madVR - high quality video renderer (GPU assisted)

[chros]

I think it's realeted to resizing. E. g. play a 2160p video on 1080p screen:
- normally chroma will be upscaled to 2160 and then both to 1080
- if it's enabled then there won't be chroma doubling

[Asmodian]

Is this really lower quality? And why or in what way would it be lower quality when enabling the option? Why shouldn't we always have that option enabled?

You do get different chroma... whether or not it is worse is more debatable. In testing it does seem better to handle chroma the same as luma but it is one of the least loss of quality options in trade quality for performance.

[Wilmar]

The screenshot shows "HDR: unknown properties", meaning it can't find the HDR metadata.

Use LAV Video decoder. Disable the internal video decoder and video processor in Potplayer.

Thank you for your answer. But I already use LAV Video decoder:
https://i.imgur.com/HZOJZpJ.png
https://i.imgur.com/1MmvsgE.png
My exported Potplayer settings are here (https://drive.google.com/open?id=16DcL4IOoCvSvJJk25XbXRbCACFaXSZgo).

[IngramAU]

Having problems with repeated frames on 24.000fps content. I've nuked my drivers and reset Madvr to default still seeing the same issue (1060GTX.)

Please see attached images. I am not able to access 24.000hz fine tuning. 23.976hz I can.

I nuked my drivers and tried 24 again before fine tuning 23.976 incase it was causing issues didn't make a difference.

[hannes69]

Please see attached images. I am not able to access 24.000hz fine tuning. 23.976hz I can.
Attached images in this forum can take a long time until we can see them, so it would be better to upload them to an image hoster.
24.000Hz and 23.976Hz sometimes 'interfere' in certain setups/drivers.
A possible workaround could be to use multiples of these values (so you´ll have e.g. 23.976Hz and 48.000Hz or 47.952Hz and 24.000Hz or 72.000Hz...
Of course this depends on your display device´s capabilities, so your monitor/tv/beamer has to accept these refresh rates.

[tp4tissue]

Hi,this is rough representation of what I am seeing with RX 5700 bug, beginning of bladerunner 2049

https://1drv.ms/u/s!AgvFafeelEBij75edbautjSVhll3yw?e=nM5M1Y

Use Madvr's tonemapping, The Tv's tonemapping is prolly poop anyway.

:devil:

[tp4tissue]

I have now a full 10-bit path madVR-GPU(gtx 1060)-Display(4k monitor). Is there any disadvantage which I have to pay attention to ?

You need to test it with 10bit/8bit gradient test files. See which one is smoother.

But since 444 60hz only works in 8bit, and autoswitch 23/24hz is hit or miss depending on driver, I've stuck with 8bit.

I've not seen 10bit input produce noticably smoother gradient either because madvr's dithering is great.

[tp4tissue]

the current in progress version killed 1080 by just tonemapping with a version to saying it will be released like that but still.


:scared: is this true ? we all have to buy 1080tis now ?

[tp4tissue]

In short, I'll stay to 8 bit.

Since you own a colorimeter, you can test both and see if your accuracy increases/ decreases/ contrast/ gradient smoothness etc.

[nsnhd]

You need to test it with 10bit/8bit gradient test files. See which one is smoother.

But since 444 60hz only works in 8bit, and autoswitch 23/24hz is hit or miss depending on driver, I've stuck with 8bit.

I've not seen 10bit input produce noticably smoother gradient either because madvr's dithering is great.
By switching madVR's dithering off I can see 10bit gradient files smoother by 10bit path. Hope that some browsers can take advantage of it when playing videos, but I'm not sure.
You have to connect via Displayport to have 444/60Hz 10bit and some newer driver versions don't like it.

[huhn]

yeah and what else should happen if you disable dithering?

[nsnhd]

I haven't noticed anything else except 10bit path playing gradient files smother than 8bit path. I just did that for testing only, and always enable dithering in madVR. I have to find out how to test videos such as football matches in browsers to see if 10bit path might have any effect.

[huhn]

that's not a "test" how could there be another result.

you have to test if 8 bit dithering or 10 bit dithering is smoother.
yuor test is comparable to taking a shower and figuring out if you get wet or not.

[nsnhd]

Yeah, it's not a test, just what I can see. I don't know how to test browsers such as Edge or Chrome which don't tell about bit depth and dithering options/settings.

[huhn]

if the web browser doesn't dither(they don't) 10 bit rendering would be always better but it's not that easy to render 10 bit so they generally don't do that.

10 bit rendering with dithering on a screen that doesn't support it will always win against 8 bit none dithering. so what is the point of looking at 8 bit none dithered ever why should 10 bit none dithered not always be better? i mean the screen doesn't even matter in this case.

[IngramAU]

Managed to get MadVR to accept a 24hz custom mode by going for 23.999xxx hopefully after a few runs I can get it closer to 24. Not sure why I can't start off from my EDID it just won't accept it with an unknown error.

[mclingo]

Use Madvr's tonemapping, The Tv's tonemapping is prolly poop anyway.

:devil:


There is a well documented bug with AMD HDR API which causes this, its nothing to do with tone mapping, AMD's private API isnt sending BT2020, we have to use windows HDR API for now.

I'm already using MADVR's tone mapping, however I have to turn on WINDOWS HDR first which isnt ideal but there is no fix in sight for this bug yet.

[Asmodian]

10 bit rendering with dithering on a screen that doesn't support it will always win against 8 bit none dithering. so what is the point of looking at 8 bit none dithered ever why should 10 bit none dithered not always be better? i mean the screen doesn't even matter in this case.

I wanted to expand on this because it is an interesting and more subtle point than I thought.

With dithering disabled you are not reality testing your 8 vs 10 bit paths. 10 bit will always look smoother even if, with dithering, 8 bit would look smoother.

[VBB]

If you're unsure whether your display is 8-bit or 10-bit, there really isn't a test to determine that 100% anyway. But if you know the bit depth, you can certainly test for banding in different desktop and TV modes. And for that, I would recommend disabling dithering in madVR. You still can't be sure that the video card itself doesn't do any dithering, but it's a start. Just make sure to turn it back on when you're done testing.

[mrmojo666]

@Mclingo, I'm very curious to follow this bt2020 bug, could you please lnk something regardig the documentation you are referring to?

[Asmodian]

But if you know the bit depth, you can certainly test for banding in different desktop and TV modes. And for that, I would recommend disabling dithering in madVR.

But if you disable dithering what do the results of the test tell you? 10 bit will always be better than 8 bit with dithering disabled, even if 8 bit would be better with it enabled. You have to do your testing with the same settings you actually use when watching, with dithering enabled.

[mclingo]

@Mclingo, I'm very curious to follow this bt2020 bug, could you please lnk something regardig the documentation you are referring to?

When I say documented i'm mean the lesser meaning of this, i.e written down, there are no afficial documents detailing this, however the is a bug logged on MADVR here:

http://bugs.madshi.net/view.php?id=630

We moved the discussion about this to the driver thread as we had a complaint about spamming this thread:

https://forum.doom9.org/showthread.php?t=176013&page=50

This issue has dominated the last few pages of this thread, its now pretty clear the API is causing it, we're just waiting for AMD to fix it, here is DMU explanation of the issue:

You have entered support for HDR mode. The agsSetDisplayMode() function used to set a specific display in HDR mode, does its job perfectly: it sends metadata to the display device, which is defined in section 6.9 «Dynamic Range and Mastering InfoFrame» according to Table 5 of the CTA-861 standard. But in the same Table 5 there is also «Auxiliary Video Information (AVI)» defined in section 6.4. And all display devices are required to use the color space (colorimetry) from this data section (AVI InfoFrame) for the current video signal.
Suppose we are in SDR mode with the standard sRGB color space. And we want to switch to the HDR mode with the BT.2020 color space, which is the main one for this mode. By calling the agsSetDisplayMode() function, we put the display device in HDR mode. And we see distorted or unsaturated colors. This is because the display device did not receive the corresponding flag from the GPU in the AVI InfoFrame and is trying to display our BT.2020 color space in its sRGB.
Please tell me, do you think that such HDR support in AGS_SDK is sufficient? If yes, then advise what else needs to be done so that the display device passes into the correct color space when activating the HDR mode using AGS?

The outcome, MADVR is not sending BT2020 to your display so your colours are very unsaturated. Best workaround is to turn on windows HDR before starting your movie player.


*ONLY AFFECTS NAVI BASED CARDS

[VBB]

But if you disable dithering what do the results of the test tell you? 10 bit will always be better than 8 bit with dithering disabled, even if 8 bit would be better with it enabled. You have to do your testing with the same settings you actually use when watching, with dithering enabled.

You're not testing for bit depth, though. You're trying to find the best combo in regards to banding, and from what I've seen, the best combo is the best with dithering on and off. I guess what I'm trying to say is that I find 4:2:2 10-bit without dithering to be smoother even compared to 4:4:4 or RGB with dithering on. Hope that makes sense.

...and not to confuse anyone here: the above is meant for LG OLEDs only.

[Asmodian]

But you would come to exactly the same conclusion if you left dithering enabled for all your testing. I assume 10 bit 4:2:2 with dithering is still smoother than 4:4:4 or RGB with dithering, if it is smoother without dithering. :p

Aren't we testing bit depth? 8 v.s. 10?

I am arguing against testing without dithering, it never tells you anything useful about your video path.

Edit: In my testing, on a C9 in PC mode, 10 bit without dithering is better than 8 bit without dithering (for banding) but 8 bit with dithering is better than 10 bit with dithering.

[DMU]

You do get different chroma... whether or not it is worse is more debatable. In testing it does seem better to handle chroma the same as luma but it is one of the least loss of quality options in trade quality for performance.

I was also very interested in this question, so I did some tests.
Pic 1 (https://drive.google.com/open?id=15YRyjkCM-ciXo9BUIErdXDgrVm4bi0qO) - "scale chroma separately, if it saves performance" - OFF
Pic 2 (https://drive.google.com/open?id=10Yneo04JxYs-gMQq4S7JIJp8eDNQNB9M) - "scale chroma separately, if it saves performance" - ON
As you can see, double conversion of chroma noticeably degrades image quality. Correct me, please, if I did something wrong.

[VBB]

But you would come to exactly the same conclusion if you left dithering enabled for all your testing. I assume 10 bit 4:2:2 with dithering is still smoother than 4:4:4 or RGB with dithering, if it is smoother without dithering. :p

Aren't we testing bit depth? 8 v.s. 10?

I am arguing against testing without dithering, it never tells you anything useful about your video path.

Edit: In my testing, on a C9 in PC mode, 10 bit without dithering is better than 8 bit without dithering (for banding) but 8 bit with dithering is better than 10 bit with dithering.

For me it's more about seeing the display's raw performance and what works best with its built-in processing. My final testing is always with dithering on. PC mode is special ;)

[Asmodian]

That is the thing though, this myth about "seeing the displays raw performance". 10 bit will always be better. You are not testing your displays raw performance, you are noticing that 10 bit has more steps than 8 bit.

[VBB]

But I'm not even bringing up 10-bit. You did :) I wholeheartedly agree that madVR's 8-bit dithering is indistinguishable from 10-bit. It was merely a coincidence that I found the best mode for me happens to be 10-bit. I would have happily picked 8-bit if it was better in this case.

[Asmodian]

This was a discussion in response to nsnhd about turning off dithering and noticing that 10 bit was smoother than 8 bit and thinking that it meant something. :)

Edit: You decided YCbCr 10 bit 422 looked better than YCbCr 8 bit 422 with dithering disabled or enabled? There is an implicit choice between 8 and 10 bit in your preferred mode.

[mrmojo666]

@mclingo, thank you for the explanation, I'm going to read links

[VBB]

This was a discussion in response to nsnhd about turning off dithering and noticing that 10 bit was smoother than 8 bit and thinking that it meant something. :)

Edit: You decided YCbCr 10 bit 422 looked better than YCbCr 8 bit 422 with dithering disabled or enabled? There is an implicit choice between 8 and 10 bit in your preferred mode.

That's why I said in response to nsnhd's post that there is no good way to test for bit depth. Instead, you can test for banding. And again, my final choice for using 10-bit wasn't based purely on bit depth, but with 4:2:2, 10-bit is banding-free while 8-bit is far from it. I did not find the same to be true using any other combo. 8-bit always wins (compared to 12-bit, when there is a choice).

Maybe I should rephrase: When testing for banding, whether you do it with or without dithering, your final choice should always be based on what looks best the way you actually watch it: with dithering enabled.

[Asmodian]

Ok, totally agree. :)

But I still think testing with madVR's and/or the GPU's dithering disabled is pointless, except academically.

[nsnhd]

After your discussions I've redone the testing, from now on madVR's dithering is always enabled (Ordered Dithering).
When GPU output is set to 10bit RGB Full in NCP, no difference between 8bit/10bit outputs from madVR FSW can be seen by playing gradient files.
But, when GPU is set to 8bit and of course madVR output FSW also changed to 8bit, I can now see more banding compared to the above 10bit GPU setting.

So, what can I know about my monitor now, or still nothing ?

[huhn]

if the test is done correctly this would mean your screen is incorrectly handling 8 bit input so it will add additional banding for pretty much every application.

the current stable version of madVR shouldn't stop you from sending 10 bit even when the GPU is set to 8 bit output actually there is quite a good reason to still do that but i guess madshi wants to work around an nvidia driver bug i haven't checked in a log time.

you should use full screen exclusive for this test and not windowed full screen because WFS is know to produce a lot of banding when the GPU outputs 8 bit but get's a 10 bit input.

[glc650]

I've noticed with 23.976 HDR content played back at 23.976 sections of the screen are extra blurry (almost like they are pulsating/flickering) when panning (eg. fence or branches/leaves on a tree in background). If I play back at 59.940 (with smooth motion off - with it on the issue is still there) this almost goes away.

If I turn off HDR (tone map hdr using pixel shaders), this issue is completely gone at any frame rate and with or without smooth motion.

This happens even after restoring default madvr settings.

[chros]

If you're unsure whether your display is 8-bit or 10-bit, there really isn't a test to determine that 100% anyway. But if you know the bit depth, you can certainly test for banding in different desktop and TV modes. And for that, I would recommend disabling dithering in madVR. You still can't be sure that the video card itself doesn't do any dithering, but it's a start. Just make sure to turn it back on when you're done testing.

I understand what @VBB does and why, and I agree with him.

[jasonwc18]

I have my JVC RS600 calibrated for both BT.709 and BT.2020. In the past, I've accomplished this by using 3DLUTs for both color spaces, and madvr will automatically select the correct 3DLUT based on the video metadata. However, I upgraded to a Stewart ST130 G4 screen and the manual calibration result was so good, there's no real need for a 3DLUT. If I select "Disable calibration controls for this display", will that allow me to switch between BT.2020 and BT.709 content by just selecting the appropriate mode on my PJ? I thought that mode was equivalent to BT.709, Gamma Power 2.2.

[huhn]

no it will not.

in sdr output "Disable calibration controls for this display" is pretty much bt 709 gamma 2.2.

[Manni]

I have my JVC RS600 calibrated for both BT.709 and BT.2020. In the past, I've accomplished this by using 3DLUTs for both color spaces, and madvr will automatically select the correct 3DLUT based on the video metadata. However, I upgraded to a Stewart ST130 G4 screen and the manual calibration result was so good, there's no real need for a 3DLUT. If I select "Disable calibration controls for this display", will that allow me to switch between BT.2020 and BT.709 content by just selecting the appropriate mode on my PJ? I thought that mode was equivalent to BT.709, Gamma Power 2.2.

It depends if you're using the same calibration for both, and which device you're using.

Unfortunately the rs600 on its own is unable to detect SDR BT2020 vs rec-709 and switch automatically between a rec-709 and a BT2020 calibration. It's only able to detect HDR10 and apply its awful gamma D curve, which you definitely don't want.

So you have three ways to use madVR to automatically display the content correctly:

1) You're using an SDR BT2020 calibration for all content. This means that unless you need your iris fully open even for SDR content, you are most likely sacrificing black floor / contrast in SDR rec-709 to privilege brightness in HDR, running all with the iris fully open and using the P3 filter to get a wide gamut. In that case, assuming your two 3D LUTs use this same BT2020 baseline, madVR can switch between each LUT automatically. If you don't want to use the LUT anymore, simply specify "this display is already calibrated to"in the calibration tab, and specify SDR BT2020 and whichever gamma you've used in your calibration (usually 2.4, but it can be anything, as long as your target for the 3D LUT is 2.2 if you're using one for HDR tonemapping, though in that case it's recommended to target P3 rather than BT2020 to avoid posterization). madVR will automatically convert rec-709 content so that it displays correctly with your BT2020 calibration. There is no need for profiles.

2) You're using an SDR Rec709 calibration for all content. This means that you are most likey sacrificing brightness and gamut cover in HDR in order to maximise black floor / native contrast, not using the filter (unless you can't reach 709 wihtout it) and closing the iris further to reach 50/60nits in sdr rec-709 (or brighter if you don't have a dedicated room). If you don't need the 3D LUTs anymore, you simply specify "this display is already calibrated to" and you specify rec-709 and whichever gamma the display is actually calibrated. MadVR will automatically tonemap BT2020 content to fit into your rec-709 calibration. There is no need for profiles.

3) You don't want to compromise either content type, so you have two calibrations on the JVC (rec-709, no filter, iris closed to get best black floor / native contrast and around 50-60nits peak) and SDR BT2020 (2020, P3 filter, iris open to get the highest peak brightness, hence headroom for HDR highlights, at the cost of a higher black floor / reduced native contrast). You will need profiles to tell madVR which calibration you're using for each content type. If you're not using 3D LUTs, you can specify "this display is already calibrated" to either rec-709 or BT2020 using a different profile for each content. In that case, you can either switch manually between the two calibrations on the JVC, or if you have an nVidia GPU and a Vertex/Maestro/Diva you can select the "report BT2020" box in the calibration tab for your BT2020 profile, and make sure it's unchecked in your rec-709 profile. The HD fury device will detect if the content is rec-709 or BT2020 according to the flag, and provided you have installed an rs-232 cable and programmed the device accordingly, it will select the correct calibration automatically, according to content. MadVR will select the correct profile according to content, and will know which calibration you're using thanks to each profile. See here for more info: https://www.avsforum.com/forum/24-digital-hi-end-projectors-3-000-usd-msrp/2222586-official-jvc-rs600-rs500-x950r-x750r-x9000-x7000-owners-thread-931.html#post55408090

There are other ways to achieve this using batch files instead of an HD Fury device but they are even more complicated and don't work reliably in my experience.

[Wilmar]

Hi everyone!
I asked a question about not working HDR some time ago but it seems to got buried under the discussion above, so I decided to up it :)
Here’s the issue:
I can’t make HDR work properly with madVR and PotPlayer + 4K HDR-capable TV.
I enabled HDR in madVR settings – screenshot (https://i.imgur.com/0j55G89.png).
My MadVR settings config is here (https://drive.google.com/open?id=1_ElvCWSDIMAIyPlnN328od9qza6b_mFr).
The TV says it’s getting HDR image. However, the colours don’t look HDR at all. Moreover, when played via the TV directly, the video looks properly. But playing it via the PC results in faded, dull and dark image.
I tried enabling and disabling Windows HDR – to no avail.
Here’s what PotPlayer says about the video with Windows HDR enabled and disabled respectively.
No Windows HDR (https://imgur.com/z0sOTXr)
Windows HDR (https://imgur.com/CYCsUZo)

My configuration:
Windows 10 64 bit 1903
Nvidia Geforce 1080Ti (Driver ver. 441.87)
MadVR ver. 0.92.17
PotPlayer 1.7.18346

clsid suggested that I disable the internal video decoder and video processor in Potplayer and use LAV Video decoder. However, I already use it as seen on the screenshots:
https://i.imgur.com/HZOJZpJ.png
https://i.imgur.com/1MmvsgE.png
My exported Potplayer settings are here (https://drive.google.com/open?id=16DcL4IOoCvSvJJk25XbXRbCACFaXSZgo).

Am I doing something wrong?
Thanks.

[mclingo]

Hi in beta 113 tone mapping what is the different between peak luminance and Dynamic target nits, just wondering what the best settings should be for my 400 nit OLED for both of these to maximise bright small objects in HDR movies, I know they are not going to be super bright but when I recently watched the EXPANSE in HDR on Amazon it looked way better than any movie i've watched through MADVR. I know I cant do an A-B test for this material but I'd like to know what the best settings should be.

[SamuriHL]

Your real display nits should be set to 400. Dynamic target is NOT what people seem to be thinking it is. I would start at 75 and if you find that too dark/not to your liking, drop it to 50. I have mine set to like 150 or something crazy right now. It's personal taste on that particular setting...

[mclingo]

thanks.

I tried 50, 75, 100, 150 and 400, i genuinly could see no difference. you said if 75 is too dark drop it to 50, does this mean the lower the value the bright the highlights, what exactly is this doing?

edit: ok i see it now, I had to go to either extreme to really see the difference though, I compared 0 to 500, 500 being much duller more like SDR.

if you have an OLED, surely you want the brightest image possible, surely you'd set this at 0?

edit 2: ok so the lower it is the more highligh detail you loose, theres a balance you have to find, i settled on 25 believe it or not :)

[SamuriHL]

Yea, exactly what you said, you lose detail so it's a matter of preference to find the balance you're looking for. Fun game isn't it? :D

[mclingo]

Indeed, I spent last nigt runing through some 4k demos and movies with my new settings, never looked so good, to think i'd been languishing with dim HDR performance for so long thinking my TV was pants :), this doesnt say much for LG's own tone mapping though as MADVR now looks loads better the passthrough HDR, well happy with the results. :)

[SamuriHL]

LG's DTM is an acquired taste. One that sadly I never acquired. LOL I VASTLY prefer madvr's tone mapping to LG's. We can't get rid of LG's entirely on our OLED's, but, we can limit its impact by tone mapping to 700 nits or less (or in your case 400). The work that's been done is tremendous.

[x7007]

Anyone with AMD 5700XT with working HDR? in Movie Games and Youtube/Chrome 4K HDR? I have the 1080GTX now and there are no issues. but will I will change gpu for FPS improvement.

Because someone said there is an issue with 5700XT HDR and VEGA IIV or something works fine. I am waiting for the big Navi, I hope there won't be issues.

[WuNgUn]

Big Navi isn't expected anytime soon...? I thought like summertime?

[mclingo]

[QUOTE=x7007;1897782]Anyone with AMD 5700XT with working HDR? in Movie Games and Youtube/Chrome 4K HDR? I have the 1080GTX now and there are no issues. but will I will change gpu for FPS improvement.

Because someone said there is an issue with 5700XT HDR and VEGA IIV or something works fine. I am waiting for the big Navi, I hope there won't be issues.[/HDRQUOTE]

HDR shouldnt be working properly for anyone in that its not sending BT2020 and saturation is low if the app uses AMD API. Workaound is to turn on windows HDR first, this seems to solve the issue for MADVR playback but I havent tried this with games as nobody has been able to give me an example of a game that uses AMD's Private API for me to test.

I cant see big navi not having this issue but I guess its 50/50, if you dont play 3D movies i'd consider an NVIDIA card right now.

[huhn]

maybe just wait for the GPU. it's not even clear if it is a consumer card or a professional card or if we will even see one at all.

the 5700 XT is eat 225 watts a bigger card would be insane with the same node.