madVR - high quality video renderer (GPU assisted)

[Knight77]

Hi guys, just some quick questions:

1) in this page... what did you select?

http://i.imgur.com/yAUrqVY.jpg

I understood that someone is suggesting to use specific settings (not letting the application decide) and change the power management mode. Is it necessary?

BTW I do not have a mobile, I've got a desktop with a GTX1070.

2) About diethering: 10bit 4K TV, GTX1070 running things, I should pick Ordered Diethering without "use colored noise"....right? I'm a bit confused because I was sure that the best quality settings was ED1 but then in the last pages of the thread I read that OD is like the best choice for my set-up.

3) Image downscaling: am I wrong or it's like I'll never use it? I mean I upscale 720p or 1080p to the 4K and I'll never watch anything over 4K....

Thanks

[JarrettH]

NGU doubling is really impressive with 720p animation @ 1080p. After playing back an episode of Family Guy in 720p and then 1080p version, it was hard to tell the difference. It's easy to tell the difference with Jinc and NNEDI3, 720p version looks blurry. Switch to NGU and bam, back in focus. The only thing that stands out is aliasing on some diagonal lines with NGU, otherwise they look almost identical.


I'm not stopping anyone from doing anything. The differences to me are not noticeable in 99% of real world circumstances. All these comparisons of red text are not applicable to what you see watching in the majority of a video. I still haven't come across anything where I paused a video, changed chroma, then saw any difference at all. With cartoons, movies, sports or anything else.

But, if the only way to tell the difference is with these extreme fringe examples, then more power to them. That will be my last comment on the matter. It really doesn't matter to me what other people use. I'm just glad something like madVR exists for us videophiles. :)

And I don't watch cartoons apart from the odd Family Guy. I get what you mean. I've noticed a lot of examples on this board aren't real life content.

[Asmodian]

2) About diethering: 10bit 4K TV, GTX1070 running things, I should pick Ordered Diethering without "use colored noise"....right? I'm a bit confused because I was sure that the best quality settings was ED1 but then in the last pages of the thread I read that OD is like the best choice for my set-up.

ED1 is higher quality than ordered. However, with a 10-bit 4K TV you are even less likely to notice the improvement over ordered dithering. I really do not like "use colored noise" but others do so that is a matter of preference.

If you have to lower other settings ED1 probably isn't worth it, but if you don't then ED1 is never bad in itself.

3) Image downscaling: am I wrong or it's like I'll never use it? I mean I upscale 720p or 1080p to the 4K and I'll never watch anything over 4K....

You are correct, there are two downscaling options now and you will never need to use the one in "image downscaling".

There is the downscaling option for "algorithm to use after doubling:" on the image upscaling page. This would be used after quadrupling 720p to 5120x2880 to downscale to 3840x2160. I suggest leaving it on "let madVR decide". :)

[ryrynz]

I don't see a mod title next to your name. I'll do whatever I like. :thanks:


You fail to get my point and "witty" remarks do nothing to change it.


I'm not going to waste gpu/electricity on something that might pop up a couple of times a month for a few seconds. You can choose whatever algorithm you like, I'll do the same.

I guess you won't "waste" money on a screen that shows these differences quite clearly either. If you have a low quality screen then I guess it makes sense to use the lower quality algorithms.
Just because you can't see differences in your ancient setup doesn't mean they don't exist.

[fedpul]

Please guys, we need to calm down. The aggressiveness has been increasing in the last few days. We do not know the other economical situation, I think we all have the right to share our opinion and that we can do it without being rude and avoiding offensive comments. We are all trying to do madVR better day after day, we spent hours doing comparisons and helping madshi in every way we could. So I think everyone deserves a chance to say something and doing it in the right way is the way to do it.

[cokeefe]

I'm having a couple of issues with fast forwarding/rewinding on my bluray remuxes. The screen flickers on and off for a while and I get screen corruption until it seems to catch up on itself. No such issues with encodes. The other issue I have is if I put 1080p23 or 1080p24 in my OLED TV's display modes only half the video displays on the TV. I have to leave out those entries for MadVR to play them back correctly in fullscreen exclusive mode on the TV.

Any ideas?


You don't say what GPU you are running, but I've had similar behavior with my RX 480 (loss of signal/sync, corruption). Combined with the poor NGU performance and drivers, I am regretting buying my first non-NVidia card for my dedicated HTPC.

Two things did help in my case:
-force vsync always on in the Crimson driver - it does not seem to honor the "let application decide" flag with MadVR.

-decrease both CPU and GPU queues to the minimum 4

These have reduced but not completely eliminated the flickering (DX11 exclusive).


On another note, @Madshi, I'm one of the silent users who have enjoyed MadVR for several years but never post, as I am not as technical as most folks here. THANK YOU for this remarkable software (especially NGU!)

I also appreciate the move toward simplification and self-configuration - it makes MadVR so much more accessible to the "dull average" crowd I count myself among!

[pirlouy]

All these heuristics like downscaling an image and upscaling it again with different algorithms to see which is closest to the original, are just that, heuristics. They are useful for comparing algorithms, or even designing new ones, but they're no substitute for actually looking yourself.
Comparing screenshot is useful. And there have been a lot of interesting posts these days with useful screenshots comparison.
But asking for an old algorithm to be included without screenshot comparison, I'm not sure it will motivate a dev to accept the request that easily.

As to "personal looking" in the end, I agree with you, but this has to be judged by people who know what to look. Non professional people will not judge the algorithm and are subject to placebo effect. And it is unsure they have done all good settings in the complicated chain of elements to do right !
Again, screenshots avoid endless discussion with nothing.

@others: good job at doing screenshots comparison ! (and also for not embedding big images and using links instead !)

[HillieSan]

You don't say what GPU you are running, but I've had similar behavior with my RX 480 (loss of signal/sync, corruption). Combined with the poor NGU performance and drivers, I am regretting buying my first non-NVidia card for my dedicated HTPC.

Two things did help in my case:
-force vsync always on in the Crimson driver - it does not seem to honor the "let application decide" flag with MadVR.

-decrease both CPU and GPU queues to the minimum 4

These have reduced but not completely eliminated the flickering (DX11 exclusive).


On another note, @Madshi, I'm one of the silent users who have enjoyed MadVR for several years but never post, as I am not as technical as most folks here. THANK YOU for this remarkable software (especially NGU!)

I also appreciate the move toward simplification and self-configuration - it makes MadVR so much more accessible to the "dull average" crowd I count myself among!

I never experienced these issues with my RX 480. In the beginning I had issues with DX9 and fixed them by switching to DX11. DX11 works flawless with latest versions of mpc-be and bluesky frc (flicker free filter). Only the perfomance of NGU medium and up is poor. Yesterday I installed the latest relive drivers and they have a bug at boot time showing snow for a few seconds and lost of audio via hdmi. Otherwise these work fine. I have reported the bug and I guess this will be fixed soon. Be sure you use good quality hdmi cables. In the past I had sync issues or half screens with my nvidia gtx 660 and amd hd 8570 due to drop out (= poor quality) hdmi-to-dvi cables with 1080/75 hz..

[Patrik G]

Madshi should get the nobel prize for NGU !

watching blu ray movies on my 3D LUT calibrated KRP-500M now with NGU almost makes them look like 4K resolution!
and then im not using luma doubling only chroma upscaling

its amazing how clean and sharp the picture is now and with NO sideeffects or any sign of artificial created image.
damn amazing!

i just wonder how the hell did he do it and why the picture quality is so much higher than with NNEDI3 ? :)

btw NGU for luma doubling doesnt work here
to much moire effects
almost like bumping up sharpness +20
but since NGU at high for chroma is so sharp already you dont need it

Scene from i still know what you did last summer blu ray with NGU high
Pioneer KRP-500M
btw the photo has 22 stops of dynamic range so everything from black to white is captured ;)
you should be able to spot the frame of the tv if your blacklevel is setup right
http://privat.bahnhof.se/wb192876/NGU_High.jpg

[har3inger]

try to compare bilateral and super XBR 150.

most obviously the red triangles but not limited to them.

i don't think it is possible to hide the differences with a sub sampling screen. maybe vivid mode or something like that can hide this but...

i'm pretty sure firefox can't hide this what so ever.

this is not a chroma heavy image but it still shows a huge difference with bilateral.

Yeah. Bilateral absolutely shits the bed when there isn't any high contrast information in the luma channel to draw from. It uses this weird, fuzzy, dithered, edge blurring algorithm in these cases which causes the problem with the red triangles. I wouldn't use bilateral unless with chroma SR2 or higher, but bilateral with SR2 does become really quite nice.

How does that same image look with recon soft?

[CruNcher]

Madshi should get the nobel prize for NGU !

watching blu ray movies on my 3D LUT calibrated KRP-500M now with NGU almost makes them look like 4K resolution!
and then im not using luma doubling only chroma upscaling

its amazing how clean and sharp the picture is now and with NO sideeffects or any sign of artificial created image.
damn amazing!

i just wonder how the hell did he do it and why the picture quality is so much higher than with NNEDI3 ? :)

btw NGU for luma doubling doesnt work here
to much moire effects
almost like bumping up sharpness +20
but since NGU at high for chroma is so sharp already you dont need it

Scene from i still know what you did last summer blu ray with NGU high
Pioneer KRP-500M
btw the photo has 22 stops of dynamic range so everything from black to white is captured ;)
you should be able to spot the frame of the tv if your blacklevel is setup right
http://privat.bahnhof.se/wb192876/NGU_High.jpg

Yes but please keep in mind a relative low quantized Blu-Ray in H.264 is a virtual perfect source you can work with up-scaling to 4K which should need very little overall processing.
Neo-XP got some very nice results already even @ NGU-Med which should save some processing and he used much more inferior source Lower Resolution 2nd Generation Blu-Ray Transcodes instead of a 1st Generation 1080p H.264 Source:)

This is where Profiles actually could really become handy todo source based setup decisions in a fast switching way :)

Nice would be these Presets then would become easy accessible through the Player Interface itself (not the MadVR Icon only)

Your result even from the Screen Capture looks nice especially with Samsungs Cinema Magic Bright PP (which unfortunately has its own sharpening filter ontop of it's Cinema color tone mapping which though doesn't create that much over sharpening based on your Nikons camera focus and jpg result at all).

[burfadel]

You don't say what GPU you are running, but I've had similar behavior with my RX 480 (loss of signal/sync, corruption). Combined with the poor NGU performance and drivers, I am regretting buying my first non-NVidia card for my dedicated HTPC.

Two things did help in my case:
-force vsync always on in the Crimson driver - it does not seem to honor the "let application decide" flag with MadVR.

-decrease both CPU and GPU queues to the minimum 4

These have reduced but not completely eliminated the flickering (DX11 exclusive).


On another note, @Madshi, I'm one of the silent users who have enjoyed MadVR for several years but never post, as I am not as technical as most folks here. THANK YOU for this remarkable software (especially NGU!)

I also appreciate the move toward simplification and self-configuration - it makes MadVR so much more accessible to the "dull average" crowd I count myself among!

What driver are you using? The new 16.12.1 drivers are much better than older drivers. I got shot down by saying that a couple of pages back, but seeing as it is working 'perfectly' for me I'm happy with them. D3D 11 Exclusive mode, with thin edges, enhance details, adaptive sharpen shaders, no trading quality for performance, using NGU etc.

Now there are two schools of thought regarding GPU capability. he first is the max out all the settings including NGU etc, so when using a side by side comparison with your eyes 2 inches from a 55 inch screen you may be able to tell a slight quality difference between those settings and lower settings, in the right scene. The second is to use more realistic settings and let the GPU peacefully tick over, but without compromising perceptive quality. The last point is important!

Basically there is no point (in my opinion) to use higher settings when there is no perceptible difference (and not psychological) between actually watching the video as you would (not doing zoomed in analysis or looking at still images closely), especially when you may be using 3 times the GPU power to do so.

Download the latest GPU-z from here:
https://www.techpowerup.com/downloads/SysInfo/GPU-Z/

Once running (you don't need to install, just select 'No' or 'Not Now'), click on the sensors tab. This will show you the GPU clocks, temperature, load, and GPU power draw (on the newer cards). Assuming you have dual screens, play back on the screen you wish to play back on, and have GPU-z on the second screen. Watch the GPU use and adjust settings as necessary. I use NGU Medium (or high) for Luma, Jinc AR for upscaling after doubling (only used if requried), and for downscaling 'use "image downscaling" settings'. For those SSIM is the sharpest but introduces aliasing, 2D 100% is the sharpest but the most GPU unfriendly. I think a good compromise is Jinc with linear light and anti-ringing enabled. For Chroma upscaling having Jinc with anti-ringing and superres (1 or 2) is good.

With upscaling refinement I have 'add grain', and for dithering I have 'Random dithering' enabled. Error Diffusion may be 'better', but have others have pointed out unnecessary when outputting to a 8 or 10 bit display.

So, with the latest drivers, try:
Devices (set to highest display depth the display can handle)
Artifact removal: Debanding Low/high or medium/high, reduce ringing artifacts
Image enhancements: thin edges (1.0), enhance detail (1.0), adaptivesharpen (0.3, for example), anti-rining.
Chroma upscaling: Jinc, anti-ringing, SuperRes (1 or 2)
Image downscaling: Jinc, linear ligh, anti-rining relaxed
Image upscaling: NGU medium (or high IF you can actually see the diffence when watching a video), after doubling algorithm Jinc AR, use image downscaling settings for downscaling.
Upscaling refinement: add grain (maybe 1 or 2 depending on your taste). A tiny bit of grain added can perceptively bring out detaisl whilst hiding artifacts.
Rendering. General settings. Delay playback until queue is full, overlay mode (not sure if that does anything in Win 10), Fullscreen exclsuive mode, separate device both enabled. Large CPU and GPU queue size seems to work fine (say, 64 and 8 or whatever).

The windowed and exclusive mode settings are fine as default (present 8 in advance etc), random dithering, and no trade quality for performance settings.

Now on drivers before 16.12.1 these settings wouldn't work, mainly because of DX 11 Exclusive. It would also not be smooth etc. On 16.12.1 with these settings the GPU just ticks over nicely. You will see the GPU only power draw being low as a result of the GPU clock remaining low, which means the card runs cool, and remains under 60C (for me). It may be so that the fans don't even have to run!

If you need to zoom in to see if something is better, use a ridiculously low resolution (320x240 for example) and blow it up to 4k resolution on a 60 inch screen and then look closely at a still image to see the difference, or any variation of not actually watching it normally, to be able to tell the difference, then the difference is not worth quibbling over if the settings differences require much more processing power.

Back to the new driver, I can even have the computer turning off the display etc when in paused windows mode, and now it will just wake up and play perfectly again! :). Actually, all faults I reported earlier seem to be fixed, so it probably was a driver issue. Just strange Nvidia users were affected as well, it's probably related to updated API support (such as SM 6.0, DirectX 11.4, WDDM 2.1 etc and other changes) in the driver no being perfected, particularly since for both Nvidia and AMD you need the new cards (GTX 1000 series/RX 400 series) to make use of it.

When starting playback after monitor turn off/resume you do need to go back to full screen mode, or windowed mode then fullscreen if you paused in fullscreen, to reset the output. CLSID mentioned this earlier, it's not a driver or madVR thing, it's just a thing with Windows. There haven't been any associated issues with this (such as MPC-HC freezing etc) with the new drivers.

[ryrynz]

Please guys, we need to calm down.

Claiming that these time spent and provided to end-user options are next to pointless is the height of rudeness.

we spent hours doing comparisons

Well if it's not for luma then it's "pointless and amounts to nothing/placebo" apparently.

Saying that he can't see any difference on his outdated setup is all he needed to do. Not the following drivel..

The real world impact to any of those chroma comparisons is so close to nothing/placebo it's barely even worth discussing.


Really, then why are there pages of them in this forum? Why have experts like 6233638 weighed in on such?


As madshi has said a million times, luma is way more noticeable/important.

*facepalm* I'm done.

[burfadel]

The argument is between differences in still images where you have to look closely to tell the difference, or zoomed in output, or low resolution images upsampled, versus actually watching a video and being able to tell the difference. You seem to be in the former category, I'm in the latter. By the sounds of it, Madshi is also in the latter category. If the differences when playing a video and sitting back and not 10 cm from the screen are not perceptible, then putting an extra 100 percent load+ on the GPU for high end Chroma or Luma settings is pointless. Sorry, that's reality. If you can achieve those fine improvements that are only visible when zooming/looking closely 10 cm from the screen on still image without sacrificing GPU performance (or extremely little) then the argument gains for validity.

Spend more time using it for playing videos rather than nitpicking quality settings that have no perceptible difference when ACTUALLY USING it to watch videos normally, then complaining using those settings puts too much load on the GPU. Duh! lols.

[70MM]

Guys I wonder if someone can offer some help please...
I have thee 1080 card and the JVC X9000 projector. Ive always had horrible banding and none of the banding settings on madvr make any difference at all for me.
Im still only using 1080p for all my ripped BDs and NEEDI3, haven't tested NGU yet.
As I play everything at 60Hz RGB 8bit to avoid the horrible slow sync of the JVC for mixed content, is it possible that 60Hz introduces the banding?

Any help would be greatly appreciated.

[mogli]

The real world impact to any of those chroma comparisons is so close to nothing/placebo it's barely even worth discussing.
I do see quite a difference between e.g. Bicubic60 and suber-xbr 100 with the latter beeing much sharper, brighter and thicker (artificial/digital look).

However the question is what's actually better?

People tend to prefer sharper and brighter images but that doesn't mean it's closer to intended reproduction. And with the tone changes caused be different chroma algorithms I'm unsure how much sense stuff like color calibration even makes.

In other words I think that the differences are so big that we'll never come close to correct (whatever that may be). So we're very much back to personal taste.

[Neo-XP]

I have just noticed a weird behavior with the dark halos removal :

http://screenshotcomparison.com/comparison/193397

The processed lines are not straight anymore as it creates artifacts all around them. It is very distracting, even more in motion.

Is there room for improvement there @madshi ? Maybe a less aggressive version ?

There are some artifacts too with just the ringing removal, but they are less noticeable :

http://screenshotcomparison.com/comparison/193398

[HillieSan]

@burfadel Are you using bluesky FRC or SVP?

[StinDaWg]

I guess you won't "waste" money on a screen that shows these differences quite clearly either. If you have a low quality screen then I guess it makes sense to use the lower quality algorithms.
Just because you can't see differences in your ancient setup doesn't mean they don't exist.
Yes, I'm using an "ancient" 2014 60" plasma. Regarded as one of the best tvs available at the time.

It is not worth it to me to use 2-3x more gpu so that a couple blocks of red are slightly sharper on some scenes, and nothing that I would be able to see in motion from a 6-7' seating distance. I don't know why you can't get that through your thick skull.

I will probably switch to superxbr 150ar or NGU low, since render times are only 2-4 ms higher than bicubic 60ar even if it's just placebo in motion.

You have a horrible attitude problem, and are just trolling at this point. Take a step back from the keyboard. It will be ok, I promise.

[nomakewan]

So I've noticed an odd behavior in MadVR. It was there in 0.89, and it's still there in 0.91.4. For files that are encoded (perhaps incorrectly?) with RGB color space, MadVR turns the whole video a deep magenta tint. Switching MPC-HC to EVR Custom Presenter (while changing no other settings) resolves the issue, as does playing the files with VLC, so it does appear to be a MadVR issue. I've been poring over options in MadVR but I can't find anything that fixes it. Is this a known issue? I couldn't find anyone mention it with the search function, or on Google for that matter.

Here's the mediainfo output for one of the affected files:

http://pastebin.com/21bp4Bct

System Specs:
MPC-HC 1.7.9
madVR 0.91.4
LAVFilters 0.66.0
Windows 7 Ultimate SP1 x64
i7-4790K
GeForce 760GTX

[HillieSan]

Guys I wonder if someone can offer some help please...
I have thee 1080 card and the JVC X9000 projector. Ive always had horrible banding and none of the banding settings on madvr make any difference at all for me.
Im still only using 1080p for all my ripped BDs and NEEDI3, haven't tested NGU yet.
As I play everything at 60Hz RGB 8bit to avoid the horrible slow sync of the JVC for mixed content, is it possible that 60Hz introduces the banding?

Any help would be greatly appreciated.

I have a X5000 and it works fine here.
Important: Turn off the dynamic contrast in de nvidia drivers. This causes banding! Use PC 0-255 full color space in driver setting.
Set rgb 4:2:2 (not 4:4:4) in driver. Leave X5000 on auto color space.
I leave most options in the X5000 as default except for blur reduction setting which I have turned off. They introduce artifacts. I play everything at 60 Hz. Even 24hz movies. This works best.

I hope this helps.

[burfadel]

You LAVFilters is a really old version! 0.68.1-53 is the latest interim build (from 10/12/2016). Also your MPC-HC is out of date (1.7.10.269 is the latest). In LAVFilters you can select to output as whatever you like. If the input was RGB, you can force output in YV12, RGB32 or whatever else and see if that helps. If you go MPC-HC BE, do note that it has a different numbering system, such that 1.5.0 is the latest and is 'newer' than the MPC-HC build.

Are you using MPC-HC 32-bit or 64-bit?

[nomakewan]

You LAVFilters is a really old version! 0.68.1-53 is the latest interim build (from 10/12/2016). Also your MPC-HC is out of date (1.7.10.269 is the latest). In LAVFilters you can select to output as whatever you like. If the input was RGB, you can force output in YV12, RGB32 or whatever else and see if that helps. If you go MPC-HC BE, do note that it has a different numbering system, such that 1.5.0 is the latest and is 'newer' than the MPC-HC build.

Are you using MPC-HC 32-bit or 64-bit?
32-bit. Also do note that, despite everything you said, I can leave all of those things the same, then switch the renderer to EVR Custom (or, really, anything that isn't madVR) and there are no issues with the video color. Only when I'm using madVR is this an issue.

Also, I -did- try changing the output settings in LAV, but that did not affect the horrible magenta color output by madVR. For the sake of argument, however, I'll go ahead and update my MPC-HC and LAV. I highly doubt it will change anything, but better safe than sorry, right?

EDIT: Am now on MPC-HC 1.7.10x86 and LAVFilters 0.68.1 Stable. No change, as expected.

[HillieSan]

32-bit. Also do note that, despite everything you said, I can leave all of those things the same, then switch the renderer to EVR Custom (or, really, anything that isn't madVR) and there are no issues with the video color. Only when I'm using madVR is this an issue.

Also, I -did- try changing the output settings in LAV, but that did not affect the horrible magenta color output by madVR. For the sake of argument, however, I'll go ahead and update my MPC-HC and LAV. I highly doubt it will change anything, but better safe than sorry, right?

EDIT: Am now on MPC-HC 1.7.10x86 and LAVFilters 0.68.1 Stable. No change, as expected.

I remember that I had magenta ouput once. It was caused by a setting in de device section of your monitor/tv/projector in madvr. The rgb/ycbcr must match the device.

[nomakewan]

I remember that I had magenta ouput once. It was caused by a setting in de device section of your monitor/tv/projector in madvr. The rgb/ycbcr must match the device.
My settings are correct (E248WFP; PC levels, 8-bit, no 3D, no calibration, passthrough HDR) for my monitor and had never been an issue until I ran into this odd files that had RGB color space. Every single other file I have ever played plays through perfectly fine. So it's not exactly killing me--I can just play these few files in VLC if I have to--but I figured it was worth reporting since it appears to be a legitimate bug in how madVR is handling them.

[nevcairiel]

My settings are correct (E248WFP; PC levels, 8-bit, no 3D, no calibration, passthrough HDR) for my monitor and had never been an issue until I ran into this odd files that had RGB color space.

Instead of "no calibration", its usually advised to set it to "this display is already calibrated" and set the screens properties below that selection appropriately. Only this way madVR can know how your screen behaves and properly show any and all content - which might as well fix your video.
With "no calibration" all videos are passed to the screen without changes, which would result in wrong output on anything that doesn't natively match your screen.

Otherwise, madshi will likely need a sample video.

[Magik Mark]

Guys,

Did somebody posted a test image for Nedi3 64 neuron vs NGU all variants?

I don't know if it was just my eyes. Nedi3 64 produces cleaner and more natural video. It manages noise very well as depicted in video such as TWD. TWD is very grainy and Nedi3 doesn't sharpen this. Can anybody concur? The only downside is that it hugs too much resources but if you have something to spare then this algo seems to be the best

[Knight77]

a greeting to all, activate 200% super sampling you think it is helpful when upscaling to 1080p to 4K? what do you advise me to use as an algorithm after doubling when upscaling to 4K 1080p and use as a quality algorithm NGU very high to me like a very defined image sharp

I'm a bit confused about this...should I use it when upscaling 1080p to 4K or is better to be used when you watch things at native screen resolution, like 4K video on 4K TV ?

And how much rise up the quality of the image?

[mueslibrown]

Please guys, we need to calm down. The aggressiveness has been increasing in the last few days. We do not know the other economical situation, I think we all have the right to share our opinion and that we can do it without being rude and avoiding offensive comments. We are all trying to do madVR better day after day, we spent hours doing comparisons and helping madshi in every way we could. So I think everyone deserves a chance to say something and doing it in the right way is the way to do it.

Well said ...

[nomakewan]

Instead of "no calibration", its usually advised to set it to "this display is already calibrated" and set the screens properties below that selection appropriately. Only this way madVR can know how your screen behaves and properly show any and all content - which might as well fix your video.
With "no calibration" all videos are passed to the screen without changes, which would result in wrong output on anything that doesn't natively match your screen.

Otherwise, madshi will likely need a sample video.
I had already tried changing this setting just for kicks, but changing it to "this display is already calibrated" and then going through all the possible combinations of calibrations (none of which my monitor has--because it's not calibrated) merely either makes the problem worse by making the magenta a little closer to 'neon' vividness or merely makes the magenta a little softer. The actual fact that magenta is outputted where black should be does not change. As before, this setting has never caused me any issues in all the years I've been using madVR, nor with any of the probably hundreds of thousands of video files I've thrown into my player over those years. I just happened to stumble upon a few clips with RGB color space that threw it for a loop, that's all.

I have samples I can provide if truly necessary, that is, if the mediainfo output alone is not sufficient to pinpoint the issue at play. As this is my first time posting on Doom9 (I had an account years ago but it's apparently been pruned), what method is generally accepted for submitting samples nowadays?

[Werewolfy]

I just tested NGU on 720p anime and I still prefer NNEDI3, it hides aliasing better.
About the new UI I'm quite happy with it, I don't have the impression to be limited in my choices so if it's easier to understand for the newcomers, I'm fine with it.

And thanks to ryrynz, Asmodian and Neo-XP I can clearly see now that chroma has an impact on real life contents. I wasn't aware of that, I used Jinc most of the time without thinking that there was a real better option. So thank you guys!

[Neo-XP]

I have just noticed a weird behavior with the dark halos removal :

http://screenshotcomparison.com/comparison/193397

The processed lines are not straight anymore as it creates artifacts all around them. It is very distracting, even more in motion.

Is there room for improvement there @madshi ? Maybe a less aggressive version ?

There are some artifacts too with just the ringing removal, but they are less noticeable :

http://screenshotcomparison.com/comparison/193398

In addition, a comparison between madVr's dark halos removal VS FineDehalo (http://avisynth.nl/index.php/FineDehalo) :

http://screenshotcomparison.com/comparison/193421

madVR is better at preserving details while FineDehalo is better at cleaning the edges without artifacts.

Maybe you can get the best of both worlds @madshi ?

[CruNcher]

Not really better it blurs it again way to much though their is no perfect way todo it for fixed content you can predict it might work but you never will get without trained AI or at least some metric analysis the results you want and the AI has to analyse even each specific scene and make a guess how the output might look good for your output target.

Also we are still fixating to much to spatial results only wee need to get more actual moving results :D

Some input surely could be very well predicted Blu-Ray for example and based on this wee could buildup relative low complexity chains which wouldn't kill your resources in 1 second for no reasonable balanced result overall.

But we really need a more convenient way (apart from the MadVR icon in the Systray) to edit and to chose those presets with 1 or 2 clicks and or Keyboard shortcuts cycling through them this will be also most important for making it easier for AVG joes overall.

But this needs overall tighter integration into a player interface and im not sure if Madshi wants this to happen overall but usability wise the Systray idea seems sub optimal for preseting stuff having the windows taskbar all the time open is not really content centric nice todo.

Defining some kind of input quality strategy to work with as a base for the Presets seems to be a good start, Blu-Ray (offline)/ Youtube/Netflix/Vudu/Amazon (online) are very interesting targets because they have some predictability you could buildup on content wise :)

[Neo-XP]

Not really better it adds blur again way to much

Neither is the best.
The good thing is that you can change a lot of variables with FineDehalo, so if you don't like some aspects if it, you can fine-tune it as you wish.
It is under WTFPL licence, so no problem for madshi to use the code :cool:

[CruNcher]

I absolutely agree and indeed it has some nice properties in your overall spatial result overall the result could in motion be better indeed you would surely do some nice presets but i wouldn't even say it's better dehaloing in that specific case of yours much better aliasing prevention on edges and overall it doesn't hurt the end result so much after all :)

But now is the question can't you get a similiar result without it and a different MadVR chaining and maybe then with less overhead at all ;) ?

Ringing seems not really the problem here then Aliasing also the question is does it holdup in motion as well with your 4K target ? :)

[Neo-XP]

But now is the question can't you get a similiar result without it and a different MadVR chaining and maybe then with less overhead at all ;) ?

No, because the ringing and dark halos are unfortunately already in the source, before any processing.

Ringing seems not really the problem here then Aliasing also the question is does it holdup in motion as well with your 4K target ? :)

Yes ! And I've made some comparisons (with original included :p) in 4K (Original / madVR / FineDehalo) :

https://s23.postimg.org/prwfllimf/Original.jpg (https://postimg.org/image/prwfllimf/) https://s23.postimg.org/wh2z1m3yf/mad_VR.jpg (https://postimg.org/image/wh2z1m3yf/) https://s23.postimg.org/g4sxbvpmv/Fine_Dehalo.jpg (https://postimg.org/image/g4sxbvpmv/)

And lossless in FHD with 100% zoom :

https://s27.postimg.org/sw3vr2eof/Original.png (https://postimg.org/image/sw3vr2eof/) https://s27.postimg.org/4ebs36c3z/mad_VR.png (https://postimg.org/image/4ebs36c3z/) https://s27.postimg.org/ko1y62mrz/Fine_Dehalo.png (https://postimg.org/image/ko1y62mrz/)

I used those parameters for FineDehalo (probably not the best, I was just testing) : FineDehalo(thmi=128, thma=128, thlimi=50, thlima=50, contra=1.0).
I disabled the "add grain" enhancement of madVR for the test, to compare them more easily without added randomness.

Clearly, my vote goes to FineDehalo, in still and in motion.

[Warner306]

Guys I wonder if someone can offer some help please...
I have thee 1080 card and the JVC X9000 projector. Ive always had horrible banding and none of the banding settings on madvr make any difference at all for me.
Im still only using 1080p for all my ripped BDs and NEEDI3, haven't tested NGU yet.
As I play everything at 60Hz RGB 8bit to avoid the horrible slow sync of the JVC for mixed content, is it possible that 60Hz introduces the banding?

Any help would be greatly appreciated.

The refresh rate wouldn't introduce any banding. That is the fault of the mastering process of the Blu-ray.

Are you outputting correct RGB levels? Your PC should be set to RGB Full and so should your projector.

[plasma]

I liked to use upscaling refinement such as adaptive sharpen and line thinning but now as I upgraded to a 4k tv it uses way too much performance. Using "image enhancements" does not give similar results, albeit much better perf.

I assume that upscaling refinements uses the upscaled 4k images and then applies to refinements while image refinements does it at the source resolution, whatever it is. Is there any way to apply these refinements at a half upscaled image (aka 1080p) so you get better balance in terms of performance penalty.

[BluesFanUK]

Ever since the new update i'm getting dropped frames on everything I watch.

Using a 980ti and 1440p monitor. 1080p@60fps seems to come off worst. I put all settings even to the lowest settings and still get 40-50ms.

It's unusable for me, and I've an inkling it's something to do with this new combined image upscale feature.

What's strange is despite the ridiculous number of dropped frames in GPU-Z my GPU is using barely 50% of power....

[AntonP]

Hello, which AMD videocard is better for BD playpack on 1080p screen?
R9 Nano or RX480? both overclocked of course
NGU high chroma upscaling + all other cool things on...
AMD was chosen because of no 24p problem (23.971 - nvidia) and rx480 and nano are only short pcb boards (have not much space in htpc case)

[pirlouy]

What's strange is despite the ridiculous number of dropped frames in GPU-Z my GPU is using barely 50% of power....
GPU-Z causes dropped frames. The same for f.lux (default setting).
Try to use Open Hardware Monitor for GPU infos. Way better than GPU-Z IMO thanks to useful graph.

[CruNcher]

No, because the ringing and dark halos are unfortunately already in the source, before any processing.



Yes ! And I've made some comparisons (with original included :p) in 4K (Original / madVR / FineDehalo) :

https://s23.postimg.org/prwfllimf/Original.jpg (https://postimg.org/image/prwfllimf/) https://s23.postimg.org/wh2z1m3yf/mad_VR.jpg (https://postimg.org/image/wh2z1m3yf/) https://s23.postimg.org/g4sxbvpmv/Fine_Dehalo.jpg (https://postimg.org/image/g4sxbvpmv/)

And lossless in FHD with 100% zoom :

https://s27.postimg.org/sw3vr2eof/Original.png (https://postimg.org/image/sw3vr2eof/) https://s27.postimg.org/4ebs36c3z/mad_VR.png (https://postimg.org/image/4ebs36c3z/) https://s27.postimg.org/ko1y62mrz/Fine_Dehalo.png (https://postimg.org/image/ko1y62mrz/)

I used those parameters for FineDehalo (probably not the best, I was just testing) : FineDehalo(thmi=128, thma=128, thlimi=50, thlima=50, contra=1.0).
I disabled the "add grain" enhancement of madVR for the test, to compare them more easily without added randomness.

Clearly, my vote goes to FineDehalo, in still and in motion.

you should have made the cut higher above the hairdresser hat the top part was also problematic ;)

[BluesFanUK]

GPU-Z causes dropped frames. The same for f.lux (default setting).
Try to use Open Hardware Monitor for GPU infos. Way better than GPU-Z IMO thanks to useful graph.

GPU-Z is not the culprit. It's the combined image doubling function. Anything above 30fps and it becomes a complete mess.

[Gagorian]

GPU-Z is not the culprit. It's the combined image doubling function. Anything above 30fps and it becomes a complete mess.

I haven't heard of anyone with your problem after the update so I would not blame the change in madVR options menu. Have you set your GPU for maximum power when running your media player from the NVIDIA control panel? NVIDIA drivers do not handle power mangement properly in optimized mode.

[BluesFanUK]

I haven't heard of anyone with your problem after the update so I would not blame the change in madVR options menu. Have you set your GPU for maximum power when running your media player from the NVIDIA control panel? NVIDIA drivers do not handle power mangement properly in optimized mode.

Nowt wrong with my card.It was all working dandy till I got the latest update. The 980ti is a powerful GPU, but for anything above 30fps it cannot handle many cranked up settings.

Average frame time is 16ms @ around >30fps, with the new update i've tried to scale everything back to how it was and it's still dropping frames.

Settings atm for 1080 >30fps

Super XBR 125 SR1
SSIM 1D 100%
Jinc AR
No refinement

It's now rendering at 30ms and dropping frames. I've ditched MadVR for the time being otherwise it's going to kill my card.

[BluesFanUK]

Right I think i've narrowed down what the problem was. I installed a much older version and reset everything to default, slowly going through each setting. The artifact removal 'reduce ringing artifacts' had no baring on my older config, now it adds a good 10ms when watching anything over 30fps.

I've unticked it and it seems to have done the trick... for now anyway.

[flashmozzg]

Hello, I've enabled NGU-med for both chroma and luma upscaling and noticed that with "let madVR decide" option for alg after NGU it chose Bicuber AR 60. Is it really good or should I choose another? Also, I have R9 290 and it only takes 5ms avg to render 10bit 4:2:0 480p at 1080p with NGU-med. Nice!

[Asmodian]

It's now rendering at 30ms and dropping frames. I've ditched MadVR for the time being otherwise it's going to kill my card.

I am not sure what the "kill my card" is about but in their newer drivers Nvidia changed the default power management mode from "adaptive" to "optimal power". Optimal power clocks too low for madVR though so you need to change back to adaptive, after I do this I don't get any dropped frames with the newest madVR at ~85% GPU usage.

Edit:
Right I think i've narrowed down what the problem was. I installed a much older version and reset everything to default, slowly going through each setting. The artifact removal 'reduce ringing artifacts' had no baring on my older config, now it adds a good 10ms when watching anything over 30fps.

I've unticked it and it seems to have done the trick... for now anyway.

How odd, though I don't use reduce ringing artifacts normally it only adds ~2.3ms on my system. I don't think the Titan X (pascal) is 4x faster than the 980Ti. ;)

Hello, I've enabled NGU-med for both chroma and luma upscaling and noticed that with "let madVR decide" option for alg after NGU it chose Bicuber AR 60. Is it really good or should I choose another? Also, I have R9 290 and it only takes 5ms avg to render 10bit 4:2:0 480p at 1080p with NGU-med. Nice!

Can you use NGU-high for luma? I really do recommend leaving everything on "let madVR decide" and picking the highest luma option you can. Once you have maxed out luma you can start turning up the other settings a little, if you still have performance available.

Edit2:
The point of "let madVR decide" is to use reasonable options given the performance the choice for luma implies. Anything below NGU-veryHigh uses Bicubic60 AR for chroma doubling, which is a very reasonable option for chroma doubling.

[BluesFanUK]

I am not sure what the "kill my card" is about but in their newer drivers Nvidia changed the default power management mode from "adaptive" to "optimal power". Optimal power clocks too low for madVR though so you need to change back to adaptive, after I do this I don't get any dropped frames with the newest madVR at ~85% GPU usage.

Edit:


How odd, though I don't use reduce ringing artifacts normally it only adds ~2.3ms on my system. I don't think the Titan X (pascal) is 4x faster than the 980Ti. ;)



Can you use NGU-high for luma? I really do recommend leaving everything on "let madVR decide" and picking the highest luma option you can. Once you have maxed out luma you can start turning up the other settings a little, if you still have performance available.

That's weird. I checked this setting before I had a moan here and changed it to Adaptive, then got a weird buzzing sound from my GPU when I started testing videos.

I'll guess it was just overheating at the time or something.

Anyway, i'm back to default settings working my way back to my original settings, and i'm getting a huge number of dropped frames when I open videos again and fast forwards/rewind (but they aren't dropping as the video plays thankfully). I thought I elminated this issue a year back.

I've stopped it opening in Full Exclusive mode, enbabled Windows Overlay and User Separate Device for presentation but the issue persists. Any ideas?

[Asmodian]

It is supposed to drop frames on open or seek, that is normal. You can reduce the number with the rendering -> general settings option "Delay playback start until render queue is full". The frames dropped on playback or seek probably aren't important though, I prefer faster response times without that setting even though it drops a lot of frames during open/seeks.