Dropped Frames on new M2 MacBook Pro Max - not using processor power

Finally I got a new MacBook Pro Max M2 with 32 GB RAM and I am trying to edit Sony XAVC-HS UHD 50p to a 50i HD timeline. I copied the footage to the internal SSD an just applied a little color correction and scrolled around the timeline and hit playback and fast forward and I am getting a "dropped frames" warning. When I check in activity monitor during playback, cpu is used by 10 percent at most. The fan of the MacBook isn't kicking in, it doesn't even get warm. I was expecting that FCPX is able to use all of apple system resources to the max. Did anyone have the same experience with HEVC-footage?

The M2 Max is a great machine. I have an M1 Max MBP and an M1 Ultra Mac Studio. In general they handle Sony 4k/50 XAVC-HS well. 

What you are seeing is likely a misleading warning caused by the FCP setting "If a frame drops, stop playback and warn". That is more intended to warn of dropped frames during normal 1x speed playback. 

Each press of the L key during playback doubles the frame decoding rate required. So if it's playing at normal speed, that is decoding 50 frames per second. If you then press the L key three times, that is 8x speed, or 400 4k frames per second. 

For that reason on FCP and other NLEs I usually disable any dropped frame indicators. 

Re Activity Monitor, during playback of XAVC-HS it is probably waiting on the hardware decoding accelerator. In such cases it is not CPU-bound, not GPU-bound, not I/O-bound but the accelerator is doing most of the work. Unfortunately on MacOS there are no instrumented parameters to monitor the hardware video accelerators.

A related item is each NLE places different priority on updating the playhead position vs updating the viewer. Premiere Pro places a high priority on updating playhead position, even if the program monitor playback is jerky and laggy. By contrast FCP places a high priority on smooth viewer playback, even if the timeline playhead becomes laggy.

FCP uses a different software code path for the skimmer. You can often skim material and get more fluid response than using the playhead.

Here is MacBreak Studio #331 on how to edit faster using the clip skimmer:

Thanks a lot Joema! There are still so many things I didn't know about FCPx.

I think I will risk it and record my documentary in HEVC. It's a 30 minute feature for broadcast and the FX3 and my MacBook is brand new so I hope it will not slow down excessively close to the finish line. But anyway I will work from an external SSD and create proxys on the internal SSD to work on the go and before finishing and color correcting I think I will optimize the media of my entire timeline to my internal drive anyway. Then there should be no issues at all.

The FX3 has two different HEVC codecs, I think 10-bit 4:2:0 and 10-bit 4:2:2. The 4:2:2 one might be slightly harder to decode, but normally on the M1 and M2 machines it's no problem.

If you are using timecode, FCP cannot read that from the default FX3 MP4 container format. A solution is process that before importing with the utility QTChange: www.videotoolshed.com/handcrafted-timecode-tools/qtchange/

Make sure all your filenames are globally unique, IOW unique across all cameras, the project, the library and across all projects in the future.  The FX3 has an adjustable video filename format which helps, but if multiple cameras are used it can be necessary to add a "uniqueifier" suffix such as an incrementing serial number. 

If you are shooting in S-Log3, the built-in FCP conversion LUT does not work that well in some cases. An alternative is apply the S-Log3-to-Rec.709 conversion with the built-in "Custom LUT" effect and use a 3rd-party LUT which has exposure offsets. Here is one example by Alister Chapman:  www.xdcam-user.com/2018/08/new-venice-lo...er-v-look-v3-s-log3/

It is normally not necessary to have FCP "background rendering" enabled, nor to use optimized media. This has no image quality benefit on the final product but in a few cases might help performance. 

Sony's recommended exposure for S-Log3 is 61% IRE on a white target. You can set that by setting Zebra 2 to 61%. Most conversion LUTs are designed for that, but the above Chapman LUTs have offsets up to +/- three stops at 1/2-stop increments. He also has a "G1" and "G2" version which slightly diminish any green color cast.

I suggest importing media using "leave files in place", not "copy to library". Be cautious about using third-party plugins, as they can sometimes cause problems.

In general use multicam clips, not sync clips -- even if you have only one camaera and one external audio source. Sam Mestman discussed that here:  www.fcpworks.com/sync-or-multicam-clips/

Fully leverage FCP's ability to organize the material before putting any content on a timeline. Use keywords, favorites and rejects. Since a multicam does not inherit keyword or ratings from the parent clips, first sync the multicams, then mark keywords and favorites on the multicam, not the parent clips.

To avoid accidentally ending into the timeline a single-cam parent clip, after creating the multicams you can mark all parent clips as rejected and run the viewer using the "hide rejected" filter.

Thanks again!
Usually this is my workflow for documentaries:
I copy all my media cards to two drives using hedge, which automatically renames them. You can set it to add the "label" of the media file plus date plus project name etc..
Then I use tentacle sync studio to sync video to audio and export them as combined media files - it just copies video without transcoding and transcodes all audio tracks to linear pcm 24 bit mono tracks plus it converts timecode from the file or from the LTC audio to a "real" timecode track which is readable by FCPX.

This eliminates a lot of problems you can encounter later in the process.
In FCPX I don't have to sync anything.
In the past I had some problems when I sync clips in any NLE - for example often when I hit "reveal clip in browser" from the timeline it only leads me to the video clip without synced audio and so on.. So for me this is the cleaner way to do it.

The only slight disadvantage: You loose gyro-metadata of the sony files for stabilisation by syncing in advance. But as I keep the originals I could easily come back to them if needed.

FinalCutdown wrote:

The only slight disadvantage: You loose gyro-metadata of the sony files for stabilisation by syncing in advance. But as I keep the originals I could easily come back to them if needed.

You don't lose the stabilization data. A benefit of using a multicam (vs a sync clip) is you can later replace the parent clips composing the multicam, and all instances of that multicam in all projects will immediately inherit that. 

E.g, Say you sync a XAVC-HS CamA, CamB and ExtAudio clips in a multicam clip, then you edit 20 pieces of that multicam into five different timelines. Then you later decide you want to use Catalyst stabilization on a certain CamA clip within the multicam. You just import that original camera clip into Catalyst Browse, stabilize it, export as XAVC-I, import to FCP, open the multicam and replace that angle with the stabilized clip. All instances of that multicam in all timelines will instantly inherit the stabilized version.

Or you can add an additional angle to the existing multicam, then instances of that multicam in all projects will have that new angle available as an option, in case you only want some of the timelines to use the stabilized version. This is demonstrated at 25:17 in this video "How to Multicam edit Final Cut Pro": 

You can also use gyro stabilization within FCP using Gyroflow Toolbox, but I haven't tested this:  gyroflowtoolbox.io/

Re gyro stabilization, it can work well but you must have all other forms of stabilization off. E.g, lens optical stabilization and IBIS sensor stabilization must be off. Gyro stabilization works better at 60 fps due to the frame blurring that happens at 24 fps. Here is an example:  vimeo.com/598896853/2bde2933ac

Thanks. I meant in MY workflow I loose Gyro Data when I sync in Tentacle Sync and export the synced files so that video and synced audio are merged into one file. Then the new files don't contain stabilization metadata anymore.

The cameras record gyro data even when "active" stabilization is turned on - which is a combination of a stabilised sensor plus digital stabilisation. You can still use gyro data later to stabilise even further.

I tried gyro flow as well - its horrible. It doesn't come even close to catalyst browse and it doesn't use sony metadata so you have to apply all settings manually. I thing its more made for GoPros.

If you use a zoom lens on your sony camera and zoom while filming its almost impossible to get the settings right in gyro flow as you would have to constantly adapt the focal lens.. catalyst automatically obtains the focal length via metadata.

I really hope there will be an official plugin from sony in FCPx soon. I tried the Sony premiere plugin as well and it was constantly crashing. I wouldn't use it on a large project.

FinalCutdown wrote:

Thanks. I meant in MY workflow I loose Gyro Data when I sync in Tentacle Sync and export the synced files so that video and synced audio are merged into one file. Then the new files don't contain stabilization metadata anymore.

That is caused by the Tentacle Sync Studio desktop app. A possible solution is using the FX3's ability to use true "file timecode" which you can then sync in FCP using the "sync on timecode" option. You would need to use two adapter cables to connect the Tentacle Sync E to the FX3:

Tentacle Sync to BNC cable: 
www.bhphotovideo.com/c/product/1327595-R...ble_tentacle_to.html  

Sony VMC-BNCM1 cable:
www.bhphotovideo.com/c/product/1717858-R...e_adapter_cable.html

Here is a discussion of using file timecode on the FX3 using Deity devices. It's about the same for Tentacle, except you need different adapters:



Long term, using true file timecode is a better solution than using audio LTC timecode. If possible, make sure all your cameras and recorders support this.

FinalCutdown wrote:

...The cameras record gyro data even when "active" stabilization is turned on - which is a combination of a stabilised sensor plus digital stabilisation. You can still use gyro data later to stabilise even further...

Yes I noticed that The FX3 & A7SIII gyro data work in Catalyst Browse IF in active stabilization mode, but not in regular mode. I wasn't sure if that was a feature or a bug

On our FX6 cameras we don't have IBIS sensor-based stabilization, so we were interested in the various tradeoffs between gyro, lens optical stabilization and FCP stabilization in post, and how this varies between 23.98 fps (our normal documentary frame rate) and 59.94 fps. Below are some tests using all combinations:

FX6 stabilization tests at 200mm, 23.98 and 59.94 fps using combinations of gyro, OIS and FCP:

Thanks again!
I had the FX6 here for one week for testing as well. Its a great camera with much better ergonomics but the missing OIS was one of the main reasons for me to choose the FX3. Plus the FX3 needs less memory cards because of long GOP, need much less battery overall and so saves weight and AF tracking is better.
Mostly I am using the FX3 on a gimbal - PLUS active stabilization. I experimented with gyro stabilization a lot but the workflow is still so complicated and time consuming that I try to avid it at all costs.

However the timecode cable doesnt work right.
Its off by about 2,5 Frames in 50p constantly when the cable stays connected and when you pull it out it drifts very quickly. Sony knows about it and doesnt communicate it. Even Tentacle warns you about using the cable - they say they have the information from sony - on their own website theres nothing about it.
tentaclesync.zendesk.com/hc/en-us/articl...-my-Sony-FX3-camera-
NOTE:We have current information from Sony that there may be deviations of +/- 1 frame or even more when using the Sony timecode adapter cable VMC-BNCM1 with external timecode generators. Therefore we recommend the timecode synchronization via the 3.5mm Mic input as audio timecode if this is feasible in the workflow.
I know that loosing gyro data is caused by tentacle sync, thats what I tried to say in the first place. Its a little tradeoff of my workflow, but I feel much more comfortable when audio and video is in one file combined.

Do you get to the synced clip in Final Cut when you click "reveal in browser" for or does it lead you to the unsynced video clip?
In the past I had trouble with different LNEs not to loose my synced audio while editing, so syncing in advance is a little safer.

FinalCutdown wrote:

...Do you get to the synced clip in Final Cut when you click "reveal in browser" for or does it lead you to the unsynced video clip?..

We have several FX6 and RED cameras, also a DJI Ronin 4D and several Sound Devices audio recorders, each using a continuously-connected Tentacle Tracke E for file timecode, not audio timecode. We jam-sync our Denecke slate from a Tentacle. We do not use the Tentacle Desktop app to create an XML sync file, nor do we transcode the camera files. Rather we import them and sync using file timecode.  We've never had any sync problem in FCP except those which we caused by human or procedural error during production.

So with this method they all devices have file timecode and they are synced as needed in a multicam. In a few cases where we needed extra coverage from an A7SIII, instead of using Tentacle audio timecode, we manually preset the timecode by hand, which is usually within about 1/4 of a sec. That provides true file timecode and avoids using the Tentacle Desktop app. All clips from that camera are off the same amount so they can all be adjusted in a multicam with a single operation. Thus far, our five A7SIIIs have maintained good timebase stability over several hours.

All cameras must use the same DF vs NDF settings. Once we had an A7SIII with the wrong settings and it appeared to drift, but it was a setting problem. 23.98 by definition is NDF but the other frame rates can be set to DF or NDF, which is a possible source of sync error.

We don't use the "sync clip" type in FCP, only multicam clips. After creating and syncing the multicam, if I open it and use "Reveal in Browser" it points to the parent clip. 

Thats a great setup! For sure FX6 is the more professional camera.
For me Timecode with the Ultrasync one has been very reliable, especially as they constantly wirelessly stay in sync there is no drift at all - as long as I dont use the ne Sony-Cable.
For multicam setups time of day is perfectly fine in most situations.

By the way: FX3 file-timecode cannot be read in FCPX natively, in some other threads people were talking about a converter but I guess this will remove gyro metadata as well.

As I dont rely too much on gyro stabilization I would use it for single clips only and in this case I can still go back to the original clip and then sync manually with audio.

If you can always reveal the parent clip syncing in FCPX could be an option but as I have to convert timecode anyway its no more effort to sync and copy audio as well. Its just two clicks more for a whole shooting day.

FinalCutdown wrote:

...By the way: FX3 file-timecode cannot be read in FCPX natively, in some other threads people were talking about a converter but I guess this will remove gyro metadata as well.

Yes, I mentioned above the tool QTChange which will fix the FCP problem with Sony timecode in an MP4 container. It doesn't technically remove the gyro metadata (although it functionally appears so).  A binary compare of the two files shows that the gyro data is unchanged. However it somehow exposes a bug in Catalyst Browse where it will no longer read the gyro metadata. 

But that might be a non-issue. To use the gyro data you have to stabilize it in Catalyst Browse then export it. If you simply pick the MXF All Intra format (which is what the FX6 uses natively) the gyro-stabilized clip will also have timecode readable in FCP.

That is another reason we don't use our A7SIII cameras much, because the FX6 MXF format handles timecode better and has much richer metadata for troubleshooting various problems. That plus the FX6's built-in variable ND filter. We used the A7SIIIs a lot for gimbal work before we got the Ronin 4D, which has timecode support and built-in ND.  [/quote]

You are right Too many threads.. I forgot from whom I got the information.
Vario ND of the FX6 is great!
I was thinking about the ronin 4d before I bought the fx3 but I was worrying about ergonomics and the massive weight you cannot put on your shoulder plus I needed four audio channels so I could do without timecode and sound bag as much as possible. But the 4d is definitely the future I am sure much more cameras like it will come out in the future.

Its a pity that FCPX is still not able to convert LTC to timecode natively like resolve does.

joema wrote:

Gyro stabilization works better at 60 fps due to the frame blurring that happens at 24 fps.

Thanks for the video - the frame blurring is caused by shutter speed and not by the frame rate, but its much easier to dial the shutter speed down with a higher frame rate without getting a stuttering video. If you stay over 1/100 you should be finde. If you dial in a shutter angle in your camera, changing the framerate will automatically reduce shutter speed and blurring so it's the same.
I guess I will try to change my workflow, until now I was using Sony cameras on a gimbal only but maybe its better to shoot handheld with active stabilization turned on and fix some single clips with gyro stabilization in post. Still have to try out if it works even on interviews while walking with a very wide lens (16 or 20mm) I have to deliver in 1080i at the end so there is some resolution left in 4k to crop in.
I tried the new ZV-E1 as well with "dynamic active" stabilization, a feature that will come out in other cameras in the next months as well for sure, but its more for bloggers, slow tilt or pans are stuttering because the camera tries to stabilize first and then jumps back when digital crop window hits the border of the sensor.
In active stabilization its not that extreme.

FinalCutdown wrote:

...I guess I will try to change my workflow, until now I was using Sony cameras on a gimbal only but maybe its better to shoot handheld with active stabilization turned on and fix some single clips with gyro stabilization in post. Still have to try out if it works even on interviews while walking with a very wide lens (16 or 20mm) I have to deliver in 1080i at the end so there is some resolution left in 4k to crop in...

Each stabilization method has pros and cons. E.g, optical stabilization works well at longer focal lengths. Gyro stabilization can also work well there, but it can inject some slight image warping. At wider angles, optical stabilization does not work so well. Each IBIS sensor stabilization implementation is different. On the Panasonic GH5, the sensor is small and sensor housing is proportionally large, so it can be moved around a lot and achieve good results. With Sony Alpha mirrorless cameras, the sensor is larger and housing smaller, so it can't move as much, giving more limited results, especially at wider angles.

At wider angles, combining gyro stabilization with NLE stabilization can produce good results. Here is a test I did walking with the FX6 with a 28-135 f/4 lens at 28mm, 59.94 fps, 180-deg shutter, and no gimbal. I think I accidentally used a 23.98 timeline frame rate, so you can probably disregard the frame judder that introduced.

vimeo.com/617097074/aa54cef30c?share=copy

Yes, I have already seen some comparison videos of optical and digital.. I just tested the gyro plus active on the new 20-70mm f4 at 20mm. Looks good no wobbling at the frame edges although camera was shaking quite hard when walking around. But still I think I will try to combine gimbal and handheld. I think I will put stick the gimbal in a pocket of my belt and turn it off most of the time so I can use the Fx3 handheld and use it just to put the weight on my hips and when I want to walk during an interview I can quickly turn it on and use aperture control on the control wheel. The dji Rs2 seems a little weak for my Fx3 plus two wireless receivers though. Maybe I will upgrade to Rs3 pro for that and have bluetooth control for the camera so I can skip the cable which is always the weak point..

Render times are still very long even on a new m2 max.. More then 10 minutes for about 1,5 Minutes stabilized and exported in XAVC-L.

Btw I guess we are getting a little off topic here. maybe we should open a new thread.

Back to the main topic: The problem persists. Sony XAVC-S UHD 10 Bit 422 footage is slow while scrolling in the browser and playing back on a 1080i Timeline 25i only with a lut applied I get dropped frame warnings. I have three monitors connected, two for the application one for A/V Outupt.. I switched to high performance on the m2 max, still the same.. Whats going on?

Narrow down the problem. Do the exact same test but with all external monitors disconnected. Also if you have any wireless audio such as bluetooth speakers, headphones, etc, as a temporary troubleshooting step, disable those.

If you have a hard drive with external media connected via a hub, as a temporary troubleshooting step, connect that directly to your Mac. Many problems have been caused by hubs.

In general you can disable the dropped frame warnings. The issue is NOT whether a frame is dropped during playback, but whether it significantly affects your editing.

Also, please restate your codec specifics. Initially you said XAVC-HS (HEVC) @50p for (I think) 25i delivery, now you're saying XAVC-S (H264). Nothing wrong with that, since the FX3 won't shoot 25p in HEVC, but it will in XAVC-S. 

I just tested Sony 4k/25p XAVC-S 10-bit 4:2:2 and 4k/50p XAVC-HS 10-bit 4:2:2 on a 1080/25i timeline on an M1 Ultra Mac Studio running FCP 10.6.6 and Ventura 13.4.1. With one LUT applied, it seemed to run OK with no playback lag. That was without background rendering. I had the drop frame indicator enabled, and it never tripped.

My M1 Ultra should perform about equal to your M2 Max (maybe not even as fast) because for single-stream decoding, the parallel hardware accelerators cannot be used.

Re slow scrolling in the Event Browser, please define what you mean. In "filmstrip mode" (vs "list mode"), the clip thumbnails and waveforms must be generated as you scroll. It may be a little laggy on the first pass, afterward it should be fast. If you have View>Browser>Waveforms on, try turning those off and see if it makes a difference.

FCP I/O breaks down into two broad categories: (1) large sequential reads for media, and (2) small random I/Os for the library SQL data, thumbnail generation, and waveform generation. 

Ideally you want a "lean" library with all media imported using "leave files in place", then place your library (which by default contains SQL data, thumbnails, waveforms and cache) on a fast SSD.