madVR - high quality video renderer (GPU assisted)

[nevcairiel]

I don't suppose it is possible to have 10 bit output with windowed mode?

No, thats currently not possible. Windows itself would need to run in 10-bit mode, and it doesn't support that.

[ryrynz]

Has anyone see any difference on their 10 bit panel when using error diffusion, now that madVR allows it?

[mindz]

Im using the Intel HD4600 iGPU and therefor use DXVA upscaling (great results imo). Am i understanding correctly that for me 10 bit output will never work because DXVA scaling is always done in 8 bit? Ive got a fairly new Sony TV (50W829B) and think its 10 bit capable. What would be the best settings for me bit-wise? I dont want any conversions from 8 to 10 and back to 8 bit if its not needed/not doing anything.

[ryrynz]

Madshi, is that NNEDI3 pixel shift going to be fixed up? Been there for a while.

[Qaq]

What we need to output 12 bit picture? I mean true 12 bits form madVR and not zero padding from videocard.

[omarank]

My preferences haven’t changed regarding debanding algorithms. I prefer the “high” preset, with the trade quality option “don’t analyze gradient angles for debanding” selected, over the latest Shiandow’s script. Without the trade quality option selected, I see a slight blurring in the images. It may be due to stronger debanding.

I find that madVR’s debanding algos make the images look better than Shiandow’s algo does. It is perhaps due to madVR’s debanding being monochrome, as it was pointed out by 6233638. In the case of dithering algos too, I prefer mono dithering. My eyes probably like mono debanding/ dithering better.

[madshi]

Is there any significant benefit to higher cpu/gpu queues and presenting more video frames in advance over something small like 4/4/4?
Bigger queues / more prepresented frames means more protection against frame drops when your PC gets busy doing something in the background. Also, might be useful when using smooth motion FRC or when re-rendering frames when a fade is detected (for debanding).

Im using the Intel HD4600 iGPU and therefor use DXVA upscaling (great results imo). Am i understanding correctly that for me 10 bit output will never work because DXVA scaling is always done in 8 bit? Ive got a fairly new Sony TV (50W829B) and think its 10 bit capable. What would be the best settings for me bit-wise? I dont want any conversions from 8 to 10 and back to 8 bit if its not needed/not doing anything.
Scaling will be 8bit. Some other stuff after scaling might produce more bits, though, e.g. calibration stuff, levels conversion, sharpening etc. So 10bit could still be useful. The best setting is that which produces the best quality. Which nobody can judge for you without seeing your specific GPU -> display setup. You'll have to let your eyes be the judge.

Madshi, is that NNEDI3 pixel shift going to be fixed up? Been there for a while.
Fixing it is not a problem, but can potentially lose a bit quality. I do fix the shift when you activate SuperRes, because SuperRes works better that way. But otherwise I don't fix it, if I don't have to, because not fixing it produces better quality.

[madshi]

What we need to output 12 bit picture? I mean true 12 bits form madVR and not zero padding from videocard.
Windows does not offer any APIs to output 12bit. It supports 16bit output, but in a weird way and the GPU drivers don't handle that well. Because of that it's not supported by madVR.

My preferences haven’t changed regarding debanding algorithms. I prefer the “high” preset, with the trade quality option “don’t analyze gradient angles for debanding” selected, over the latest Shiandow’s script. Without the trade quality option selected, I see a slight blurring in the images. It may be due to stronger debanding.

I find that madVR’s debanding algos make the images look better than Shiandow’s algo does. It is perhaps due to madVR’s debanding being monochrome, as it was pointed out by 6233638. In the case of dithering algos too, I prefer mono dithering. My eyes probably like mono debanding/ dithering better.
Thanks for your feedback!

Can you play with the custom deband settings? You can activate them via custom keyboard shortcut (see settings dialog), and then you can modify the settings by using the arrow keys. You can then modify the "angleBoost" and "maxAngle" values to fine tune the angle analyzation stuff. Maybe you can find a setting for "high" which you like better than the old and current "high" preset?

[ryrynz]

Windows does not offer any APIs to output 12bit. It supports 16bit output, but in a weird way and the GPU drivers don't handle that well. Because of that it's not supported by madVR.


Maybe it's gotten better since then? I have it enabled in MPDN on the 750Ti and I haven't encountered any issues that I know of.


Fixing it is not a problem, but can potentially lose a bit quality. I do fix the shift when you activate SuperRes, because SuperRes works better that way. But otherwise I don't fix it, if I don't have to, because not fixing it produces better quality.

Potentially or definitely? If it's an easy fix why not just try it and see?

[huhn]

Potentially or definitely? If it's an easy fix why not just try it and see?

you have to do an extra scaling operation so it's not good.

[Nevilne]

you need to scale after NNEDI3 anyway so it's only origin shift (note that i'm not sure if that's madvr's shift, or this is about chroma)

[Ava Pug]

v0.88.9/10 has fixed my issue with DX11 and my Intel GPU. Thanks!

edit: I have noticed it still goes glitchy gif like when its playing in windowed mode but I can live with that since I normally watch fullscreen.

[huhn]

you need to scale after NNEDI3 anyway so it's only origin shift (note that i'm not sure if that's madvr's shift, or this is about chroma)

you don't need to scale after nnedi3 luma doubling.

chroma just needs to be scaled differently but that's it.

[MS-DOS]

@huhn, at some point you said you knew registry tweaks to force the driver to not dither, didn't you? Can you write up a small summary on those registry keys? Do you know that only for NVidia or also for AMD and Intel? Thanks!


I'll answer for huhn: https://www.doom9.org/showthread.php?t=172128 My quote in the first post.

[SecurityBunny]

One final question. Apologies for my ignorance.

Do I need to make sure the average rendering time is below the vsync interval for smooth motion to function as intended and be judder free? Or do I only need to worry about being under the movie frame interval?

[madshi]

Maybe it's gotten better since then? I have it enabled in MPDN on the 750Ti and I haven't encountered any issues that I know of.
MPDN does some extra processing, which is GPU specific, to work around driver issues. And it seems the GPU drivers are doing extra stuff behind the scenes when using 16bit output. I don't like any of that. I want untouched output, not GPU driver post-processed output. Furthermore, most displays out there today don't even support 10bit natively. So why using a badly supported and obviously GPU driver post-processed 16bit output mode, if the display's native bitdepth is at best 10bit, anyway?

Potentially or definitely? If it's an easy fix why not just try it and see?
You can't move an image by 0.5 pixels losslessly. Moving the image does come at a certain cost. How much cost is hard to say. NNEDI3 produces somewhat "soft" upscales, so maybe because of that the loss wouldn't be big. But tell me: Why should I be doing extra processing, which not only costs extra GPU power, but also may lose some image quality? Just so that you can compare NNEDI3 upscales better to other image upscaling configurations? That makes no sense. An extra processing step should only be performed if it's useful. I see no usefulness in fixing the NNEDI3 pixel shift, for normal video playback.

you need to scale after NNEDI3 anyway so it's only origin shift
If you play e.g. Blu-Rays on a 4K display, there's no further scaling needed after NNEDI3. However, in many other situations you're right, and NNEDI3 is followed up by one more linear scaling algo. In that case fixing the NNEDI3 pixel shift would be possible without losing much (if any) quality. At one point madVR did exactly that. But it caused problems (artifacts) with the anti-ringing filter, so that's why I disabled that again.

Once again: Why fixing the pixel shift? There's no good reason to do that. At least none that comes to my mind.

I'll answer for huhn: https://www.doom9.org/showthread.php?t=172128 My quote in the first post.
Ah, thanks! How about NVidia and Intel?

One final question. Apologies for my ignorance.

Do I need to make sure the average rendering time is below the vsync interval for smooth motion to function as intended and be judder free? Or do I only need to worry about being under the movie frame interval?
Let's make it easier: As long as you have no frame drops and glitches, you're fine.

[huhn]

And what happens if you go back to v0.88.8 and test that with the 32bit madVR first? I suppose probably you'll get a black screen then, too?

yes just test 0.88.8 32 into 64 bit. still broken so nothing wrong with your new code i guess. which is very good.

32 bit nnedi3 on nvidia is broken anyway that's to there driver so no huge lost here.
it would be nice if some can reproduce this issue like it is.

you should have gotten an error report from me yesterday.

this error comes when a file is played to the end with the nnedi3 error.

edit:
about dithering.

there is a chance nvidia isn't dithering.
James Freeman stated that the GPU driver "damaging algorithm" add banding so the results is not dithered. which is an indicator for this.

but i haven't test this yet myself. i need my TV to do a proper test and that's not connected to my nvidia PC.

[iSunrise]

It's not harmful like Lanczos8 is, just not the best option to use.
To be fair, I also used Lanczos 8 quite a lot, but ONLY because we didn't have a native (fast and max. quality) and mild sharpening filter within madVR.

The results with Lanczos 8 and AR were the sharpest of them all and still relatively cheap performance-wise, without ever having to invoke NNEDI3. But it was never really used other than to just sharpen content by me.

As such, I would not really miss it though.

EDIT: madVR now has sharpening, need to retest.

[huhn]

there is finesharp and lumasharpen in madVR

[leeperry]

I think that's a bug in the Intel GPU driver. Doesn't happen with AMD or NVidia. However, it doesn't happen on my HD4000, either. So it's hard for me to work on this. I'll get a newer laptop with Intel GPU later this year, though, then I can investigate this in depth.
When I moved from XP to W7 I complained about the flashing black frames when switching between FSE and FSW(which didn't occur on XP) and you said that they were expected behavior? The zombie frames on top of it when enabling Aero are also "normal" apparently? XP made pristine FSE/FSW transitions, that was so cool :o

you said you knew registry tweaks to force the driver to not dither, didn't you? Can you write up a small summary on those registry keys? Do you know that only for NVidia or also for AMD and Intel?
It's all here for AMD. (http://forum.doom9.org/showpost.php?p=1721434&postcount=7)

[iSunrise]

there is finesharp and lumasharpen in madVR
Yes, you are correct, thanks for reminding me, I must have completely overlooked that. I edited my above post so that it reflects only my specific usage of Lanczos 8.

[Xaurus]

Can I see a screenshot of the complaint, please?
Certainly.

http://home.halden.net/mordor/madvr.png

[XMonarchY]

Yey! The latest madVR 88.10 fixed D3D11 presentation glitches problem on my TV. With 88.8 release, I was able to enable D3D11 and Sync sub-option only on my 120Hz monitor, but not my TV @ 23Hz (I would get presentation glitches). This new build fixed it!

Someone else was asking whether they could see dithering when 10bit+ mode selected. On my 12bit TV (can select 12bit in nVidia CP) and with madVR set to use 10bit+ depth, I do NOT see any dithering what-so-ever, even though Error Diffusion Option 2 (+ both sub-options) is selected. Is that normal? Maybe at 12bit, ER or any kind of dithering is practically not noticeable?

[huhn]

Yey! The latest madVR 88.10 fixed D3D11 presentation glitches problem on my TV. With 88.8 release, I was able to enable D3D11 and Sync sub-option only on my 120Hz monitor, but not my TV @ 23Hz (I would get presentation glitches). This new build fixed it!

Someone else was asking whether they could see dithering when 10bit+ mode selected. On my 12bit TV (can select 12bit in nVidia CP) and with madVR set to use 10bit+ depth, I do NOT see any dithering what-so-ever, even though Error Diffusion Option 2 (+ both sub-options) is selected. Is that normal? Maybe at 12bit, ER or any kind of dithering is practically not noticeable?
ED is not used at 10 bit output.

and getting 12 bit in a TV means nothing. it is totally default that this is possible.

[madshi]

When I moved from XP to W7 I complained about the flashing black frames when switching between FSE and FSW(which didn't occur on XP) and you said that they were expected behavior? The zombie frames on top of it when enabling Aero are also "normal" apparently?
I don't remember what you said or I said ages ago. The switching between FSE and windowed mode was improved a while ago. With Intel GPUs there are still some weird things going on for some users, when using DX11 presentation or "use a separate device for presentation". But that's limited to Intel.

Certainly.

http://home.halden.net/mordor/madvr.png
That doesn't look like v0.88.10 to me. Are you sure you really updated? Maybe there are more than one madVR installations on your PC? E.g. JRiver MC comes with its own installation.

I do NOT see any dithering what-so-ever, even though [B]Error Diffusion Option 2 (+ both sub-options) is selected. Is that normal? Maybe at 12bit, ER or any kind of dithering is practically not noticeable?
madVR outputs 10bit, nor 12bit. The rest is probably zero bits added by NVidia. But it's quite possible that you can't see dithering at 10bit. Even at 8bit dithering noise is already quite low. At 10bit it's 4x lower than at 8bit. The more important question is whether the image is still perfectly smooth and banding-free at 10bit, compared to 8bit. That will depend on the display.

ED is not used at 10 bit output.
It is now, with v0.88.9/10.

[XMonarchY]

madVR outputs 10bit, nor 12bit. The rest is probably zero bits added by NVidia. But it's quite possible that you can't see dithering at 10bit. Even at 8bit dithering noise is already quite low. At 10bit it's 4x lower than at 8bit. The more important question is whether the image is still perfectly smooth and banding-free at 10bit, compared to 8bit. That will depend on the display.


Oh God - yes, the image is perfectly smooth, not a single band even in low quality content. It is just hard to really tell whether 12bit with previous release (10bit madVR setting = Ordered Dithering, even if ER is selected) looks better/worse with current release (10bit madVR setting = ER, if selected).

[shaolin95]

Might be that your driver is dithering. I don't know.

@huhn, at some point you said you knew registry tweaks to force the driver to not dither, didn't you? Can you write up a small summary on those registry keys? Do you know that only for NVidia or also for AMD and Intel? Thanks!


One extra piece of information. When I do the same ramp 10 bit test with MPDN instead of MPC with MADVR, I cannot get the "banding" to show smooth even when switching to full screen in 10 bit or higher. It just looks 8 bit and clearly visible bands.
So I only get the nice smooth image with MPC/MADVR but that still leaves the question open about why it gets smooth if Nvidia Control Panel is not giving me the option for 10bcp at all.

A user just PMed me telling me he sees the 10bcp option at 24Hz but not at 60Hz. I tried yesterday at 48Hz so I will check. Perhaps somehow my project is still capable of 10bcp after all.

As a side not, this latest Nvidia driver seem to mess up the TV levels or at least it was making the projector detected Auto incorrectly after the upgrade. No other setting was changed.

In any case, more testing to come later tonight.


BTW, when in the Chroma options I am selecting the NEDDI3 32 setting but what is the SuperRes option in there supposed to do? I know it increases the performance hit but not sure what I am gaining.

Thanks and sorry for the many Qs!

[Budtz]

Im experiancing a short black screen when switching from FSE to windowed mode in mpchc in d11 mode. It switches instantly in d9 mode. Is this intended?

[Siso]

Im experiancing a short black screen when switching from FSE to windowed mode in mpchc in d11 mode. It switches instantly in d9 mode. Is this intended?

Same here, along with the high present stats in D3D11 windowed mode.

[Anime Viewer]

BTW, when in the Chroma options I am selecting the NEDDI3 32 setting but what is the SuperRes option in there supposed to do? I know it increases the performance hit but not sure what I am gaining.

Thanks and sorry for the many Qs!

madshi wants to focus on debanding effects/comparison at the moment, so he may not be to interested in answering that question for you. Chroma's SuperRes option effects the sharpness, anti-alaising, softness, and anti-ringing chroma. Most people have a hard time seeing differences. From full screen video check and uncheck it clicking apply each time to see if you can notice a difference. If you don't then you might as well leave it unchecked and save resources for other features. NNEDI3 is very resource demanding, and many people (myself included) see better performance and visual quality when they combine a lesser chroma scaler with ChromaSuperRes then they do with NNEDI3. Your opinion may be the same or differ, so experiment and make your own conclusion.

[Werewolfy]

Glad it's better now. For that previous test picture you might want to try setting "margin" to a high value (like 0.5 or 1.0) that might work better for lower quality sources, it shouldn't be used for high quality sources though.

Shame that the grain is too strong, in theory it should add exactly as much grain as was removed by the banding algorithm (if the original was dithered). Did the grain improve the image at all, or was it just "grainier"?

Yes I played with the settings and post my opinion just below ;) (I take a lot of screenshots but I didn't post all of them

Honestly it's just grainier and I can't see an improvement. The grain is a lot stronger than dithering here.


Looking forward to the results of your further tests!


I've done a lot of comparisons. Shiandow's algo does a very good job at debanding but the high preset blurs less details... Well it's not that simple. Shiandow's algo preserves more the shape of the original picture but blurs the details.

[img=http://s5.postimg.org/b2buwdqpz/high5008.png] (http://postimg.org/image/5e5k5hmdf/full/)
[/url]
Power : 0.60 - margin : 0.30 : [img=http://s5.postimg.org/g67pybr7r/shi5060030.png] (http://postimage.org/index.php?lang=french)

Power : 0.80 - margin : 0.30 [img=http://s5.postimg.org/6cqilip3b/shi5080030.png] (http://postimage.org/index.php?lang=french)


But if I use high values it can beat sometimes the high preset.

[img=http://s5.postimg.org/inaae3wcn/high3008.png] (http://postimage.org/index.php?lang=french)

0.60-0.30 : [img=http://s5.postimg.org/5zuj1lhfr/shi3060030.png] (http://postimage.org/index.php?lang=french)

0.80-0.50 [img=http://s5.postimg.org/w9k6y4sqv/shi3080050.png] (http://postimage.org/index.php?lang=french)


It's difficult to choose because they don't act the same way, they both have pros and cons... For example, I think it's impossible to set them at the same level of details.
Personnaly I think I still prefer the high prest because you have to use high values with Shiandow's algo to beat it and when you do that, details become too blurry to my taste.


How much better (roughly)? Which GPU?


Actually, I was wrong they are mostly the same : 19.80ms vs 19.60ms but before I had for example 1 frame drop every 7 hours and now it's more like 1 frame drop every 1 day. It's just the counter, I didn't experience any frame dropped. By the way, I have a Geforce GTX 970.

My preferences haven’t changed regarding debanding algorithms. I prefer the “high” preset, with the trade quality option “don’t analyze gradient angles for debanding” selected, over the latest Shiandow’s script. Without the trade quality option selected, I see a slight blurring in the images. It may be due to stronger debanding.


Can you try to set the Maxangle at 0.10 or even 0.08?
At 0.08 I don't think that details are lost but the debanding is still stronger than with the old high preset.

Old high preset : [img=http://s5.postimg.org/as4cd1c3r/oldhigh5.png] (http://postimage.org/index.php?lang=french)

[img=http://s5.postimg.org/fs1skzhqf/oldhigh3.png] (http://postimage.org/index.php?lang=french)


Maxangle at 0.10 : [img=http://s5.postimg.org/qzamt3j47/high5010.png] (http://postimage.org/index.php?lang=french)

[img=http://s5.postimg.org/kkblwfcef/high3010.png] (http://postimage.org/index.php?lang=french)


Maxangle at 0.08 : [img=http://s5.postimg.org/b2buwdqpz/high5008.png] (http://postimage.org/index.php?lang=french)

[img=http://s5.postimg.org/inaae3wcn/high3008.png] (http://postimage.org/index.php?lang=french)
[url=http://postimage.org/index.php?lang=french]

[shaolin95]

madshi wants to focus on debanding effects/comparison at the moment, so he may not be to interested in answering that question for you. Chroma's SuperRes option effects the sharpness, anti-alaising, softness, and anti-ringing chroma. Most people have a hard time seeing differences. From full screen video check and uncheck it clicking apply each time to see if you can notice a difference. If you don't then you might as well leave it unchecked and save resources for other features. NNEDI3 is very resource demanding, and many people (myself included) see better performance and visual quality when they combine a lesser chroma scaler with ChromaSuperRes then they do with NNEDI3. Your opinion may be the same or differ, so experiment and make your own conclusion.

Thanks for the information.
Is there an easy way to compare the effects of each chroma scaler, like particular images that make it easier to compare what works best for system and or preferences?

I was using Lanczos before because I cannot run NEDDI3 AND use SVP at the same time as it makes my system lag but lately I have been debating the SVP artifacts vs the inherent issues with non "smoothing" the movies so I have been watching without SVP as well and at that time, I decided to go back to NEDDI32 but I never did any real tests to see which one was best.

Thanks again!

[6233638]

I'm not sure I agree. Can you show a difference in screenshots?The issue is that I just don't have much in the way of 4K test patterns.
That's why I'm hesitant to see such a change made right now.

But as I said: if this only applies to downscaling, and using linear light downscaling forces the old behavior anyway, I don't really have any objections. It seems like a smart way to improve performance of 4K playback on non-4K displays.

To be fair, I also used Lanczos 8 quite a lot, but ONLY because we didn't have a native (fast and max. quality) and mild sharpening filter within madVR.

The results with Lanczos 8 and AR were the sharpest of them all and still relatively cheap performance-wise, without ever having to invoke NNEDI3. But it was never really used other than to just sharpen content by me.

As such, I would not really miss it though.
EDIT: madVR now has sharpening, need to retest.Yes, I think the sharpening options do eliminate any reason that someone might try to use something like Lanczos8, which is why I am not opposed to its removal.

Here's an old example from when anti-ringing was first introduced. (http://www.abload.de/img/rippleezlw1.jpg)
On the left is Lanczos 3/AR
On the right is Lanczos 8/AR

You can see that while the anti-ringing filter removes most of the ringing from Lanczos 8, it does not remove the distortion that the ringing has introduced on the right edge of that red circle.

We have come a long way since then however. NNEDI3 in particular makes a huge difference with that low resolution, low quality source.

http://abload.de/img/nnedifurft.png

Lanczos3 AR on top, NNEDI3 on the bottom.
Sure it can look a little "smoothed over" or artificial in places, but that's just the nature of the source being so low quality.
Though NNEDI3 and other options such as debanding have been in there for quite some time now, it still impresses me with just how much of an improvement madVR can make to low quality, low resolution sources.

So I only get the nice smooth image with MPC/MADVR but that still leaves the question open about why it gets smooth if Nvidia Control Panel is not giving me the option for 10bcp at all.If you are using an HDMI output, NVIDIA only seems to have 8-bit or 12-bit available via the Control Panel.
It seems that you need DisplayPort (and possibly DVI?) to select a 10-bit output there.
However madVR seems to be able to output 10-bit over HDMI via the DX11 presentation path if your display supports it, even though NVIDIA do not make it available via the Control Panel.

[Deim0s]

madshi,
If not copy the folder "legal stuff", an error occurs "OpenCL compiler error".
This folder is needed now?

[huhn]

If you are using an HDMI output, NVIDIA only seems to have 8-bit or 12-bit available via the Control Panel.
It seems that you need DisplayPort (and possibly DVI?) to select a 10-bit output there.
However madVR seems to be able to output 10-bit over HDMI via the DX11 presentation path if your display supports it, even though NVIDIA do not make it available via the Control Panel.

madVR isn't output anything to a screen the GPU driver is doing this.

the best way to find this out is reading the EDID.

One extra piece of information. When I do the same ramp 10 bit test with MPDN instead of MPC with MADVR, I cannot get the "banding" to show smooth even when switching to full screen in 10 bit or higher. It just looks 8 bit and clearly visible bands.
So I only get the nice smooth image with MPC/MADVR but that still leaves the question open about why it gets smooth if Nvidia Control Panel is not giving me the option for 10bcp at all.

A user just PMed me telling me he sees the 10bcp option at 24Hz but not at 60Hz. I tried yesterday at 48Hz so I will check. Perhaps somehow my project is still capable of 10bcp after all.

As a side not, this latest Nvidia driver seem to mess up the TV levels or at least it was making the projector detected Auto incorrectly after the upgrade. No other setting was changed.

In any case, more testing to come later tonight.


BTW, when in the Chroma options I am selecting the NEDDI3 32 setting but what is the SuperRes option in there supposed to do? I know it increases the performance hit but not sure what I am gaining.

Thanks and sorry for the many Qs!

you don't have an old amd card flying around didn't you?

it's way easier to test with AMD cards thanks to the disabling dither trick.

AMD is for sure dithering 10 bit input to 8 bit output if 8 bit is the only possibility or is selected.

[shaolin95]

If you are using an HDMI output, NVIDIA only seems to have 8-bit or 12-bit available via the Control Panel.
It seems that you need DisplayPort (and possibly DVI?) to select a 10-bit output there.
However madVR seems to be able to output 10-bit over HDMI via the DX11 presentation path if your display supports it, even though NVIDIA do not make it available via the Control Panel.

That will explain why I get the clear improvement in the ramp image when on 10bit with madvr even though Nvidia does not give me that option.
Tonight I will try 24Hz as someone told me they get the 10bcp option at 24Hz but not at 60Hz for some odd reason.

I appreciate your help. there is a TON to learn for me yet. :)

[MS-DOS]

Ah, thanks! How about NVidia and Intel?


I asked (http://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU?pid=4007#pid4007) ToastyX (the developer of CRU) about Nvidia a while ago:
No, not in Windows. They dither in Mac OS X, but I haven't found a way to enable dithering in Windows.
He also was the one to provide the info about dithering on AMD in the first place (http://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU?pid=3314#pid3314).

[krille]

Just a question about plasma screens. Should dithering in madvr be enabled ,even though plasmas does dithering by its own?

[huhn]

i did the test with nvidia and nvidia is for 100% dithering.

first test. 6 bit tn asus vg248qe. nvidia can only send 8 bit to this screen.

8 bit madVR without dithering shows a ton of banding.
d3d11 10 bit FSE doesn't show banding.

TV with 12 bit input support disabled madVR dithering.
8 bit window mode banding.
d3d11 10 bit FSE 12 bit nvidia no banding
d3d11 10 bit FSE 8 bit nvidia no banding

Just a question about plasma screens. Should dithering in madvr be enabled ,even though plasmas does dithering by its own?

yes 100%

[Xaurus]

That doesn't look like v0.88.10 to me. Are you sure you really updated? Maybe there are more than one madVR installations on your PC? E.g. JRiver MC comes with its own installation.
It sure is 88.10.

http://home.halden.net/mordor/madvr.jpg

However I figured out what the problem was. I had to extract the two folders from the .zip as well (legal stuff and developers), until now I've not bothered about extracting those. Now it works fine. I didn't think madvr needed anything from those folders to function.

Thanks.

[leeperry]

I don't remember what you said or I said ages ago. The switching between FSE and windowed mode was improved a while ago. With Intel GPUs there are still some weird things going on for some users, when using DX11 presentation or "use a separate device for presentation". But that's limited to Intel.
Well least year this seemed to be a known issue (http://forum.doom9.org/showpost.php?p=1664865&postcount=22140):
So is there any way to avoid the black frames flashing during FSE/FSW transitions? This is happening whenever I want to open a context menu and that never occured on XP.
I've only ever used madVR on Windows 7/8, and this has always been the case. I was not aware that XP didn't do this.
XP was a bit faster and more seamless when doing FSE/FSW transitions. There's nothing we can do about that. Well, maybe I could find a way (or maybe not), but I consider that a cosmetical problem and as such not very important at the moment.
Using 88.10 and the 13.12 drivers I do get a pretty long black frame flash when rolling DX9 FSE/FSW or DX9 FS/windowed with Aero disabled in W7SP1 and if I enable the latter I randomly get previous frames from the movie or black frames.

I remember ppl complaining about the random previous frames a while ago but I was on XP at the time so I didn't pay much attention.

I take it that the black frame without Aero and random previous frames wirh Aero are known non-urgent cosmetical problems?

[Vyral]

I don't know if it's intended or not, but the OSD show D3D9 fullscreen windowed mode even though I checked "use Direct3D 11 for presentation".
I haven't checked "present a frame for every sync" (it's for exclusive mode if I remember correctly) and native display bitdepth is 8 bit.

[madshi]

Im experiancing a short black screen when switching from FSE to windowed mode in mpchc in d11 mode. It switches instantly in d9 mode. Is this intended?
Every GPU/driver behaves differently there. I've done my best to make the switch as fast as possible. It should be faster with 8bit output compared to 10bit. But there's probably nothing I can do to make it faster than it currently is.

Same here, along with the high present stats in D3D11 windowed mode.
If you have high present stats, try lowering the number of prepresented frames.

I've done a lot of comparisons. Shiandow's algo does a very good job at debanding but the high preset blurs less details... Well it's not that simple. Shiandow's algo preserves more the shape of the original picture but blurs the details.

But if I use high values it can beat sometimes the high preset.

It's difficult to choose because they don't act the same way, they both have pros and cons... For example, I think it's impossible to set them at the same level of details.
Personnaly I think I still prefer the high prest because you have to use high values with Shiandow's algo to beat it and when you do that, details become too blurry to my taste.
Thanks for your tests and feedback!

The issue is that I just don't have much in the way of 4K test patterns.
You don't need 4K test patterns. I don't do anything special for 4K content. What I'm doing solely depends on the downscaling factor and nothing else. So you can take any Blu-Ray, downscale it just enough to have madVR show "luma <" in the OSD instead of "image <", and then compare the image quality to v0.88.8.

But as I said: if this only applies to downscaling, and using linear light downscaling forces the old behavior anyway, I don't really have any objections.
However, you do think that it might hurt image quality. And that is a concern, so I'd still welcome if you could do a quick test to see if you can find a situation where image quality is lost in v0.88.10 compared to v0.88.8, when downscaling just enough to activate the new rendering path in v0.88.10 (without linear light downscaling, obviously).

You can see that while the anti-ringing filter removes most of the ringing from Lanczos 8, it does not remove the distortion that the ringing has introduced on the right edge of that red circle.

We have come a long way since then however. NNEDI3 in particular makes a huge difference with that low resolution, low quality source.

Lanczos3 AR on top, NNEDI3 on the bottom.
Sure it can look a little "smoothed over" or artificial in places, but that's just the nature of the source being so low quality.
Though NNEDI3 and other options such as debanding have been in there for quite some time now, it still impresses me with just how much of an improvement madVR can make to low quality, low resolution sources.
You might like the chroma SuperRes processing, too, with the right settings. But that's for later. For now I would really like to finish up the debanding discussion, so we can move on to discuss the next new feature.

If not copy the folder "legal stuff", an error occurs "OpenCL compiler error".
This folder is needed now?
Oh well. I changed this to satisfy LGPL requirements. But I guess I could change it once more to not require the folder. Will put that on my to do list.

i did the test with nvidia and nvidia is for 100% dithering.

first test. 6 bit tn asus vg248qe. nvidia can only send 8 bit to this screen.

8 bit madVR without dithering shows a ton of banding.
d3d11 10 bit FSE doesn't show banding.
Same here.

I've tried to find any API that would tell me whether the GPU actually outputs 10bit or not, but I've not found anything, at least not nothing official GPU independent from Microsoft. I can see that Direct3D11 actually tells the GPU driver to output 10bit, and the driver accepts it, although I know for a fact that my display doesn't support 10bit. So basically it's a black box and I don't know what the GPU is doing with my 10bit output... :(

It's good that we have the dither registry tweaks for AMD. Wish we had something like that for NVidia and Intel, too.

It sure is 88.10.

However I figured out what the problem was. I had to extract the two folders from the .zip as well (legal stuff and developers), until now I've not bothered about extracting those. Now it works fine. I didn't think madvr needed anything from those folders to function.
Ah, makes sense. I'll fix that for the next build.

Using 88.10 and the 13.12 drivers I do get a pretty long black frame flash when rolling DX9 FSE/FSW or DX9 FS/windowed with Aero disabled in W7SP1 and if I enable the latter I randomly get previous frames from the movie or black frames.

I remember ppl complaining about the random previous frames a while ago but I was on XP at the time so I didn't pay much attention.

I take it that the black frame without Aero and random previous frames wirh Aero are known non-urgent cosmetical problems?
It's not so bad here, at least with Windows 8.1. There's no improvement in that area planned, not sure if there is anything I can do. Maybe the issue only occurs on Windows 7. I don't know if what you're seeing is the same what I'm seeing.

I don't know if it's intended or not, but the OSD show D3D9 fullscreen windowed mode even though I checked "use Direct3D 11 for presentation".
I haven't checked "present a frame for every sync" (it's for exclusive mode if I remember correctly) and native display bitdepth is 8 bit.
Which OS are you on? Do you have the Windows 7 platform update installed?

[Vyral]

Which OS are you on? Do you have the Windows 7 platform update installed?

I'm on Windows 7.
I'm not sure Windows 7 platform update is installed. Is it the KB2670838 update ? If it is, then yes I haven't installed it yet.

EDIT : Yep, it was this update. It's ok now. Thanks.

[leeperry]

It's not so bad here, at least with Windows 8.1. There's no improvement in that area planned, not sure if there is anything I can do. Maybe the issue only occurs on Windows 7. I don't know if what you're seeing is the same what I'm seeing.
I could try to make a video but I guess it's just like you said: XP was a lot snappier....pretty ironic huh.

Basically I get a black frame for quite a while during every transition and with Aero enabled I get a random frame from the movie between the black frames, which is a lot more annoying :o

So on W8 you do get the slow updating black frame but no previous frame from the movie? The latter comes with Aero on W7SP1, I would enjoy a clear confirmation whether upgrading to W8/W10 would greet me with forced previous frames during transitions as Aero can't be disabled anymore.

[Siso]

If you have high present stats, try lowering the number of prepresented frames.


I've tried with different numbers and nothing changed, even it got worse with lower and higher numbers...

[madshi]

I'm on Windows 7.
I'm not sure Windows 7 platform update is installed. Is it the KB2670838 update ? If it is, then yes I haven't installed it yet.
I'm not sure about the KB, but the Windows 7 platform update is needed for madVR's DX11 rendering option.

Basically I get a black frame for quite a while during every transition and with Aero enabled I get a random frame from the movie between the black frames, which is a lot more annoying :o

So on W8 you do get the slow updating black frame but no previous frame from the movie? The latter comes with Aero on W7SP1, I would enjoy a clear confirmation whether upgrading to W8/W10 would greet me with forced previous frames during transitions as Aero can't be disabled anymore.
I don't seem to have this problem here with Windows 8.1 and AMD. Can't guarantee that there will not be any problem for you, either, though, if you update to Windows 8.1. Have seen too many weird problems from different users. So I don't feel safe predicting what your experience will be with Windows 8.1/10.

I've tried with different numbers and nothing changed, even it got worse with lower and higher numbers...
How high are your present numbers exactly? Anything lower than 1ms should be ok.

[Siso]

How high are your present numbers exactly? Anything lower than 1ms should be ok.

With D3D9 fullscreen windowed new path they are: 0.12 ms

with D3D11 they are: around 0.55 ms

[6233638]

You don't need 4K test patterns. I don't do anything special for 4K content. What I'm doing solely depends on the downscaling factor and nothing else. So you can take any Blu-Ray, downscale it just enough to have madVR show "luma <" in the OSD instead of "image <", and then compare the image quality to v0.88.8.That's true, though it can be difficult to test at that sort of scale. It would be easier to test with 4K scaled down to a full-screen image.
I will have a look into it, but from some initial testing it doesn't seem to be problematic with all of the scaling/processing options we have available to us now.

Edit: Just to be clear though, if chroma is being upscaled, it is always being upscaled 2x, yes? (e.g. 720p > 1080p)


I did manage to get some testing done at 4K (though not with proper test patterns) but my system doesn't really handle it that well.
After changing the CPU priority of the player and adjusting the queue sizes, I was able to get smooth 4K video playback though.

I set up some new profiles to change the dither algorithm from ordered dither to random dither at 4K, and some of the other render/queue/trade quality for performance options had to be adjusted as well, which worked great.
I'm very happy with how flexible the profile system is.

This had me wondering if there was a specific reason that some settings are locked out of profiles.
For example, I had to drop the present queue size down to 6 from 8 for 4K, and I'd prefer to set that with a profile instead of it being set globally.

And though it's not specific to 4K playback, I'd really like to set up profiles that change the calibration and color & gamma settings based on the source content.
I'd like to use 2.40 gamma for 480/576/1080p content (movies) and 2.20 gamma for all other source resolutions.
That's a setting I find myself changing quite often depending on the source content, so it would be very convenient to have that switch automatically.

[flashmozzg]

Edit: Just to be clear though, if chroma is being upscaled, it is always being upscaled 2x, yes? (e.g. 720p > 1080p)

wouldn't x2 of 720p be 1440p?