necessary except to make browsing easier at the Finder level. Since FCPX is so fast I usually just import everything — usually with “leave files in place”, vs browsing at the file level. It is usually fast enough to skim and make the initial selects, then if it’s multicam or more speed is required, I generate proxy media only for those selected clips.
For whatever reason if you still have problems even after trying proxy ONLY, I’d suggest EditReady — it is very fast and works well.
-
[Walter Soyka] “You’re taking that demo out of context. It was a tech demo at a developer’s conference, not a product demo at a user’s conference.
What has Adobe ever shown at NAB or IBC and not delivered on within a couple months?”
A specific commitment was made to deliver — not just demonstrate — the Metal API performance improvements in AE, Premiere, Illustrator and Photoshop. That has yet to arrive, after approx. a year of waiting.
I’m not doubting the Premiere improvements demonstrated at NAB will ever be delivered, and as a CC user I look forward to that. I was just curious how we know that will be delivered “in a few weeks”.
-
[Herb Sevush] “Future, as in the next upgrade due in a few weeks, not future as normally understood with Black Magic or Red delivery dates, or hoped for X features by users.”
As a CC user I would love to have those badly-needed Premiere performance improvements within a few weeks. How do we know it will be delivered in a few weeks?
A year ago Adobe demonstrated some huge performance performance improvements in After Effects using Apple’s Metal API, and committed to delivering these in AE, Premiere, Illustrator and Photoshop. To my knowledge everyone is still waiting on these hoped-for improvements.
https://www.youtube.com/watch?v=Zp1aPgLx4RQ
-
[Shane Ross] “…Premiere…or Resolve…They all have proxy modes.”
The current version of Premiere CC does not have any built-in proxy mode. In fact Adobe’s Premiere CC intro video still says: “allows editors to work with 4k and beyond, without time-consuming transcoding”, and “never needing to render until your work is complete”:
This worked fine on 1080p but on H264 4K, it’s no longer “mercury quick”.
Fortunately Adobe now sees the need for a built-in proxy mode and this will be added in a future version of Premiere.
-
[Gary Huff] “B.S., I don’t buy this for a second…it cannot outshine everything to the level of 8k native over everything else. I think this is highly exaggerated, or phrased in such a way that the specifics wouldn’t be as impressive as saying “8k native”
You are correct. While the FCPX playback engine is vastly faster than Premiere on 4k H264 content it is not nearly fast enough to edit native 8K H264 content smoothly — without other compensating factors.
The answer is probably in a several phrases from the article, and understanding what cameras and codecs were used.
Two VR cameras are shown:
(1) Jaunt VR prototype which used 14 GoPros. GoPros record 4k in highly compressed H264.
(2) 8K Nokia OZO, which uses eight 2k sensors and a JPG2000 codec. This not not a long GOP or interframe codec, so each frame is coded independently that makes decoding and playback much easier.
I once again went through testing and FCPX just sailed through stuff that other programs just chugged on….I was working with proxies in FCPX as you would with any large format project.”
If he means testing on a “several generations old iMac”, of course FCPX using is a lot faster than other editors on that same hardware if using its built-in proxy support.
However there’s no magic there. FCPX — despite its efficiency — cannot smoothly edit multiple streams of 4K H264 on a top-spec new iMac 27 without using proxy, and certainly cannot edit 8K H264 smoothly. It can edit those smoothly using proxy, OR if the codec is lightweight — as in JPG2000.
The article said even after they conformed to the original 8K content for color correction, it was still fast. That may be answered by the statement:
“Once it had rendered, it played in real time.”
Of course it is fast once it’s rendered in the timeline — Premiere is fast also when you render the timeline. The problem is you can’t keep the timeline rendered while editing. So this really says nothing.
FCPX is very fast, most people know that, but there’s no magic that enables it to edit H264 camera-native 8K on an iMac. It can do this if (a) The codec is easier or (b) If using proxy.
If you pre-render the timeline to then demonstrate how smoothly it scrubs, yes that works up until you start doing real work. The article may have helped to correct the misimpression that an iMac can’t edit VR using FCPX, but there’s no magic ability to handle camera-native 8K without other offsetting factors.
-
[Dondré Gumbs] “And for Plural Eyes, I watched a quick tutorial, and I was wondering what exactly is the difference between Plural Eyes and FCPX multicam automatic synchronization? I saw a greater range of options for Plural Eyes (ex. more options with audio and using it visually to confirm sync, etc.). Is it just stronger than FCPX’s sync?”
PluralEyes is vastly superior to the audio sync in either FCPX or Premiere CC. The FCPX audio sync works pretty well for smaller cases, up to 40 or so clips — provided the audio is good. It also requires all the clips from each camera or angle be named in the metadata, which is a straightforward manual step before syncing.
If FCPX sync fails you don’t get an error message, it just puts the clip in the wrong place. I have seen it fail on four-clip projects when sound levels were good but there was a lot of echo and background noise. The PluralEyes sync algorithm must be more robust as it consistently works better in difficult cases.
FCPX or Premiere CC cannot do sync drift compensation, which happens if one camera/recorder runs slower than another on a long take. PluralEyes automatically handles that.
The latest version of PluralEyes allows importing all clips from all cameras into a single bin, where it then automatically identifies and segregates each camera and recorder.
We once shot a large multi-week project with all the field audio recorders each set to the wrong date/time. This meant it was almost impossible to match those to the corresponding video. Fortunately Pluraleyes could search across many hundreds of clips in various subfolders and matched them all in a single pass.
PluralEyes is not perfect but it works very well. There is a reason why people continue using it even though editors have built-in audio sync.
-
[AJ Epstein] “There are many folks talking generally about the possibility of taking 4:2:0 4K footage out of some of the newer cameras – Panasonic GH4 and DVX200 specifically, and compressing it down into 4:4:4 1080. The only actual workflow I could find involved using Adobe CC and a GoPro codec.
Has anyone done this within FCPX or Compressor? If so, is there a well described guide?”
This is a good question and I don’t have a specific answer. It is supposedly possible to convert 4K 8-bit 4:2:0 to 1080 10-bit 4:4:4, trading spatial resolution for bit depth and color resolution. This was widely discussed when the GH4 came out:
https://www.eoshd.com/2014/02/discovery-4k-8bit-420-panasonic-gh4-converts-1080p-10bit-444/This article says it only works if you don’t crop or stabilize the 4K material (else you don’t have 4K before the conversion begins):
https://www.provideocoalition.com/can-4k-4-2-0-8-bit-become-1080p-4-4-4-10-bit-does-it-matter/There is a long complex thread discussing various workflows (mostly using ffmpeg) but no single agreed-upon solution: https://www.personal-view.com/talks/discussion/10637/4k-downscaling-progress-topic/p1
This thread discussed a custom OS X command-line tool (no longer available) which supposedly did the conversion: https://www.eoshd.com/comments/topic/5426-mac-app-to-resample-gh4-8-bit-420-to-10-bit-444/?page=1
There is commentary about whether just doing the conversion in Resolve might be good enough, but no clear cut tests and answers.
-
[Vesko Urukov] “I think I now understand they the footage looks low res during the editing process, but why does the exported file (exported as source) look low res?”
This is a logical question and many people before you have asked the same thing. Intuitively when you select Media>Proxy in the VIEWER, you interpret that to affect only the viewer not file export characteristics. However it does literally say “Media”, so you are actually selecting proxy media vs original for all operations — including export.
-
Oliver that was a good article. In the pell-mell rush toward 4K it’s good to evaluate the cost/benefit at every stage of acquisition, editing and delivery. There are cases (as you mentioned) where 4K can hurt quality not help it. Good quality 1080p can look fantastic. ABC, Fox and ESPN still broadcast exclusively in 720p/60.
I personally like the extra compositional freedom that 4K gives for HD delivery but it has associated costs and complications.
Re 4K-HD-4K there are various interpretations of that. You are mainly talking about post in HD then upscaling. Another workflow is post in HD via proxy, then final render at 4K or HD using the 4K content. Your point is it visually may not make much difference either way, which is interesting.
This should be no surprise since on broadcast TV, good quality 720p/60 looks better than poor quality 1080i/30, even for static images where the higher temporal resolution of 720p has no benefit.
That said, the upscaled quality of some of your comparison images was so striking I looked at them closely, and it may be the scaler or related processing altered the characteristics somewhat. Note the below crops, made from my 5K iMac screen when viewing the original images full size.
https://joema.smugmug.com/Photography/4K-Upscale-Comparison/n-TCbGvr/