fbpx
Welcome, Guest
Username: Password: Remember me
{JFBCLogin}
25 Jan 2021
New boarders will have their posts moderated - Don't worry if you cannot see your post immediately.
Read More...
  • Page:
  • 1

TOPIC:

10 bit timeline question for you 05 Jul 2022 05:52 #121231

Hello group, I have a question about fcp 10.6.3, I have a couple of Canon 10 bit cameras and did a series of videos shooting 10 bit 4.2.2 in log and I am having some questions....

I am not seeing or understanding if fcp x is editing in 10 bit or does it drop the files down to 8 bit.

Then if you happen to know, when outputting in compressor, if I select HEVC 10 bit I understand compressor is using a software based compression instead of a hardware and wondered if anyone understands why and if you think this is a decent compression export format. (to keep file size down). Thank you very much in advance for any of your knowledge....

Please Log in to join the conversation.

10 bit timeline question for you 09 Jul 2022 14:56 #121286

  • DaveM
  • DaveM's Avatar
  • Offline
  • Elite Member
  • Elite Member
  • Posts: 201
  • Karma: 1
  • Thank you received: 49

Hello group, I have a question about fcp 10.6.3, I have a couple of Canon 10 bit cameras and did a series of videos shooting 10 bit 4.2.2 in log and I am having some questions....

I am not seeing or understanding if fcp x is editing in 10 bit or does it drop the files down to 8 bit.

Good question. While BlackmagicDesign is pretty up front about DaVinci Resolve's color processing pipeline (32-bit floating point, I believe), Apple is as transparent as a foot-thick concrete wall. It seems that the best one can do with respect to FCP is to infer (unless I'm missing something obvious)...

Both Library color space options (Standard or Wide Gamut HDR) support 10-bit video. The Wide Gamut HDR (Rec. 2020) Library option likely supports up to 12-bit video if it's following the specs. properly.

As a test, if you have a properly calibrated 10-bit capable display and GPU (and suitable settings), you can import a 10-bit and 8-bit grayscale ramp image. If both appear smooth, then you have a 10-bit display configuration.

Consideration of color spaces larger than Rec. 709 (DCI or Display P3 or Rec. 2020) is a bit more murky. FCP converts imported media into the Library's working color space for processing. FCP then converts outputs to the Project color space. What you see on your display is a combination of FCP's Project color space along with your display's color profile. The accuracy of what you see on your display depends on the hardware capbilities of your display along with the quality of the display profile being used. For example, if you have a display that is capable of sRG or Rec. 709 (same color gamuts but different gammas for display use), you are not going to be able to see a proper, accurate image for DCI, Display P3, or Rec. 2020 imagery.

There is a lot more to this, as far as color processing is concerned, which I won't address here.

Then if you happen to know, when outputting in compressor, if I select HEVC 10 bit I understand compressor is using a software based compression instead of a hardware and wondered if anyone understands why and if you think this is a decent compression export format. (to keep file size down). Thank you very much in advance for any of your knowledge....


The choice of output/delivery format, and encoder settings, depends on your desired result or needs. In most cases, the choice between SW versus HW encoding should not affect quality. Often, with SW encoders you have more options to choose from, whereas HW encoders may be faster.

Hope this helps...

Please Log in to join the conversation.

Last edit: by DaveM.

10 bit timeline question for you 10 Jul 2022 13:28 #121299

  • joema
  • joema's Avatar
  • Offline
  • Platinum Member
  • Platinum Member
  • Posts: 2188
  • Karma: 27
  • Thank you received: 533

...fcp 10.6.3, I have a couple of Canon 10 bit cameras and did a series of videos shooting 10 bit 4.2.2 in log...I am not seeing or understanding if fcp x is editing in 10 bit or does it drop the files down to 8 bit...


If the source codec is 10-bit 4:2:2, that should be preserved during the edit. When you export, you have a choice of codecs which may or may not preserve that. So there is a difference between the edit phase and export phase. If the export codec is 10-bit 4:2:2, the original bit depth and chroma sampling should be preserved.

...if you happen to know, when outputting in compressor, if I select HEVC 10 bit I understand compressor is using a software based compression instead of a hardware and wondered if anyone understands why and if you think this is a decent compression export format. (to keep file size down)....


I just did some tests on my M1 Ultra of importing R3D 8k 23.98 clips from a RED V-Raptor (which is 16-bit RAW), sending to Compressor and exporting to UHD 4k/23.98 10-bit 4:2:2 HEVC at various bitrates from 4 mbps to 50 mbps. It was very fast. CPU was low and GPU was high, which indicates the GPU was used for debayering the R3D footage and hardware accelerator was used for encoding HEVC.

Technically RAW footage does not have a chroma sampling since that only exists after debayering but in loose terms you can consider it to be 4:4:4:4. In this case it appeared to preserve 10-bit 4:2:2 in the HEVC output and it appeared to be hardware accelerated, at least on the M1 Ultra, FCP 10.6.3 and Monterey 12.4.

There are apparently cases where 10-bit HEVC export is not hardware accelerated, e.g, if you export from FCP and select the 10-bit Apple Devices HEVC preset. This might involve whether that format uses "open GOPs", aka dependent GOPs vs "closed GOPs", aka independent GOPs.

Long GOP formats are often pictured as being composed of independent Groups Of Pictures, IOW encoding or decoding one GOP in a file has no dependency on other GOPs. In those cases you can theoretically accelerate the encode/decode of those by processing them in parallel up to the hardware limits of the machine.

However some HEVC formats are built with dependent GOPs whereby you cannot process them in parallel. The advantage of those is higher compression, as you might want on a mobile device. There are 3rd-party tools to examine the file encoding but I don't have those.

I suspect the huge performance differences in some of the FCP HEVC 10-bit encoding cases may involve whether the export format uses dependent or independent GOPs, which in turn may allow or disallow use of hardware acceleration for encoding.

The M1 Ultra has four H264/HEVC encoders, so in theory it could segment a single input stream and encode those segments in parallel (if independent GOPs), then concatenate the segments to a single file. It does not seem to do that currently but maybe that could be a future enhancement.

The M1 Ultra seems to be using both H264/HEVC decoders in parallel on multi-stream input. E.g, if you have four 4k/23.98 10-bit 4:2:2 XAVC-S files, and scale those to 25% so they all appear in the viewer, and no other effects, the timeline responsiveness is significantly better on M1 Ultra than M1 Max. It is not CPU bound, GPU bound or I/O bound, which implies the difference is caused by the Ultra's additional decoder. But that situation only currently exists for multiple streams on input.

See attached graphics showing the various accelerators on M1 Pro, Max and Ultra.
Attachments:

Please Log in to join the conversation.

Last edit: by joema.
  • Page:
  • 1