Here is an example of the jittery playback (link is for the relevant time)
EDIT: Timestamped links appear not to work in this forum (of course). Go to 10:40 when the close shot changes to the wider shot.
(Don't watch it in this tiny window, watch it full screen at 60fps to really notice).
This clip was recorded in ProRes 422 on an Atomos. The clip directly preceding this one is actually the same clip, just zoomed in. None of the HEVC gameplay has this issue and the other ProRes shots don't have this issue either. It seems to appear at random. If someone asks why that part is so jittery, I'll answer "This is what Final Cut Pro sometimes does on my Mac" and I wouldn't be lying. I believe they screwed up with the ProRes acceleration on the Mac Studio. If it doesn't happen for you, that doesn't mean it's not the issue. Nothing about my equipment or workflow has changed since moving from my 2013 Mac Pro.
Fortunately due to the subject matter onscreen as the jitter is happening, people will probably just think it's an effect.
Here is an example of the jittery playback...This clip was recorded in ProRes 422 on an Atomos. The clip directly preceding this one is actually the same clip, just zoomed in...
It appears that clip has undergone some kind of rate conforming with an uneven cadence, or else the Atomos, FCP or something else duplicated, switched or dropped frames. In Youtube you can see that by single-stepping via the < and > commands around 10:40 while looking at your thumbs, not the on-screen part. So it's not isolated to the on-screen gameplay. I can also see it in Quicktime Player on a downloaded version of the Youtube upload.
By contrast the other non-gameplay portions show an even, consistent 2:1 rate conforming that's typical of 30 fps material in a 60 fps timeline.
Bear in mind I can only examine a downloaded version of the re-encoded Youtube video, which could have been altered. If you could use Quicktime Player to do no-encode trimming (CMD+T, then File>Save) of that small section around 10:40 -- both original Atomos ProRes 422 HQ file *and* the FCP export -- and upload those files to DropBox, etc, it would help us figure this out.
Sure, here is the clip straight from the Atomos (trimmed) as well as the file I uploaded to Youtube (trimmed). If I watch it back in Final Cut Pro, it no longer does this. Only occasionally..
This is to confirm I downloaded both clips and am examining them. I see the problem on the 1080p H264 clip, I assume that is one you exported and the behavior is now "baked in". The original Atomos 4k/29.97 ProRes 422 HQ clip looks OK. On my M1 Ultra I put the original 4k ProRes clip in a 59.94 timeline, did VLog color correction and exported it as 1080p/59.94 H264, but so far I don't see the problem manifest. I will look at it more tomorrow.
The problem appears not limited to frames in the wrong order. There are lots of duplicated frames in your 1080p H264 clip. You can see that by playing it in Quicktime Player, clicking on the timecode display until it displays frames, then single-stepping by pressing J and L while holding down the K key. There are various places where the frame number advances but the image does not change.
Do you think the duplicated frames you are seeing are part of the normal 2:1 cadence of a 29.97 clip in a 59.94 timeline?
When I put your original 4k/29.97 VLog clip in a 1080p/59.94 timeline, I only see the expected 2:1 cadence. IOW every other frame changes. I tried with no Fx or camera conversion LUT, with the built-in VLog LUT, and with that plus grading, rendered and unrendered.
I see it in your 1080p H264 clip, but whatever happened becomes "baked in" during the encode. That behavior is as you describe it -- periodic wrong frame order, but it also contains ranges where 3 or so frames are duplicated.
I've seen several reports of certain USB-C and Thunderbolt SSDs behaving incorrectly on M1 Max and Ultra. I'm studying one now where a certain 4TB QLC SSD behaves erratically on M1 Max and Ultra but more or less works on Intel. However further study is showing there's something wrong with that SSD, it only manifests quicker on the M1 CPUs. I don't have the mfg of that 4TB QLC -- it is bonded inside a NVMe enclosure.
Can you take the same files you sent, put those in a test library where library and media are all on the internal Mac Studio drive and try to reproduce it? Or do you already know that answer?
You mentioned it doesn't happen when there are unrendered Fx on the timeline. I wonder if that introduces a small but finite delay that prevents those drives from encountering this problem?
Your original description mentioned a suspicion of the Samsung 980 Pro but then you saw it on a Sandisk USB 3 SSD. In that case were all files, both library and media on the Sandisk?
I've tested several 2TB and 4TB Sandisk Extreme Portable USB-C SSDs on M1 Max and Ultra and so far haven't see any problems with those.
My limited testing of the Library on the internal drive has yielded no jitters, but I have only spent a few minutes testing. The problem shows up seemingly randomly. I wish I knew what prompted the issue into appearing.
Last night I exported a half hour video and it jittered in only one short part. I had since closed that Library. Opening it back up now it doesn't jitter at that part at all. I'm re-exporting it now which will take a few minutes.
I have to take back my assumption that having unrendered effects cures it, because I don't really render any more with this machine and the part that jittered certainly had color grading on it.
My limited testing of the Library on the internal drive has yielded no jitters, but I have only spent a few minutes testing. The problem shows up seemingly randomly. I wish I knew what prompted the issue into appearing....
This is the most difficult type of problem -- infrequent, random and with unclear causes. Simply identifying the problem is also hard -- if one frame in a 5 min video is skipped, duplicated or mis-ordered, there's no automated way to catch that. You basically must single step through every frame, then do that over and over with each test case.
There is a Python script that can attempt to scan video files for duplicate frames, but it's difficult to use, doesn't work reliably and is at best an approximation: github.com/ryanfox/retread
Have you ever seen it if exporting ProRes 422? With ProRes source, render and output files, that might eliminate Long GOP encoding issues.
Can you reformat your Sandisk or other external disk as MacOS Extended Journaled (HFS+) and try that? I normally use APFS on external SSDs but I've seen some odd problems that went away when I switched them to HFS+. Obviously the internal drive is APFS but that is a special case; we're just trying anything to give some clues on the problem boundaries.
Yes I've seen it when exporting to ProRes. I recall because I was doing a 5.1 surround video and I had to export that ProRes and then encode it with Handbrake because that's the only thing my Roku will see. I kept seeing the jittery playback and figured it was Handbrake. That was my first project on this new computer.
The SanDisk is just called "3D SSD" and it's 2 TB. I guess you can only edit 2D video on other SanDisk SSDs.
Not a bad idea to reformat one of the drives. That's gonna take a lot of copying back and forth but it's worth a shot.
The SanDisk is just called "3D SSD" and it's 2 TB. I guess you can only edit 2D video on other SanDisk SSDs.
To get more info about your SanDisk SSD, choose your Apple menu on your Mac and hold option to get "System Information" (first menu item). Once System Info's main window appears, under Hardware select USB and locate your SanDisk device. It should be labelled something like "Extreme SSD" (see attachment for an example). This info should be helpful.
There seems to be at least 4 "flavors" of the SanDisk Extreme SSD: Extreme, Extreme Pro, Extreme v2, and Extreme Pro v2. The v2 versions are faster and newer than the previous ones. Here's where you can get more info (you want to slect "Portable Drives" from the Product Category menu and SanDisk Extreme or SanDisk Extreme Pro from the Product Name menu):
Okay, cool. Thanks for the new screenshots. The product name/info provided in the System Information listing includes the SATA to USB bridge/chipset that's inside the cable. I thought it was one of the drives similar to what joema showed...
Both Sandisk Ultra 3D SATA III drive and the 2TB Samsung 980 Pro are TLC (Triple-Level Cell), not QLC (Quad-Level Cell). TLC typically has better sustained write performance and durability, which is good.
Even if it's a long shot, It makes sense to try reformatting one HFS+, then verify the problem can be reproduced there.
The 1080p clip you sent has more issues than just some out-of-order frames. The "jittery" behavior is caused by numerous groups of duplicated frames at fluctuating intervals, which means other frames were dropped.
On my M1 Ultra when I single step through that in Quicktime with select View>Time Display>Frame Count, I see the following cadence. This is not indicative of a rate conforming algorithm, but seems random. A "2" is the expected cadence of 30 fps in a 60 fps timeline. Each number is the # of times that frame was duplicated. As signposts I put some frame numbers in parenthesis.
I'm a bit concerned there might be a more widespread problem (whether hardware or software) which has so far eluded detection. It's tempting to say "I never saw this". But in this instance it's easy to see. What if certain configurations and actions caused one dropped/duplicated frame at longer but random intervals averaging around 5 minutes?
BTW it's not the game -- when the frames get duplicated, his hands holding the device freeze as well.
I have seen something vaguely like this on certain Atomos Ninja V recorders when they record ProRes 422 4k/59.94 fps on certain SATA III SSDs from a Sony A7SIII. However that is simply dropped/duplicated frames. Redifer's Atomos source material looks OK -- I checked every frame.
Redifer -- has any clip in this project been retimed or had optical flow applied? Also, what kind of Panasonic camera is that? Just as a comparison, have you tried using the internal codec in the same project, rather than the Atomos ProRes material?
I have that particular Library backed up on a removable drive at the moment but I don't think anything was retimed, certainly not any of the ProRes-sourced stuff. In a different Library which is still on a drive attached to the computer does have a couple of HEVC clips, one of them with optical flow. The ProRes clip that got jittery in this one has a Cube transition into a retimed clip, though optical flow is not applied to that particular one. Audio is not attached to either clip.
In Disk Utility, choose View -> Show All Devices. You should then see something like "Samsung SSD 980 PRO 2 TB Media" in the list in the sidebar. If you select that you should be able to choose Erase from the main window toolbar and select "Mac OS Extended (Journaled)" and GUID Partition Map...
I do not seem to format my Samsung 980 Pro drive to anything but APFS. I think once it's APFS it's stuck there unless there's a trick I'm unaware of.
It is non-intuitive. With Disk Utility in View>Show Only Volumes, it will let you erase an HFS+ volume and change it to APFS but not vice-versa. With Disk Utility in View>Show Only Volumes, you must first delete the partition. Then erase (ie format) it using the Erase button.