Hi,
I have noticed problems with UHD HDR10 encodes when using Intel Quick Sync hardware accelerated crf encoding with the UHD Ripper module.
There appear to be slightly different problems with the UHD Converter module so I'm starting with the Ripper module issues first.
Encodes using 1-pass work and playback fine so clearly the NUC8i7HNK hardware, and the Intel graphics driver it requires be installed support HDR10 encodes, and DvdFab v11.0.0.2 itself is capable of doing these encodes.
I think being able to use crf for these encodes is important because a crf encode should always have better quality than a 1-pass for a given output size and using a 2-pass software encode takes a very, very long time to complete.
First, the size of the resulting encode is much larger than it should be.
For example using a crf of 30 produced a video of 36.9 GiB, as can be seen in the mediainfo output, from a source around 60 GiB.
Second, the video metadata for HDR10 crf encodes does not include all the elements it should for correct playback.
For an HDR10 encode the colour space is different from SD and HD videos and therefore needs to be specified in the video metatdata as players will default to the colour space used for SD and HD videos.
As can be seen from the mediainfo output for the 1-pass and crf encodes the 1-pass has the expected metadata while the crf encode does not.
For example the crf encode has:
Mastering display color primaries : Display P3
Mastering display luminance : min: 0.0050 cd/m2, max: 4000 cd/m2
Maximum Content Light Level : 4471 cd/m2
Maximum Frame-Average Light Level : 1007 cd/m2
While the 1-pass has:
Color primaries : BT.2020
Transfer characteristics : PQ
Matrix coefficients : BT.2020 non-constant
Mastering display color primaries : Display P3
Mastering display luminance : min: 0.0050 cd/m2, max: 4000 cd/m2
Maximum Content Light Level : 4471 cd/m2
Maximum Frame-Average Light Level : 1007 cd/m2
The mastering display and light level metadata that was included in the source video is present but the colour space metadata has been omitted.
I originally saw these problems in dvdfab v10 and when I investigated it then I also found the resulting video appeared corrupted when playback was attempted.
I don't see the point in trying to play these back again until the problems here are resolved so I haven't done so.
I have attached fabcheck, internal logs and mediainfo descriptions for the crf encodes in this paost and will post the same for the 1-pass encodes in the next post.
Ian
I have noticed problems with UHD HDR10 encodes when using Intel Quick Sync hardware accelerated crf encoding with the UHD Ripper module.
There appear to be slightly different problems with the UHD Converter module so I'm starting with the Ripper module issues first.
Encodes using 1-pass work and playback fine so clearly the NUC8i7HNK hardware, and the Intel graphics driver it requires be installed support HDR10 encodes, and DvdFab v11.0.0.2 itself is capable of doing these encodes.
I think being able to use crf for these encodes is important because a crf encode should always have better quality than a 1-pass for a given output size and using a 2-pass software encode takes a very, very long time to complete.
First, the size of the resulting encode is much larger than it should be.
For example using a crf of 30 produced a video of 36.9 GiB, as can be seen in the mediainfo output, from a source around 60 GiB.
Second, the video metadata for HDR10 crf encodes does not include all the elements it should for correct playback.
For an HDR10 encode the colour space is different from SD and HD videos and therefore needs to be specified in the video metatdata as players will default to the colour space used for SD and HD videos.
As can be seen from the mediainfo output for the 1-pass and crf encodes the 1-pass has the expected metadata while the crf encode does not.
For example the crf encode has:
Mastering display color primaries : Display P3
Mastering display luminance : min: 0.0050 cd/m2, max: 4000 cd/m2
Maximum Content Light Level : 4471 cd/m2
Maximum Frame-Average Light Level : 1007 cd/m2
While the 1-pass has:
Color primaries : BT.2020
Transfer characteristics : PQ
Matrix coefficients : BT.2020 non-constant
Mastering display color primaries : Display P3
Mastering display luminance : min: 0.0050 cd/m2, max: 4000 cd/m2
Maximum Content Light Level : 4471 cd/m2
Maximum Frame-Average Light Level : 1007 cd/m2
The mastering display and light level metadata that was included in the source video is present but the colour space metadata has been omitted.
I originally saw these problems in dvdfab v10 and when I investigated it then I also found the resulting video appeared corrupted when playback was attempted.
I don't see the point in trying to play these back again until the problems here are resolved so I haven't done so.
I have attached fabcheck, internal logs and mediainfo descriptions for the crf encodes in this paost and will post the same for the 1-pass encodes in the next post.
Ian
Comment