encode in 10-bit way worse

[Thunderbolt8]

I did some high bitrate encode of samples of a blu-ray with the same command line for 8-bit and 10-bit, and despite filesize and bitrate specified are the same in both cases, the quants for the 10-bit encode are way worse, 25-26 compared to 15-16 in case of the 8-bit encode (and it also looks like that during playback). arent the quants / the final PQ supposed to be a little bit better when encoding in 10-bit instead of 8-bit?

how can this be possible?

script: (source is 1080i60 and cannot be deinterlaced during playback with the usual freeware software, so I decided to encode to 23.976fps with dropping those other frames)

LoadCPlugin("G:\ffms2.dll")
ffvideosource("G:\00009.mkv")
LoadPlugin("G:\TIVTC.dll")
tfm(pp=0).tdecimate()
SelectRangeEvery(20000, 100)


cmd line

x264 --preset veryslow --tune grain --pass 1 --bitrate 45000 --stats "G:\00009.stats" --deblock -3:-3 --open-gop --bframes 8 --ref 6 --merange 32 --me esa --subme 10 --colorprim bt709 --transfer bt709 --colormatrix bt709 --output NUL "G:\00009.avs"
x264 --preset placebo --tune grain --pass 2 --bitrate 45000 --stats "G:\00009.stats" --deblock -3:-3 --open-gop --bframes 8 --ref 6 --merange 32 --me esa --subme 10 --colorprim bt709 --transfer bt709 --colormatrix bt709 --output "G:\test.264" "G:\00009.avs"

[LoRd_MuldeR]

Different quantizer scales?

[Thunderbolt8]

dunno what that means. is that really a question or an explanation and if the latter, what could I do here to improve the quants of the 10-bit encode?

[LoRd_MuldeR]

Well, the quantizer scale (i.e. the maximum possible quantizer value) depends on the bit-depth. See "common.h" for details.

This means that you can't compare quantizer values from an 8-Bit encode to those from a 10-Bit encode directly. They have different meanings, so to say.

Also why do you think anything needs to be improved?

Did you actually compare a 8-Bit and a 10-Bit encode of the same source and, apart from the bitness, identical settings - at the same average bitrate? :confused:

[Thunderbolt8]

yes, at same average bitrate.

so how can I translate the meaning of the 10-bit to 8-bit quants language?

as written in the OT, it needs to be improved, because the 10-bit video has blocking like hell

http://thumbnails22.imagebam.com/16095/9cb2f5160946634.jpg (http://www.imagebam.com/image/9cb2f5160946634)

[sneaker_ger]

Use a 10bit-compatible decoder.

[Thunderbolt8]

I failed, still had an old version of coreavc on my other comp -.- it looks ~same now in both cases :o

but coming back to the different quants, how do 8-bit quants translate to 10-bit quants, so how can I compare both on the same level?

[LoRd_MuldeR]

Why do you want to translate them back? Quantizer values alone aren't anywhere near to a good quality indicator.

(If they were, the "constant quantizer" mode would be a "constant quality" mode and adaptive quantization would be useless. We all know that isn't the case ^^)

[Dark Shikari]

Each extra bit of depth adds +6 to all quantizers for the same relative precision.