madVR - high quality video renderer (GPU assisted)

[webs0r]

Just want to do it automatically... :)

I use a profile to use NGU Sharp (with oversampling) when
(scalingFactor.x < 2)
And NGU AA (with oversampling) in the other case.

Not exactly a proxy for what you're looking for but somewhat along those lines. I find the lower the source res is, the more I want NGU AA's approach.

[Bloodred]

Just installed the newest MadVr. From a pure image quality point of view, which offers better quality NGU Anti-Alias or NGU Sharp? (for both Chroma Upscaling and Image Doubling)
My GTX 1080 card is fast enough to run any of them at Very High, I just can't deside which one is superior.

There's no clear-cut answer, it depends on the source as well as your own preference. NGU Sharp tends to work very well with high-quality sources, not so well with lower quality stuff. NGU AA is better on low-quality stuff but the image it produces is significantly less sharp than what NGU Sharp manages. So have a look at both of them and just pick the one you like best.

In terms of performance they are both very similar with the one exception that NGU Sharp is faster if you're doing quadrupling, because it can do it directly in a single step rather than applying image doubling twice.

[Magik Mark]

Madshi,

Does NGU uses CPU or GPU?

[Georgel]

Madshi,

Does NGU uses CPU or GPU?

Mostly GPU power.

[huhn]

Thanks for the explaination, Pirlouy.

But they however contribute to the total amount of the power consumption by the GPU and in my case - due to a known issue of the AMD RX480 on some old Mobo (they absorb to much current from the PCI slot instead from the 8pin connector) - I discover that doing so I could save some power consumption and some extra heating of the GPU, preventing the PC from going into blank screen and freezing during the vision of some contents when the madVR settings are high.

your GPU is nearly for sure broken.

only the reference cooler RX 480 and the very bad powercolor red dragon draws "to much" power if it is over PCIe or not depends on the card but it doesn't matter what you draw over the PCIe connector anyway it is directly connected to the PSU so only a problem for the PSU. the mainboard doesn't care and doesn't know.

all other RX 480 do draw to much power from the PCIe slot they have 8 pin connectors for a totally power draw of 225 watt.
i know what i'm talking about i'm not a RX 480 user because hand shake issues with HDMI under load (black screens)

[HillieSan]

your GPU is nearly for sure broken.

only the reference cooler RX 480 and the very bad powercolor red dragon draws "to much" power if it is over PCIe or not depends on the card but it doesn't matter what you draw over the PCIe connector anyway it is directly connected to the PSU so only a problem for the PSU. the mainboard doesn't care and doesn't know.

all other RX 480 do draw to much power from the PCIe slot they have 8 pin connectors for a totally power draw of 225 watt.
i know what i'm talking about i'm not a RX 480 user because hand shake issues with HDMI under load (black screens)

The PCIe power issue is fixed with the driver. No hand shake problems here. Be sure your PSU is strong enough.

[huhn]

you know that a PC is shutting down if the PSU isn't keeping up (or the PSU dies).

even a proper 350 watt PSU can easily run RX 480 with a consumer grade intel CPU.

and the PCIe problem was never fixed even the PCIe spec has be revoked. adding a compatibly mode is not a fix and it still draws to much over the 6 pin connector. this was never a problem anyway.

[HillieSan]

you know that a PC is shutting down if the PSU isn't keeping up (or the PSU dies).

even a proper 350 watt PSU can easily run RX 480 with a consumer grade intel CPU.

and the PCIe problem was never fixed even the PCIe spec has be revoked. adding a compatibly mode is not a fix and it still draws to much over the 6 pin connector. this was never a problem anyway.

I understand. The driver keeps the power consumption within the specs of PCIe 3.0. I have not read any serious problems with the 6 pin connector of the RX 480, except that you need a good PSU. Same problem with nVidia.

A 350 watt PSU is low for the total system with an rx 480. My system had a i7 CPU with a 350 Watt PSU and a GTX 660. It failed on booting or it crashed in Windows. It was unstable. I bought a new 550 Watt PSU, which fixed the problem.

I use my RX 480 in another system with a 500 Watt PSU and an ASUS F2M85-M LE motherboard. I have no problems.

I don't buy a PSU under the 500 Watts anymore.

[ashlar42]

madshi, a very humble suggestion from me: I think we should take a page from Hydrogenaudio and set up blind tests for scalers comparisons. More often than not I look at screenshots comparisons and I like one over the other depending on the specific portion of image I look at. Taking away (unconscious) bias might be good to establish what's really better for people.
Please, nobody take offense for this suggestion. If you don't like it, fine. :)

[leeperry]

I'll be running more tests but so far I think SSIM 2D 100% LL has been quite a game changer because on 360*288 movies I now seem to prefer sxbr25+SR2 AR LL doubling, quad sxbr looks too sharp, NNEDI3 looks noisy and both NGU's look way too oil-painty and not meant for hopeless low-res upscales at all. On 720p@1080p NGU Sharp looks fantastic but I'm using Bicubic60 LL downscale coz my HD7850 can't afford SSIM at all. I see little use for NGU AA compared to sxbr75 but I'll keep A/B'ing them all then, all this said I would seem to agree that NGU Sharp renders NNEDI3 useless indeed, well done. I'll be sure to reconfirm my initial impressions ASAP after rolling more content.

[Patrik G]

i only want to say thanks for NGU with AA filter
now my 500M looks perfect!
no more artificial sharpness just a smooth clean picture that still is sharp!

i used NGU low before now i use NGU with VHQ AA
damn i looks good!

and alot better than the old NNEDI3 128 i used before NGU

GPU usage slight higher but who cares :)

so for Kuro users
NGU with AA filter gives the most accurate picture on these already sharp 1080P tvs.
use it

thanks Madshi
you naild it

[YGPMOLE]

you know that a PC is shutting down if the PSU isn't keeping up (or the PSU dies).

I'm quite sure that the PSU is not the problem: Seasonic 650W Gold. The CPU is an Intel I7 860. GPU is RX 480 Nitro 8GB. Mobo is Gigabyte GA-P55A-UD3.

Sorry, I forgot to mention that I'm using Smooth Video Project and I play everything at 60fps.

The PC doesn't shut down, it freeeze: first I got a blank video with 2 or 3 more seconds with the audio still playing, then it freeze and I have to force the reboot; after rebooting, the fans runs for some seconds at the max speed and then all goes back to normal (all this using "very high" both for luma / chroma / image upscaling).

Using "High" make all works, even if the fans runs fast and it's possible to ear them (really noisy, I'll say)

As HillieSan said, I too read that the problem had should be fixed by the driver, but it's not my case!

[LexSfX]

See here:

https://forum.doom9.org/showpost.php?p=1799028&postcount=42797


I only ever included it for "educational capacity", as you say. I don't think it's useful for anything else, even for pixart, unless you absolutely need to check every single pixel separately in pixart, but why you need that? Check out this one:

http://screenshotcomparison.com/comparison/202526

Why would you ever want nearest neighbor, if you can get the quality shown in that screenshot comparison, when using madVR?


I'm confused. You're linking to some algorithms which distort the pixels but you say you don't want distored-shaped pixels?

Why exactly do you need pixels to be exactly the way nearest neighbor produces them? It seems weird to me. I can understand that you don't want blurry, but you don't need to have blurry.


Huh?

http://screenshotcomparison.com/comparison/202526

To my eyes nearest neighbor is not sharper in that screenshot comparison. It's just dramatically more aliased.

No, sharpness is not hiding the pixels. The original pixels must be distinct from each other. Aliasing is good in this case. I don't want the pixels to be blended into each other, unless on a very small scale, such as in the algorithms I linked, which only blend a maximum of 1 screen pixel column or row on upscale, keeping the upscaled pixels intact and looking evenly-sized even at non-integer upscale values. Seeing the individual pixels helps to allow visual analysis of the game state, especially frame by frame. I'll just stick with this version of madVR then.

[HillieSan]

I'm quite sure that the PSU is not the problem: Seasonic 650W Gold. The CPU is an Intel I7 860. GPU is RX 480 Nitro 8GB. Mobo is Gigabyte GA-P55A-UD3.

Sorry, I forgot to mention that I'm using Smooth Video Project and I play everything at 60fps.

The PC doesn't shut down, it freeeze: first I got a blank video with 2 or 3 more seconds with the audio still playing, then it freeze and I have to force the reboot; after rebooting, the fans runs for some seconds at the max speed and then all goes back to normal (all this using "very high" both for luma / chroma / image upscaling).

Using "High" make all works, even if the fans runs fast and it's possible to ear them (really noisy, I'll say)

As HillieSan said, I too read that the problem had should be fixed by the driver, but it's not my case!

I experienced similar problems.
1. I use DirectX 11 setting in madvr. DirectX 9 is default and was never stable.
Enable
rendering -> general settings -> use Direct3D 11 for presentation (Windows 7 and newer)
rendering -> general settings -> present a frame for every VSync --below
Disable the rest
2. Disable all options in Quality for performance tab in madVR.
3. I set LAV 0.69 to DXVA native.
4. I moved away from SVP to Bluesky FRC. Bluesky FRC uses AMD's fluid motion, which is very stable. Now, I don't uses Bluessky FRC any more. I uses the exclusive mode and it automatically sets the original framerate of the movie (23 or 24 Hz). 60 Hz is a bit smoother, but 23/24 Hz is ok and it gives you more room to prevent frame drops at higher settings.
5. Use a good player, such as MPC HC, MPC BE or potplayer. Use the latest version.
6. Use the latest AMD Crimson driver.

after rebooting, the fans runs for some seconds at the max speed and then all goes back to normal

This is normal.

I hope this helps.

[huhn]

if the PC freezes you should find out what freezes event viewer is a good start for example.
it is very unlikely that it is madVR that is freezing.

if it is the GPU/GPU driver try to contact AMD per phone that usally works some how.

doing a workaround is not a good "fix".

[kitame]

madVR v0.91.6 released

3) When taking speed into account, how would you rate NGU Anti-Alias vs NNEDI3 at the various quality levels, for image upscaling?
4) Same as 2) and 3), but for chroma upscaling.

3) NGU-AA quad low is faster than NNEDI3 quad 16Neurons, quality difference is nearly negligible but in favor for NNEDI3.
NGU-AA quad medium is equal speed to NNEDI3 quad 16Neurons, quality difference is nearly negligible but in favor for NGU-AA.
i get occasional frame drops on either NGU-AA quad medium or NNEDI3 quad 16Neurons, not enough GPU performance?
GPU hardware = HD7950 @ 950Mhz.

4) which chroma upscaling? the one on a different tab or the one bundled in image upscaling?
i'm using ReconSharp on the chroma upscaling tab and it works great.
while on the image upscaling, the selection is locked when NGU-AA is set to low.

the combination of these barely works for me without causing frame drops.
chroma upscaling : ReconSharp
Image downscaling : SSIM 2D 100% LL + AR strict(soft)
Image upscaling : Luma Double = NGU-AA Low (set to "if needed")
Image upscaling : Luma Quad = NGU-AA Low (set to "if needed")
Image upscaling : chroma = N/A (disabled - defaults to Bicubic60 AR)
Image upscaling : upscaling = Jinc AR
Image upscaling : downscaling = use "image downscaling" settings

is there any reason for disabling chroma selection when NGU-AA is set to low?

also, how do i know whether quadrupling is working or not?

PS: i think you should rename chroma upscaling to chroma supersampling.

https://i.imgbox.com/AasRlpZ8.png

[Asmodian]

is there any reason for disabling chroma selection when NGU-AA is set to low?

If you can use a higher power chroma scaling you should use NGU-AA med instead, only if you are already using >low should you use a higher quality algorithm for chroma.

PS: i think you should rename chroma upscaling to chroma supersampling.

Maybe rename it to RGB conversion or something along those lines? Supersampling implies sampling from a higher resolution which is not what is happening in chroma upscaling. It also doesn't run if the video is YUV 4:4:4 or RGB, while the chroma options in image upscaling do.

[kitame]

If you can use a higher power chroma scaling you should use NGU-AA med instead, only if you are already using >low should you use a higher quality algorithm for chroma.



Maybe rename it to RGB conversion or something along those lines? Supersampling implies sampling from a higher resolution which is not what is happening in chroma upscaling. It also doesn't run if the video is YUV 4:4:4 or RGB, while the chroma options in image upscaling do.
good point, although i was thinking setting both luma and chroma to NGU-AA low to be superior to NGU-AA low + Bicubic60 AR.


RGB conversion doesn't sound right, how about chroma upsampling?

[XMonarchY]

NVidia just released their latest drivers a day ago and they mentioned introduction of a bunch of new OpenCL CUDA stuff, SDK, etc. I just wonder if any of those new things could help madVR in some way...

[Asmodian]

good point, although i was thinking setting both luma and chroma to NGU-AA low to be superior to NGU-AA low + Bicubic60 AR.


RGB conversion doesn't sound right, how about chroma upsampling?

Chroma scaling is harder than luma (2x the data to scale) so I am sure you could run NGU-med for luma if you could run NGU-low + NGU-low.

While chroma upsampling and chroma upscaling are synonymous I do like the idea of two different names for those settings, it would probably reduce confusion. :)

[Neo-XP]

Anyway. If I could remove spline, that would solve the problem, but in my experience, whenever I remove something, somebody will cry. You see, I *thought* I would be safe removing such a dramatically bad algorithm like nearest neighbor, but even for that some users are crying. So you can bet with spline there would also be some.

Remove Spline ? It is one of the best algorithm for upscaling... It is almost as sharp as Lanczos, but without the strong ringing artifacts.
Why is it not proposed anymore for this in new versions ? For those who can't afford Jinc, it is a valid choice.

Those are my new quality settings (v0.91.1) after years of comparing ;) :

artifact removal : reduce banding artifacts medium & high
chroma upscaling : super-xbr125 (super-xbr100 is good too for a more neutral image)
image downscaling : Bicubic150 + AR [relaxed]
luma doubling : NGU-med 2x supersampling
chroma doubling : NGU-low 2x supersampling
image upscaling : Jinc + AR
upscaling refinement : add grain 3
dithering : Error Diffusion 2 + colored noise + change dither

Any chance to add Jinc for chroma doubling (or with a "same as upscaling algo after doubling" option) ?
Upscaling algos after doubling are all good for chroma doubling. With only one option you can add them all.

:thanks:

[YGPMOLE]

I hope this helps.

All your suggestion are exactly my settings, DXVA excluded (I'm in software mode) - thank, however! I'm usng MPC-BE x86 + LAV + FFDShow + ReClock + madVR and SVP, all last version (driver included). I prefer 60fps, even at the cost of some artifacts.

@huhn

It's not madVR that is freezing, but the GPU due to the overheat that madVR induces when using "Very High" settings. Using high I'm OK, so I'll stay still this way.

[huhn]

your GPU is not supposed to over heat and it is pretty hard to over heat a RX 4X0 card.
and i never said madVR is freezing?

[jkauff]

It's not madVR that is freezing, but the GPU due to the overheat that madVR induces when using "Very High" settings. Using high I'm OK, so I'll stay still this way.
Are you using your GPU for SVP? The combination of SVP and madVR puts a lot of stress on the GPU. Try running SVP in CPU mode and see if it makes a difference.

[LexSfX]

Anyway, why remove any algorithm? If it's about GUI space, you can just add "removed" algorithms to an inconspicuous dropdown list which can grow indefinitely and put them in another source file so you don't have to look at them. The beauty of madVR is that it gives users many choices for how they want to render their videos.

[pirlouy]

You said you would stay on old version. Problem solved.

[Neo-XP]

You said you would stay on old version. Problem solved.

As I already said in the past, thanks for the "let madVR decide", your default settings make sense (I use "medium" quality in "image upscaling").
Too bad I still have to choose "chroma upscaling" and "image downscaling" settings. :)

I don't see the point of removing algorithms too, unless there are algorithms available that are better in every situation.
You want to choose the "let madVR decide" option for everything, and you will be able to. Problem solved for you.

I don't get why it bothers you (or anyone) if advanced users have the choice of using other algorithms :confused:
I don't think that madshi's goal is to have some users staying with old versions because of the limitations of the new ones.

The beauty of madVR is that it gives users many choices for how they want to render their videos.

+1

[hajosattila]

hi, Guys!

Can someone give me an idea of where to start as far as what the best settings would be for a blu-ray remux on a 1080p screen?

Hardware: GTX 1070 (OC) / i7-7700K 4.5GHz / 55" FHD TV
Source: Blu-ray REMUX or Blu-ray RIP (x264 mkv)
Current madVR settings:


Chroma: NGU-Sharp, very high quality
Downscaling: SSIM 1D 100% + AR + LL
Image: Jinc3 + AR
Upscaling refinement: Off
Artifact removal - Debanding: Medium/High
Artifact removal - Deringing: Off
Image enhancements: sharpen edges (1.0) + AR + AB 50% (I'm not sure it is necessary)
Dithering: Error Diffusion 2


sorry for my English...

:thanks: madshi!

[Asmodian]

unless you have lots of low quality blurays I would use artifact removal - debanding: low/high or low/low. I would also select Image enhancements for every source, some sources look better with sharpen edges, some worse. I don't use any image enhancements by default.

Are you using any trade quality for performance options?

Otherwise those look like good settings to me.

[Warner306]

hi, Guys!

Can someone give me an idea of where to start as far as what the best settings would be for a blu-ray remux on a 1080p screen?

Hardware: GTX 1070 (OC) / i7-7700K 4.5GHz / 55" FHD TV
Source: Blu-ray REMUX or Blu-ray RIP (x264 mkv)
Current madVR settings:


Chroma: NGU-Sharp, very high quality
Downscaling: SSIM 1D 100% + AR + LL
Image: Jinc3 + AR
Upscaling refinement: Off
Artifact removal - Debanding: Medium/High
Artifact removal - Deringing: Off
Image enhancements: sharpen edges (1.0) + AR + AB 50% (I'm not sure it is necessary)
Dithering: Error Diffusion 2


sorry for my English...

:thanks: madshi!

You might want to try crispen edges over sharpen edges. That's the only change I would make.

[Selur]

Small question when watching a 10bit clip and enabling the OSD I see:

Display 59.98126Hz
clock derivation 0.00189%
smotth motion on (settings)
D3D11 windowed (8bit)
P010, 10bit, 4:2:0
movie 25.000fps (says source filter)
movie 720x364,240:1 (35mm)
scale 0,0,720,364 -> 0,0,2559,1067
touch window from inside
chroma -> Nnedi16
image -> Jinc AR
vsync 16,67ms, frame 40.00ms
matrix BT.601 (says upstream)
primaries EBU/PAL (best guess)
limited range (says upstream)
deinterlacing off (settings)
decoder queue 16-16/16
upload queue 8-8/8
render queue 8-8/8
present queue 1-1/8
dropped frames 0
delayed frames 0
presentation glitches 2
average stats
rendering 4.87ms
present 0.06ms
max stats (5s)
rendering 0.00ms
present 0.09ms


Seeing "D3D11 windowed (8bit)" I'm wondering: So is the video presented as 8bit or 10bit?
(In case only 8bit output is used, what do I need to do to get get 10bit ?)

Cu Selur

Ps.: 'devices->HP Z27q->properties->the native display depth is' is set to '10bit (or higher)'; the display configuration inside the NVIDIA control panel is set to 10bit output; inside LAV Video Decoder all output formats are ticked.

[sneaker_ger]

Yes, madvr dithers down to 8 bit. Did you try fullscreen exclusive mode?

[Warner306]

What scaling factor enables image doubling when set to "Let madVR decide?" 1.2x and 3x??

Also, what happens when luma quadrupling is set to "Let madVR decide?" Direct quadruple??

[Damien147]

MPC HC 1.7.11

Changelog

LAV Video Decoder: Enable hardware decoding for HEVC and 4K/UHD by default


Every RX 4x0 user is gonna have a black screen by default with 10 bit HEVC.

[sneaker_ger]

Only those two checkboxes are ticked now but the hardware decoding drop-down list is still set to "none" by default. (At least for external LAV)

[Damien147]

For internal default is DXVA2(native) and that gives black screen with 10 bit HEVC and AMD RX 4x0.

[Neo-XP]

@madshi In v0.91.7, for chroma upscaling, only Sharp and Anti-Alias NGU can be selected. Other NGU variants (Soft and Standard) can't be.

[nevcairiel]

@madshi In v0.91.7, for chroma upscaling, only Sharp and Anti-Alias NGU can be selected. Other NGU variants (Soft and Standard) can't be.

Soft and Standard are not implemented yet, as was explained in the release notes. :)

[hajosattila]

unless you have lots of low quality blurays I would use artifact removal - debanding: low/high or low/low

Do You suggest that if I only watch quality footage (REMUX) or 1080p WEB-RIP TV Shows I should set them to low/low?

Are you using any trade quality for performance options?

No

http://i.imgur.com/fzsu59F.jpg (http://imgur.com/fzsu59F)

You might want to try crispen edges over sharpen edges

I will try it, thanks!

[kitame]

crispen edges amplifies grain and noise, so take note of that.

[Neo-XP]

2) When disregarding speed, how would you rate NGU Anti-Alias Very High vs. NNEDI3 256-taps, for image upscaling?
3) When taking speed into account, how would you rate NGU Anti-Alias vs NNEDI3 at the various quality levels, for image upscaling?
4) Same as 2) and 3), but for chroma upscaling.

4) NNEDI3 256-taps is better than NGU Anti-Alias Very High for chroma upscaling.
For my use, NNEDI3 64-taps (I can't do 128-taps with my graphic card) looks better than any NGU Anti-Alias.
However, NGU Anti-Alias is a very big improvement over NGU (Sharp) for chroma upscaling, but it is not quite as good as NNEDI3 yet.
Anyway, I use Jinc for chroma upscaling now, because the image is visually similar to NNEDI3 64-taps, with a lot less of resources needed (GPU and CPU).

Original :
https://s17.postimg.org/m4lsim6d7/original.png (https://postimg.org/image/m4lsim6d7/)

NNEDI 64 / 256:
https://s17.postimg.org/h3eeh8ywr/nnedi64.png (https://postimg.org/image/h3eeh8ywr/) https://s17.postimg.org/6ulxbfauz/nnedi256.png (https://postimg.org/image/6ulxbfauz/)

NGU NGU Anti-Alias low / med / high / very high :
https://s17.postimg.org/7ewd0depn/NGU_low.png (https://postimg.org/image/7ewd0depn/) https://s17.postimg.org/lzi18xz23/NGU_med.png (https://postimg.org/image/lzi18xz23/) https://s17.postimg.org/7qdtd4v5n/NGU_high.png (https://postimg.org/image/7qdtd4v5n/) https://s17.postimg.org/mqareq1ff/NGU_veryhigh.png (https://postimg.org/image/mqareq1ff/)

Also, I am still not convinced with NGU Sharp High in v0.91.7 vs NGU-med in v0.91.1 when upscaling FHD to 4K.

Here is another comparison I have made :

Original : http://www.mediafire.com/file/bo44396go9301l0/original3.png
v0.91.1 : http://www.mediafire.com/file/z09geg7cov5exid/old3.png
v0.91.7 : http://www.mediafire.com/file/0069oc381y59pch/new3.png

You can clearly see the differences on the watch and the ring of the man in the left side, and in the eyes of the three.
I hope this help you to improve NGU for next versions.

[x7007]

unless you have lots of low quality blurays I would use artifact removal - debanding: low/high or low/low. I would also select Image enhancements for every source, some sources look better with sharpen edges, some worse. I don't use any image enhancements by default.

Are you using any trade quality for performance options?

Otherwise those look like good settings to me.

So for Remux and high quality 1080p should I Low/Low or Low/High ?

[cork_OS]

I really don't understand how anybody could want nearest neighbor when looking at these results:

http://screenshotcomparison.com/comparison/202526
Yes, very impressive comparison. But you've used superres with NGU AA. So, SuperRes can improve NGA AA output? Or it's just a workaround to subpixel shift?

[galneon]

Having a fullscreen exclusive mode issue after updating to MPC-HC 1.7.11. In the past, when I'd right click, there would be a pause and the context menu would come up, then after I was done with the menu, exclusive would re-engage. Now, the screen goes black for ~6 seconds and returns to fullscreen exclusive without displaying a context menu. I think what is happening is it's switching out, then immediately switching back into exclusive without the menu ever showing. Workaround is to only access the context menu from windowed mode. This is a little annoying since I've already experienced an increased delay in fullscreen exclusive switching when I went from Win7 to 10 (I know, that's MS's fault).

MPC-HC 1.7.11, madvr 0.91.7, Windows 10 x64, GTX 970.

Anyone else?

[Asmodian]

So for Remux and high quality 1080p should I Low/Low or Low/High ?

Yes, debanding med and high do remove details. Even good blurays often have banding during fades so high is a reasonable choice for fades but I use debanding low/low myself, unless the source has obvious banding.

This is a little annoying since I've already experienced an increased delay in fullscreen exclusive switching when I went from Win7 to 10 (I know, that's MS's fault).

MPC-HC 1.7.11, madvr 0.91.7, Windows 10 x64, GTX 970.

Anyone else?

Do you need FSE? Windowed DX11 or Windowed Overlay work better for me with Windows 10.

[Warner306]

Yes, debanding med and high do remove details. Even good blurays often have banding during fades so high is a reasonable choice for fades but I use debanding low/low myself, unless the source has obvious banding.



Do you need FSE? Windowed DX11 or Windowed Overlay work better for me with Windows 10.

To me, all 8-bit sources have some banding. Especially if that source has been compressed.

[galneon]

Do you need FSE? Windowed DX11 or Windowed Overlay work better for me with Windows 10.

I suppose not... I don't output 10-bit after all. It's hard for me to disable exclusive as I'm the guy who complains when a new game only offers borderless fullscreen, but forced triple buffering is the issue there.

[leeperry]

Artifact removal - Deringing: Off

Aww man, put the deringer on already, it's one of the most mind-blowing features I've seen in mVR to the point that I'm wondering how I could watch anything before this popped up http://forum-images.hardware.fr/images/perso/1/anefay.gif

[Bloodred]

So for Remux and high quality 1080p should I Low/Low or Low/High ?

I find that medium or high debanding sometime smooths out too much and can destroy subtle gradients, as in it turns them into a large blob of a single color and removes some detail. I wouldn't (and don't) use them in any general purpose profile and would only turn them on with specific sources where the banding is really bad. That really shouldn't be the case with 1080p BD remux files, at least I personally haven't seen any yet that suffered from terrible banding. That's more something I associate with low-quality SD stuff.

[70MM]

Im upscaling BDs to 4K with madvr and want to use a small amount of sharpening since I blow the image up to a large 145" scope dia screen.
Where is the best place to use sharpening, under Processing > Image enhancements or in Upscaling refinement?

Also is it advisable to use AR, does it add any softness or artefacts?
If so where is the best place to add it, Image enhancements or upscaling refinement?
As there are a number of sections where AR is, do you add it more than once or once only from one section?