Psy RDO: Official testing thread (version 0.6 out!)

[Manao]

He never criticized x264. You're criticizing his options, yet --no-fast-pskip / --no-dct-decimate / --filter -2:-2 were still advised by Dark_Shikari. He didn't use subme 7 / b-rdo / umh (he stopped at subme 6 / hex ) but it's still a coherent choice.

In the end, though the number of bframes / vbv-bufsize seem off, they may probably have been set to match those used in Sonic CineVision, in order to make a fair comparison.

And, on a side note, if 7 years ago I had received such a welcome on doom9, I would never have bothered.

[Dark Shikari]

Originally Posted by Manao

And, on a side note, if 7 years ago I had received such a welcome on doom9, I would never have bothered.

The reason he got such an icy response was not because of the complaint about x264 quality, but rather because of the phrase "x264 apologists"; it is quite reasonable to interpret this as an attack, a troll, or both.

If you used such inflammatory and loaded language, you wouldn't have been welcomed here either. Fortunately, you're much more mature than that and I would never expect anything of that sort from you.

I do agree that most people here were far too sharp in their reaction; there seems to be a tendency for people to see others overreact, and interpret that as a signal that they can overreact too. It generally is a good idea to assume good faith at least initially, which most people seem to forget...

[foxyshadis]

Sharktooth: Calm down. You get way too riled up when anyone posts something provocative, and start flinging insults, instead of putting them in their place by calmly pointing out what's wrong. You'll end up suspended for rule 4 if you don't stop reacting angrily.

[Sagittaire]

Well just my 2 cents.

1) BR and AVCHD specifications
- HP@4.1 specification with limitation
- Buffer max for 1080p24: 30 000 bits
- Max bitrate at 40 Mbps for BR and 30 Mbps for AVCHD
- 4,3 for p,b frame reference, 3 pyramidal bframe imply 3,2 ref for p,b. Use 2 bframes with 4,3 ref is better for software compatibility.
- Max gop at 24 with IDR
- WPred
- FGM is not supported

2) At 15 Mbps you have not transparence with 42" and higher screen. From my experience x264 with good patch is by far better than Mainconcept SDK (cinevision) but only at low/medium bitrate.
x264 produce really impressive result in [6-8] MBps interval for 1080p24 content.

@REM -----------------------------------------------------------
@REM
@REM Profil BluRay 1080p23.976 extra high quality
@REM
@REM -----------------------------------------------------------


@REM Source file name (suffit de mettre la source ici)
set E_SRC=Lossless.avs

@REM Set of quality (ici la qualité 1-50)
set E_BR=21.3

@REM Set of max bitrate (ici le bitrate max)
set MAX_BR=40000

@REM Set of Buffer (ici le buffer)
set BUF_BR=30000

@REM Set credit (frame de début du générique)
set CRE_FR=201560

@REM Set end credit (frame de fin du générique)
set END_FR=207442



@REM Profil

x264.exe --threads auto --thread-input --keyint 24 --min-keyint 1 --crf %E_BR% --vbv-maxrate %MAX_BR% 
--vbv-bufsize %BUF_BR% --mvrange 511 --level 4.1 --bframe 3 --b-pyramid --b-rdo --bime --weightb --ref 3 
--mixed-refs --direct auto --deblock -2:-2 --ipratio 1.10 --pbratio 1.10 --partitions "all" --8x8dct --me "umh" 
--subme 7 --trellis 2 --no-fast-pskip --no-dct-decimate --aud --nal-hrd --sar 1:1 --cqmfile Sagittaire.cfg 
--psy-rd 1.0 --zone %CRE_FR%,%END_FR%,b=0.33 --progress --pass 1 --stats "stat.log" -o 1080p_Q1.264 %E_SRC%

pause

[smok3]

Sagittaire, what range of CRF values do you usually use? (%E_BR%), any special reason for 21.3?

[Sharktooth]

Originally Posted by foxyshadis

Sharktooth: Calm down. You get way too riled up when anyone posts something provocative, and start flinging insults, instead of putting them in their place by calmly pointing out what's wrong. You'll end up suspended for rule 4 if you don't stop reacting angrily.

i am calm, i did not offend or insulted anyone.
what may sound an offence is a mere constatation of facts.
btw trust me, you dont want to see me angry.

[bob0r]

Originally Posted by Sharktooth

..
btw trust me, you dont want to see me angry.

Trust me aswell, i once trew a blu-ray disc at Shark "Zangief" Tooth's head..... i couldn't walk/speak/blink my eyes/go to the rest room for 5 months.

[Sagittaire]

Originally Posted by smok3

Sagittaire, what range of CRF values do you usually use? (%E_BR%), any special reason for 21.3?

I make comparison with DarkNight uncompressed source.
x264 produce simply 12 Mbps at 21.3 for this encoding.
In this test x264 with psyrdo patch is a real killer.

[chainring]

Originally Posted by Manao

He never criticized x264. You're criticizing his options, yet --no-fast-pskip / --no-dct-decimate / --filter -2:-2 were still advised by Dark_Shikari. He didn't use subme 7 / b-rdo / umh (he stopped at subme 6 / hex ) but it's still a coherent choice.

Argh, my head is hurting! I've tried keeping up with all the patch threads, the current version thread, and now I see, what seems to be, conflicting info on --no-fast-pskip / --no-dct-decimate / --filter -2:-2.

So, for a person using build 920, psyRDO on (default), trellis 2 (don't care about speed loss), and using CRF; what is the correct application of these settings? I've now seen Sharktooth say --no-fast-pskip and --no-dct-decimate are bad. I've seen Dark_Shikari say to just go with default filter values (0:0). Now, there's conflicting info in what I've quoted from Manao.

Any help?

BTW, here's my commandline as it stands now.

--crf 20.0 --ref 5 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 7 --trellis 2 --partitions p8x8,b8x8,i8x8 --8x8dct --me umh --threads auto --thread-input --progress --no-dct-decimate --no-psnr --no-ssim --output "output" "input"

[Dark Shikari]

Originally Posted by chainring

Argh, my head is hurting! I've tried keeping up with all the patch threads, the current version thread, and now I see, what seems to be, conflicting info on --no-fast-pskip / --no-dct-decimate / --filter -2:-2.

So, for a person using build 920, psyRDO on (default), trellis 2 (don't care about speed loss), and using CRF; what is the correct application of these settings? I've now seen Sharktooth say --no-fast-pskip and --no-dct-decimate are bad. I've seen Dark_Shikari say to just go with default filter values (0:0). Now, there's conflicting info in what I've quoted from Manao.

Any help?

BTW, here's my commandline as it stands now.

--crf 20.0 --ref 5 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --subme 7 --trellis 2 --partitions p8x8,b8x8,i8x8 --8x8dct --me umh --threads auto --thread-input --progress --no-dct-decimate --no-psnr --no-ssim --output "output" "input"

--no-fast-pskip is not going to hurt quality. It might not help a lot, but it doesn't hurt.

[Manao]

I never given any settings on this thread. I just compared Vit's initial settings from those advised by Dark_Shikari and pointed out settings that were present in both, yet subject to polemic : no-fast-pskip, no-dct-decimate, filter -2:-2.

I don't know whether no-fast-pskip and no-dct-decimate ought to be used or not (instinctively, i'd say the first is good, and so is the second if trellis 1/2 is used). As for filter, between -3:-3 and 1:1, it's imho a subjective matter.

[chainring]

Hi Manao,

Thank you for the reply!

I wasn't meaning to imply YOU were giving the settings, per se, but just quoting what you were in turn quoting.

Originally Posted by Manao

I never given any settings on this thread. I just compared Vit's initial settings from those advised by Dark_Shikari and pointed out settings that were present in both, yet subject to polemic : no-fast-pskip, no-dct-decimate, filter -2:-2.

I don't know whether no-fast-pskip and no-dct-decimate ought to be used or not (instinctively, i'd say the first is good, and so is the second if trellis 1/2 is used). As for filter, between -3:-3 and 1:1, it's imho a subjective matter.

[CruNcher]

Originally Posted by Vit

I've encoded full STEM clip at 15 MBits/s using this patch and Sonic CineVision 2.6. Sonic's encode is consistently better, am I doing something wrong or commercial encoders are not as crappy as some x264 apologists claim?

Here's the command line:
x264.910.modified.exe --pass 1 --fps 23.976 --bitrate 15000 --level 4.1 --keyint 24 --min-keyint 1 --ref 3 --mixed-refs --no-fast-pskip --bframes 2 --b-pyramid --bime --weightb --filter -2,-2 --subme 6 --analyse p8x8,b8x8,i4x4,i8x8 --8x8dct --vbv-bufsize 45000 --vbv-maxrate 30000 --threads auto --thread-input --progress --no-dct-decimate --output "f:\stem_x264.264" "F:\Source Files\StEM_full.yuv" 1920x1080 --aq-mode 2

Second's pass settings are the same, but with --pass 2. I can post some pictures if anyone is interested

Hi and welcome do you also have a Speed Comparision i guess Cinevision 2.6 uses the new G4 Encoder ?

[Sharktooth]

he used wrong settings for x264. you cant compare speed and quality of 2 different encoders if you cant set up them (or at least one of them).

[CruNcher]

it's only important that the G4 Encoder settings where of the same nature then the X264 settings, tough we dont know what the Cinevision 2.6 settings where if they where equal you could very well compare it especialy the decission side of things b-frame decission, partition decission, scenecut and b-pyramid decission is very important here. Compression/Complexity wise i agree Vit his settings aren't as effective as Sagittaire but to call them wrong hmm sure in terms of BD profile there way of too to what you normaly would use but it doesnt matter in his comparision he did with Cinevision 2.6 in this case for this very special Film source.

[Sharktooth]

he said cinevision was constantly better (subjective, without pics or clips) so probably he used much different settings and we also dont know anything about the source and cinevision settings.
he just came here, trolled and gone away.
he just didnt contributed in any way to the discussion, so, for what concerns me, he could even not post at all.

[CruNcher]

Originally Posted by Sharktooth

he said cinevision was constantly better (subjective, without pics or clips) so probebly he used much different settings and we also dont know anything about the source and cinevision settings.
he just came here, trolled and gone away.
he just didnt contributed in any way to the discussion, so, for what concerns me, he could even not post at all.

Yes i agree his behaviour for what seems to be a Professional Post House Encoder is a little weired indeed, but it doesn't mean his experience was wrong, tough i would thrust Sagittaire 1000x times more

@Sagittaire
I hope we gonna see a Cinevision 2.6 vs latest X264 GIT and X264 GIT + Psy RD Blu-Ray restricted Film Source comparission from you in the future

[Sagittaire]

Originally Posted by CruNcher

@Sagittaire
I hope we gonna see a Cinevision 2.6 vs latest X264 GIT and X264 GIT + Psy RD Blu-Ray restricted Film Source comparission from you in the future

1) Well from Mutek (Elecard SDK enginer): "there are not quality gain for G4 but only speed optimisation"
http://forum.doom9.org/showthread.ph...67#post1149767

2) Moreover you can't use all the insane SDK setting in cinevision 2.5 and certainely not in cinevision 2.6 too. Cinevision is pro gui for high bitrate encoding. Certainely better here to use Reference or Converter for have the insane setting.

3) Psy RD patch for x264 is a real killer. Certainely the best patch that I have ever seen for x264. RDO for bframe decision don't make very higher quality like I expected but some small optimisation produce big improuvement at the end.

[CruNcher]

Originally Posted by Sagittaire

1) Well from Mutek (Elecard SDK enginer): "there are not quality gain for G4 but only speed optimisation"
http://forum.doom9.org/showthread.ph...67#post1149767

2) Moreover you can't use all the insane SDK setting in cinevision 2.5 and certainely not in cinevision 2.6 too. Cinevision is pro gui for high bitrate encoding. Certainely better here to use Reference or Converter for have the insane setting.

3) Psy RD patch for x264 is a real killer. Certainely the best patch that I have ever seen for x264. RDO for bframe decision don't make very higher quality like I expected but some small optimisation produce big improuvement at the end.

So you experience tells you --b-rdo and it's resulting 20% speed lose can be saved now when useing psy-rd alone (visualy) ? that's great news seeing that RD is allready faster then some month ago and saving the 20% from --b-rdo without much of a subjecive visual quality lose that's really a big win and also what my subjective tests are indicating (tough im still testing)
And no it wasn't my intention to compare Insane settings vs cinevision non setable settings it was my intention to compare the decissions and the speed/quality difference if the're any for a Blu-Ray Framework @ exactly those High Bitrates and gop restrictions

[Dark Shikari]

Originally Posted by CruNcher

So you experience tells you --b-rdo and it's resulting 20% speed lose can be saved now when useing psy-rd alone (visualy) ? that's great news seeing that RD is allready faster then some month ago and saving the 20% from --b-rdo without much of a subjecive visual quality lose that's really a big win and also what my subjective tests are indicating (tough im still testing)

I'm pretty sure he's talking about --b-adapt 2, not --b-rdo...