madVR - high quality video renderer (GPU assisted)

[glc650]

Hi folks,

I've come across a few videos that will eventually (usually within 15-20 minutes of playback from any starting point) drop a frame or more (as reported by madvr and my eyes) followed by a major shift in clock deviation (from 0.000xx to as much as 0.0xxxx & from 1 frame drop every 1.0x hours to a drop every few minutes). All other media plays without this issue. The issue occurs with the default madvr settings as well. Here's the media info from one of the troublesome videos:

General
Unique ID                                : 191280449393230205331942369617805654636 (0x8FE7464759CE9273CFFDDCA26ED93A6C)
Complete name                            : /Volumes/downloads/TV/Proven Innocent/Proven Innocent S01E01.mkv
Format                                   : Matroska
Format version                           : Version 4
File size                                : 1.22 GiB
Duration                                 : 45mn 19s
Overall bit rate                         : 3 842 Kbps
Encoded date                             : UTC 2019-02-16 04:39:03
Writing application                      : mkvmerge v21.0.0 ('Tardigrades Will Inherit The Earth') 64-bit
Writing library                          : libebml v1.3.5 + libmatroska v1.4.8

Video
ID                                       : 1
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4.1
Format settings                          : CABAC / 9 Ref Frames
Format settings, CABAC                   : Yes
Format settings, ReFrames                : 9 frames
Codec ID                                 : V_MPEG4/ISO/AVC
Duration                                 : 45mn 19s
Bit rate                                 : 3 392 Kbps
Width                                    : 1 280 pixels
Height                                   : 716 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Constant
Frame rate                               : 23.976 (23976/1000) fps
Original frame rate                      : 23.976 (24000/1001) fps
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.154
Stream size                              : 1.07 GiB (88%)
Writing library                          : x264 core 157 r2935 545de2f
Encoding settings                        : cabac=1 / ref=9 / deblock=1:0:0 / analyse=0x3:0x113 / me=umh / subme=8 / psy=1 / psy_rd=1.00:0.00 / mixed_ref=1 / me_range=16 / chroma_me=1 / trellis=2 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-2 / threads=12 / lookahead_threads=2 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=3 / b_pyramid=0 / b_adapt=1 / b_bias=0 / direct=3 / weightb=0 / open_gop=0 / weightp=2 / keyint=230 / keyint_min=22 / scenecut=40 / intra_refresh=0 / rc_lookahead=50 / rc=crf / mbtree=1 / crf=18.0 / qcomp=0.60 / qpmin=0 / qpmax=51 / qpstep=4 / vbv_maxrate=17500 / vbv_bufsize=17500 / crf_max=0.0 / nal_hrd=none / filler=0 / ip_ratio=1.40 / aq=1:1.00
Default                                  : Yes
Forced                                   : No
Color range                              : Limited
Matrix coefficients                      : BT.709

Audio
ID                                       : 2
Format                                   : AC-3
Format/Info                              : Audio Coding 3
Commercial name                          : Dolby Digital
Codec ID                                 : A_AC3
Duration                                 : 45mn 19s
Bit rate mode                            : Constant
Bit rate                                 : 448 Kbps
Channel(s)                               : 6 channels
Channel layout                           : L R C LFE Ls Rs
Sampling rate                            : 48.0 KHz
Frame rate                               : 31.250 fps (1536 SPF)
Bit depth                                : 16 bits
Compression mode                         : Lossy
Delay relative to video                  : 48ms
Stream size                              : 145 MiB (12%)
Language                                 : English
Service kind                             : Complete Main
Default                                  : Yes
Forced                                   : No

Text
ID                                       : 3
Format                                   : UTF-8
Codec ID                                 : S_TEXT/UTF8
Codec ID/Info                            : UTF-8 Plain Text
Duration                                 : 44mn 40s
Bit rate                                 : 124 bps
Count of elements                        : 983
Stream size                              : 40.6 KiB (0%)
Language                                 : English
Default                                  : No
Forced                                   : No

Thanks,

->g.

[tp4tissue]

Originally Posted by ashlar42

Source? The only tech that I knew had potential is MicroLED. What's this you're talking about?

Google Panasonic IPS light modulation layer.

This technology is already available, it's only used in the Eizo Reference HDR monitor right now, CG3145. $30,000 msrp, $45,000 street price, (highly in demand)

But this tech vastly surpasses OLED's capabilities, because it will EVENTUALLY allow for Backlight blink motion blur reduction. (ULMB)

OLED can not possibly match the motion clarity once that is implemented.



Hisense and TCL also have a variant of this light modulation layer tech using VA-panel. They will be introducing it under their ULED lines in a short while.

[tp4tissue]

Originally Posted by mkohman

Evening all, apologies in advance if this is a silly question however for those guys who use madVR along with a HTPC and projector, what setting is selected under calibration within madVR? BT2020 with power curve gamma 2.2 or DCI-P3 with power curve 2.2? The reason I ask is I know that 4HDR movies are supposed to be BT2020 but madVR defaults them to DCI-P3 and when I select DCI-P3 the colours seem to have a slightly more pop to them as BT2020 is slightly washed out in colour ? So was wondering in madVR is it correct to select this display is calibrated to BT2020 or DCI-P3? Here are some pics of my projector when DCI-P3 is selected . Colours are slightly more saturated and HDR has a bit more pop . Would appreciate your advice . Thank you ..

I believe it's recommended you skip the gamma setting page,

Just use HDR to SDR , and when you calibrate, run the porjector setting at its 2.2 gamma response setting, and also Select 2.2 in Dispcal/HCFR

[tp4tissue]

Originally Posted by Sunset1982

ok guys, my htpc runs with an vega 56 for 2 weeks now. I am happy with it, accept for two things:

coil whine and oversaturated reds with the new amd card...

I will try to underclock the card to get rid of the coil noise. But I don't know what to do with the overstaurated reds...

Didn't have the problem with my old Rx 480, but I used it together with my lg oled.

Now I use my vega 56 with my girlfriends old sony lcd tv and the oled moved into the living room. So the vega or the combination of amd and the sony lcd caused this errror...

Originally Posted by madjock

Sounds familiar ?

https://forum.doom9.org/showthread.p...ds#post1852154

Cough.. I kinda liked the Blown out red..

But does Calibration not rein it in ?

Also, are you guys sure HDMI intent has nothing to do with this, because I know that one of my tvs will go into a Different Gamut mode with significantly wider red and green with a different render intent.

So if the driver is triggering something in the TV to switch gamut space, that may be the issue, and not necessarily caused by madvr/amd

[tp4tissue]

Originally Posted by Siso

Is it worth pumping NGU Sharp high, instead of NGU Sharp medium for luma, or using error diffusion 2 for dithering. Both settings increase rendering times to ~ 24 ms.

If you're Green Conscious and have a big tv viewed from 3-4 meters away, probably not.

But if you're rich, and throw money at problems, NGU FTW. !!

[tp4tissue]

Originally Posted by clsid

A two hour 10bit 60 fps UHD movie would require 6.8 terabyte of storage space in uncompressed format. With lossless compression approximately a quarter of that. So still a very large amount of space.

Apart from the above, what you want is not possible yet.

Originally Posted by madjock

So if we forget that this is not possible at all, how would you re-encode a whole movie with "super-demanding settings to be played on MUCH lower powered hardware"

If you don't have the hardware to play a file, what is going to re-encode it.

Ok, so, If we had say a 750ti. benches 3747.

It won't be able to push NGU upscaling 1080p to 4K.

But if we had it render out into a 24fps 4K file, It should be able to playback the file straight with no upscaling algorithm required, no ??

I mean, this is probably niche, maybe a novelty, it sounded plausible when I first thunk it..

[el Filou]

You would have to then store that file which means - if you don't want to consume terabytes storing it losslessly as clsid noted - you'll have to encode it again with a lossy codec which means loss of quality.

[ryrynz]

Originally Posted by tp4tissue

Google Panasonic IPS light modulation layer..

Three years though? Not even close. OLED is here for much longer than that.

[Msarc]

Originally Posted by tp4tissue

But if you're rich, and throw money at problems, NGU FTW. !!

Even though I can afford NGU, to be honest, I still prefer the ReconSoft+JincAR combo, quality-wise. It preserves the original picture far better to my eye. And when source artifacts are a problem, RCA is just the tool for the job, regardless of the upscaler.

It might be different for extreme cases like upscaling 480p to 4K, though. Haven't had anything like that to watch in a while.

[Sunset1982]

Originally Posted by madjock

Sounds familiar ?

https://forum.doom9.org/showthread.p...ds#post1852154

yah, seems to be the same error. I will do some test and report back...

[ashlar42]

Originally Posted by tp4tissue

Google Panasonic IPS light modulation layer.

This technology is already available, it's only used in the Eizo Reference HDR monitor right now, CG3145. $30,000 msrp, $45,000 street price, (highly in demand)

But this tech vastly surpasses OLED's capabilities, because it will EVENTUALLY allow for Backlight blink motion blur reduction. (ULMB)

OLED can not possibly match the motion clarity once that is implemented.

Hisense and TCL also have a variant of this light modulation layer tech using VA-panel. They will be introducing it under their ULED lines in a short while.

Yeah, thanks to you and nevcariel I arrived at that product... what can I say? I think your forecast of three years is wildly optimistic. But I obviously hope you are right.

[Warner306]

Originally Posted by mkohman

Thank you so in simple terms, do I select DCI-P3 or BT2020 in madVR calibration setting for my JVC x9900 projector? thank you

PS: I am using madVR pixel shader option not passthrough.. so I am doing HDR to SDR via pixel shaders

Either DCI-P3 or BT.2020 could work. If you are seeing a significant difference between the two, one or the other may not be working correctly.

There could be a couple of reasons:

- The projector is misinterpreting DCI-P3 or BT.2020. It may not recognize a DCI-P3 source, so it would more likely be DCI-P3. HDR content is supposed to report BT.2020 to the projector because that is dictated by the HDR specification standard. You own an AMD GPU, so you can't check report BT.2020 to the display to make DCI-P3 appear as BT.2020.

- The AMD video driver is causing over or undersaturation with DCI-P3 or BT.2020.

I would test SDR BT.709 content and see if it looks accurate. The colors should look natural without any oversaturation. To test accuracy, also try setting both madVR and the projector to BT.709 and play some SDR BT.709 content.

[oldpainlesskodi]

Originally Posted by Msarc

Even though I can afford NGU, to be honest, I still prefer the ReconSoft+JincAR combo, quality-wise. It preserves the original picture far better to my eye. And when source artifacts are a problem, RCA is just the tool for the job, regardless of the upscaler.

It might be different for extreme cases like upscaling 480p to 4K, though. Haven't had anything like that to watch in a while.

And, NGU Sharp vs Jinc produces very different black levels too @ 23.976 (on my rig anyway)...

Test file here for anyone interested in comparing...

https://ufile.io/1vy8u

[madjock]

Originally Posted by oldpainlesskodi

And, NGU Sharp vs Jinc produces very different black levels too @ 23.976 (on my rig anyway)...

Test file here for anyone interested in comparing...

https://www.justbeamit.com/9vsmz

Link is down mate ?

[Siso]

Originally Posted by madjock

Link is down mate ?

Yes, it seems down...

[oldpainlesskodi]

Originally Posted by madjock

Link is down mate ?

Link updated

[oldpainlesskodi]

Originally Posted by Siso

Yes, it seems down...

Link updated

[madjock]

Originally Posted by oldpainlesskodi

Link updated

Yeah working now, well I use Jinc for upscaling Chroma to 2160p, but I still see 18 Flashing faintly which is as good as I could get it with this TV.

What scaler are you saying there is issues with ?

[oldpainlesskodi]

Originally Posted by madjock

Yeah working now, well I use Jinc for upscaling Chroma to 2160p, but I still see 18 Flashing faintly which is as good as I could get it with this TV.

What scaler are you saying there is issues with ?

1080p material (and below) scaled to 4k - NGU sharp Luma produces very different black levels, hence sharing the link for people to test on their setup, if they want to.

[madjock]

Originally Posted by oldpainlesskodi

1080p material (and below) scaled to 4k - NGU sharp Luma produces very different black levels, hence sharing the link for people to test on their setup, if they want to.

Ah ok but better or worse, tried NGU and I can not see a difference tbh, the numbers stayed the same as far as I can tell.