madVR - high quality video renderer (GPU assisted)

[burfadel]

Originally Posted by madshi

NGU pixart 3 is now usually sharper than NNEDI3, see e.g. Castle or AmericanDad images. So I'm wondering why you still want more sharpening for NGU pixart, while you seem to be happy without additional sharpening for NNEDI3?

Thanks...

The Pixart3 is the best of the lot overall, although the old pixart is close. It is better than NNEDI3.

So maybe a half way point between pixart 3 and the old pixart, if possible? For the sharpness I don't mean a strong sharpness, adaptive sharpness, or anything like that, I just mean a basic sharpen pass at the end to clear up the picture a bit without the need for further sharpening. In fact, the type of sharpening I'm referring to isn't an option under the post-process sharpen methods.

If you open the pixart3 images in Irfanview, available here:
http://www.irfanview.com/

and select image from the menu and sharpen, then pixart3 is the best out of all of them. Like I said it is only a subtle sharpen, but makes a massive difference. It is so effective that for the potatoes it actually makes pixart3 better than NGU. With the sharpen, the only image that looks a bit better is the outdoor kitchen scene with the old pixart, with the barbecue aliasing.

So it doesn't need a adaptive sharpen, edge sharpen or anything like that, just subtle sharpening using the most basic filters as a final pass, and if that can be done Pixart3 would be a real and clear winner.

[ryrynz]

Originally Posted by madshi

In your suzumiya image I clearly prefer NGU pixart 3 over NNEDI3, don't you?

I still prefer NNEDI3 64 neurons or higher for the sole reason that NNEDI3 is sharpening more lines within the image and the lines are darker. NGU newpix 2/3 thins lines a touch more but it's not sharper on the suzumiya image (representative of low res anime content) and NNEDI3 128 and 256 are even sharper still and are superior (and IMO probably impossible to beat) on this image. If NGU newpix 3 picked up on stuff other than just thick lines it probably would be a win. But I see a big difference still between NNEDI3 256 and NGU newpix 3. Sure the performance requirement for NNEDI3 is high but like I said, I think in order to topple NNEDI3 you need to rework the core of NGU.. for some NGU is good enough, and yes NGU newpix in some areas works better than NNEDI3 as you pointed out to huhn but it's not quite there for me., but nice improvement none the less..

[madshi]

Originally Posted by burfadel

The Pixart3 is the best of the lot overall, although the old pixart is close. It is better than NNEDI3.

So maybe a half way point between pixart 3 and the old pixart, if possible? For the sharpness I don't mean a strong sharpness, adaptive sharpness, or anything like that, I just mean a basic sharpen pass at the end to clear up the picture a bit without the need for further sharpening. In fact, the type of sharpening I'm referring to isn't an option under the post-process sharpen methods.

If you open the pixart3 images in Irfanview, available here:
http://www.irfanview.com/

and select image from the menu and sharpen, then pixart3 is the best out of all of them. Like I said it is only a subtle sharpen, but makes a massive difference. It is so effective that for the potatoes it actually makes pixart3 better than NGU. With the sharpen, the only image that looks a bit better is the outdoor kitchen scene with the old pixart, with the barbecue aliasing.

So it doesn't need a adaptive sharpen, edge sharpen or anything like that, just subtle sharpening using the most basic filters as a final pass, and if that can be done Pixart3 would be a real and clear winner.

The pixart2 build was between old and pixart3, though much nearer to pixart3. I'm not sure what irfanview's sharpening does, probably it's roughly similar to LumaSharpen in madVR.

Originally Posted by ryrynz

I still prefer NNEDI3 64 neurons or higher for the sole reason that NNEDI3 is sharpening more lines within the image and the lines are darker. NGU newpix 2/3 thins lines a touch more but it's not sharper on the suzumiya image (representative of low res anime content) and NNEDI3 128 and 256 are even sharper still and are superior (and IMO probably impossible to beat) on this image.

I'm confused. Where exactly do you see NNEDI3-256 to be better on the suzumiya image compared to pixart 3? Here's a direct comparison:

http://screenshotcomparison.com/comparison/201202

I've marked places where I could see a difference at 100% view. I've marked the differences in the NNEDI3 image. Red circles are where NGU pixart 3 looks better to me. Green circles is where NNEDI3 looks better to me. Gray circles are where I could see a difference but couldn't say which one was better. This comparison is with NNEDI3-256, btw. You must be disagreeing with my judgement here? Did I miss to find some circles? And which circles do you disagree with?

Originally Posted by ryrynz

If NGU newpix 3 picked up on stuff other than just thick lines it probably would be a win. But I see a big difference still between NNEDI3 256 and NGU newpix 3.

But where!?!? I don't see NNEDI3 picking up anything NGU pixart 3 doesn't. And IMHO NGU pixart 3 looks better on the suzumiya image. But I'm willing to change my mind, if you can point me to where you see NNEDI3 win with a big difference?

[Werewolfy]

Originally Posted by madshi

Here's a new NGU pixart test build:

http://madshi.net/nguPixArt.zip

Actually, the zip contains 2 different test builds. Both builds are almost identical to each other and to the previous build with only one key difference: The NGU "very high" algorithm has been tuned. The "low", "medium" and "high" algorithms haven't changed at all. So if you want to test the new builds, please use "very high" for all your tests.

The new builds have slightly different priorities to before: The old NGU pixart test build put highest priority on anti-aliasing capabilities, which proved to be working well. However, it resulted in some artifacts sometimes, and in results maybe not always looking very natural. The tuned NGU pixart "very high" algo now compromises a little bit. I've traded in a bit of anti-aliasing capability for more natural looking results and more sharpness. The 2 new builds are very near to each other. One build has ever so slightly higher anti-aliasing abilities, the other one is ever so slightly sharper.

Included in the zip is also a little image quality comparison, with screenshots taken with: NNEDI3-256, standard NGU very high, the original NGU pixart very high, and the 2 new NGU very high variants. Please check out the images and let me know which algorithm you prefer. Please also double check with your own content to verify your preferences.

Here's my own subjective quality evaluation, based on the screenshot images in the zip:

anti-aliasing abilities: NGU-pixart-old > NGU-pixart-2 > NGU-pixart-3 > NNEDI3-256 (see kitchen and americanDad images)
natural look: NGU-pixart-3 > NGU-pixart-2 > NNEDI3-256 > NGU-pixart-old (see potatoes and suzumiya images)
sharpness: NGU-pixart-3 > NGU-pixart-2 > NNEDI3-256 > NGU-pixart-old (see castle and americanDad images)

Overall, I prefer NGU pixart 2/3 over the other options, but I can't decide if I prefer pixart 2 or 3. With some images I prefer 2, with some 3.

What are your thoughts?

Only based on your comparison pictures : I prefer pixart 2, it seems thats this one has better anti-aliasing abilities on the aliasedAnime image. 3 is just a little bit sharper on some pictures but it has more aliasing.

Pixart beats NNEDI3 on most of your pictures, I'll try to test on some Anime your test build to see if that's still the case on my sources.

[burfadel]

If pixart2 is between pixart 3 and old, then pixart3 is the winner, however a sharpen pass will help tremendously. I believe the sharpen of irfanview is really basic, like sharpen in avisynth where it sharpens the whole image without using edge detection etc. The nature of the picture with pixart3 lends itself to responding really well to it. Surely a clarifying pass with a basic sharpen like that would be okay to do?

Seems the only complaint with pixart is the sharpness, and a basic sharpen like in Irfanview fixes that. What do others think?

[Damien147]

Originally Posted by tranfa

Thank you Madshi.

1) Done, I seem to have a similar bug to the one described previously on my Panasonic AX900, where the Win10 Ati Crimson driver recommends 4096x2160. But anyway, let's stick to 2160p, and see what happens.

2) Done. The issue is still there, introducing a 2160p29 or 30 puts in offside the p25...

3) How can I activate debug mode? I renamed the .ax files, I run the .bat file, and I get the error "file "madvr.ax" is missing" - but it's there...

Thank you for the patience...

I reproduced your problem and by adding p50 solved it.



EDIT:So your resolution is 3840x2160?Then try :

3840x2160p24, 3840x2160p25, 3840x2160p29, 3840x2160p30, 3840x2160p50

or

2160p24, 2160p25, 2160p29, 2160p30, 2160p50

[wasper]

This is it, my first post.

First, I had so much fun trying to understand all the MADVR setting even if my wife just doesn't understand why!!! I'm reading many posts for a while and it's really great to see how smart people could be. Well, enough a**-kissing.

Still, I'm having some questions that I wasn't able to figure out.

I'm using MediaCenter with Intel NUC. Mainly, I'm upscaling DVD to 1080p as my gpu/cpu is not able to step directly to 4k with double frame rate produced by DXVA deintrelacing. The TV has a correct upscaler and do the rest. It's kind of a two times upscaling process. Don't know or not if I should avoid this.

Anyway, my question is about color space. If I'm using film mode, the color space seems to work fine for all the image upsacler available (dxva, jinc, spline, etc...). In video mode, color space is fine only if I'm using DXVA. As soon, as I'm changing to software upscaler, the color space seems to get compressed between 16-235 even though everything is set to 0-255(gpu, madvr, tv). Is this a bug or I'm missing something? I have an old DVD movie in 4:3 native but shooted in 16:9. The black bar in the movie are totally black in film mode but grey in video mode if using dxva deinterlacer without dxva image upscaler. To resume, color space seems wrong if I'm using deinterlacing with CPU image upscaler.

Second question is about IVTC. How does a dvd player or a blu-ray will play a film telecined to 29.97? Will it deintrelace like video mode and play it at 29.97p or it will use kind of ITVC to play it back at 24fps (probably not as tv must accept this refresh rate)?

I'm having a DVD that will play ok in video mode but with some jitter on some sequences. It will not jitter on film mode. However, with my hardware blu-ray player, it is also playing fine... Could the jitter in video mode related to some clock deviation that are not present in film mode and totally inexistant on hardware player? I'm not using reclock or videoclock. Smooth motion is also not used for this DVD.

Sorry about the bad english

[Bloodred]

Originally Posted by madshi

Now can you guys please confirm if you like pixart 3 better than NNEDI3 or not? You know, I'd like to get rid of NNEDI3 at some point, so it's an important question for me.

I'd be fine with that, I like the look of pixart 3.

Originally Posted by madshi

However, for 4x upscaling there'd be the option of doing NGU pixart doubling, followed by standard NGU doubling. That would give you higher sharpness than doing NGU pixart doubling twice.

That actually sounds very interesting, I'd really like to see how that turns out since in my experience regular NGU had the most visible issues when quadrupling.

[cork_OS]

Originally Posted by madshi

Now can you guys please confirm if you like pixart 3 better than NNEDI3 or not? You know, I'd like to get rid of NNEDI3 at some point, so it's an important question for me.

I don't need NNEDI3, and NGU pix 3 looks good. But, just to be honest, NNEDI3 32 vs. NGU pix medium tweaked may have different relation.

[huhn]

Originally Posted by wasper

I'm using MediaCenter with Intel NUC. Mainly, I'm upscaling DVD to 1080p as my gpu/cpu is not able to step directly to 4k with double frame rate produced by DXVA deintrelacing. The TV has a correct upscaler and do the rest. It's kind of a two times upscaling process. Don't know or not if I should avoid this.

in term of quality it would be better to output at UHD but i don't think a nuc can do that.

Anyway, my question is about color space. If I'm using film mode, the color space seems to work fine for all the image upsacler available (dxva, jinc, spline, etc...). In video mode, color space is fine only if I'm using DXVA. As soon, as I'm changing to software upscaler, the color space seems to get compressed between 16-235 even though everything is set to 0-255(gpu, madvr, tv). Is this a bug or I'm missing something? I have an old DVD movie in 4:3 native but shooted in 16:9. The black bar in the movie are totally black in film mode but grey in video mode if using dxva deinterlacer without dxva image upscaler. To resume, color space seems wrong if I'm using deinterlacing with CPU image upscaler.

check the gpu driver video settings.

Second question is about IVTC. How does a dvd player or a blu-ray will play a film telecined to 29.97? Will it deintrelace like video mode and play it at 29.97p or it will use kind of ITVC to play it back at 24fps (probably not as tv must accept this refresh rate)?

depends on the player setting and the DVD.
a soft telecine DVD and a proper player can IVTC the signal easily.

I'm having a DVD that will play ok in video mode but with some jitter on some sequences. It will not jitter on film mode. However, with my hardware blu-ray player, it is also playing fine... Could the jitter in video mode related to some clock deviation that are not present in film mode and totally inexistant on hardware player? I'm not using reclock or videoclock. Smooth motion is also not used for this DVD.

and we don't know if you are changing the refresh rate to match the source refresh rate.
hard to say.

[Magik Mark]

Madshi,

We welcome innovation on image fidelity. I have been an avid Nnedi3 user. I watch a lot of HD sources. So for, nothing beats Nnedi3 64 chroma upscaling.

Comparing it to NGU, all variants, Nnedi3 still sharper, finer and depthful. Image may appear a little bit smaller because of finer lines.

It is more natural and faithful to the original.

I think what NGU lacks at the moment is depth. Maybe we can achieve this by making the lines finer

[Arm3nian]

I like pixart 3.

I also tried pixart 3 with luma sharpen in upscaling refinement, and think it looks pretty good. Gives added sharpness without the nasty artifacts that regular NGU produces with low quality sources, and does better than using luma sharpen with nnedi3. Maybe those who wanted a sharpness slider on ngu pixart can use this as an alternative.

[burfadel]

From what I can find, the sharpen as used in Irfanview applies unsharp sharpening to the image. Unsharp sharpenining is also available from the effects list (ctrl + E). Unsharp sharpening to 1 (weak) produces nice results on all the pixart3 images. Using limitedsharpen or adaptivesharpen won't be as 'good' for this blurring because they operate on different principles. For those that don't know what unsharp is, google it . If looking at images just keep in mind the strength of the filter they used, I'm suggesting it should be used with a weak setting.

Since the upscaling amount is a fixed amount (2x or 4x), having it built in with a fixed value wouldn't be a bad thing as the strength of the unsharp filter required would always remain the same. The purpose of this sharpen would not be to actively sharpen the image, it would be just to remove the softness introduced by the upscaling.

Pixart3 is preferred over the original pixart because it may be fractionally clearer, but like in the case of the outdoor barbecue handle it has bad aliasing. That type of aliasing isn't noticeable in the other pics, but I can think of many scenarios where the aliasing would be noticeable. For example, maybe car racing? The old pixart was far better with antialiasing.

Now once unsharp (weak) is used, the old pixart is the winner because it is sharper and clearer than the others without the aliasing!

So my new choice, which wasn't an option is pixartold in all cases, with a weak unsharp filter.

[ryrynz]

Should 704x480 upscaled to 1920x1200 when using NGU, show the following? NGU very high > NGU High < SSIM1D100 AR? Just wondering why there's downscaling if its supposed to be doubling.. I've reset madVR to defaults. Is the OSD correct?

[nevcairiel]

Originally Posted by ryrynz

Should 704x480 upscaled to 1920x1200 when using NGU, show the following? NGU very high > NGU High < SSIM1D100 AR? Just wondering why there's downscaling if its supposed to be doubling.. I've reset madVR to defaults. Is the OSD correct?

Its quadrupling (or doubling twice) and then scaling down to target, seems perfectly normal to me.

[ryrynz]

Ah so it's all baked into the presets. Anything medium and above is actually quadrupling..

Madshi, when looking at the doubled images of Suzumiya it's near impossible to spot any differences between them and Beyond Compare also backs this up with very little changes so I won't refer to those examples.
What I see on my own screen is quite different, what screen res do you use? Here's another comparison below.

This one is from the anime Full Metal Panic! and shows the levels of sharpness difference between NNEDI3 256 and NGU newpix3, screencapped at 1920x1200
It doesn't really show much else with with regards to handling aliasing or blurry lines etc but you'll see all the lines here are much more in focus, especially the eyes and the radio, knobs etc. with no detriment to the image quality that sharpening afterwards would usually add in the form of haloing.

This is what I see consistently when comparing the two. I've also noticed that using screenshotcomparison doesn't seem to show the exact level of change visible when viewing fullscreen.. so would recommend the ol' F11 on it.

In this case after using sharpen edges refinement of 1.5 after upscale gives a somewhat similar image, but even after that NNEDI3 is still sharper in other background areas. Even at 2.0 the background lines are still sharper using NNEDI3 256 and unfortunately sharpen edges is now oversharpening some lines now which doesn't make it a worthwhile alternative choice to NNEDI3 for anime IMO.

With the aliased anime examples I still prefer NNEDI3 overall, strangely NGU newpix 3 is sharper here but it's kinda to the detriment of the image, with it looking overall too sharp, especially around the mouth and glasses predominately.

[pose]

Originally Posted by ryrynz

This one is from the anime Full Metal Panic!

Sup with the difference in color/brightness? Can i have source image please?

[ryrynz]

Originally Posted by pose

Sup with the difference in color/brightness? Can i have source image please?

There isn't really anything really to speak of..Probably due to the upscaler differences + chroma and maybe your monitor?

http://www.filedropper.com/theguyikindalikeisasergeant Use the first frame for comparison.

[Cinemancave]

So I have a really strange problem that only occurs when playing 4k-material: My GTX 1080 doesn't clock up to its normal reference clock speed off around 1700 MHz, but stays at 800 MHz. This makes playback off 4K stuttery. When I playback 1080p upscaled to 4K the clock speed go up to 1700 MHz, as they should. I can't get my head around what's causing this :S

Does anyone know how to force clock speed on the Geforce 10x0-series?

[huhn]

change the power setting in the nvidia control panel to adaptive not optimal