madVR - high quality video renderer (GPU assisted)

[AngelGraves13]

Originally Posted by nevcairiel

Before image upscaling would be best, not sure about chroma though.
Image upscaling could otherwise create new artifacts around the block edges (or smooth them, depending on the scaler) which makes a deblock filters job ever so much harder.

That's what I was thinking too.

I wonder if the debanding is done before as well.

[huhn]

Originally Posted by AngelGraves13

That's what I was thinking too.

I wonder if the debanding is done before as well.

it is.

[Danne S]

Is there any current issues between madvr and nvidia?
It was some time since I updated the drivers...

Skickat från min GT-N7000 via Tapatalk

[aufkrawall]

afaik, the only issue should be the flawed DX11 windowed fullscreen mode on Windows 10 (most likely driver bug since 352 series).
But there can be different results with different GPU families or Optimus etc.

[Finnish Flash]

Originally Posted by aufkrawall

afaik, the only issue should be the flawed DX11 windowed fullscreen mode on Windows 10 (most likely driver bug since 352 series).
But there can be different results with different GPU families or Optimus etc.

I feel it got much better in the latest release, but then again it can be a placebo effect. At least DXVA deinterlacing performance is now what it's supposed to be in DX11 windowed mode. I'm running the latest WHQL drivers. Great work, madshi! Maximum rendering times seem to be through the roof in some cases (24 @ 60fps with smooth motion), but I don't see that affecting playback.

[kolak]

Originally Posted by madshi

I see banding problems in Blu-Rays quite often, and not limited to Anime. My impression is that authoring studios using dithering is rather the exception than the rule. I often see banding problems in dark scenes, or it's especially noticable in fade-to-blacks. Fade-to-blacks in 8bit without dithering almost always looks ugly and non-smooth.

Sometimes I'm wondering why dithering is often not used by authoring studios. Either they're just dumb. Or maybe they don't want to spend the additional video bitrate needed to encode dithering without getting compression artifacts.


Thanks to the totally different HDR transfer function, I think banding will be less a problem with HDR than it is with current 8bit Blu-Rays, because the transfer function moves the majority of the 10bits to SDR pixels. Still, I'd have loved to get 12bits to get totally rid of any banding once and forever.

Unfortunately you are correct- many authoring studios don't use dithering at all have no clue what to do with it
Been in few big places and it was very disappointing.

I use to always do dithering with avisynth- Floyd+bit of noise and this really helps with banding, specially on fades (as you mentioned).

[fedpul]

Hi, I want to ask something offtopic, related with video cards, anyone with a GTX 970 can answer me. I have recently changed my video card from: Sapphire HD 7870 GHz Edition OC to MSI GTX 970 Gaming. Everything works fine, games doesn't exhibit artifacts nor slowdowns. But I saw something really strange in whit colour images, i see like a vibrance in the form of bars, so I decided to see the Chroma Test and for my surprise i see anoying vibrations everywhere and 4:2:2 can be seen. I test with other test like the colour bars one and nothing seem to be wrong. Maybe it is a defective DVI output, beacuse everything is OK when I use HDMI output to my TV. I am using the provided DVI to VGA adapter. I will aprecciate a lot any answer from GTX 970 owners. Thanks in advance, and sorry for the offtopic question, it is not my intention to deviate the topic.

[markanini]

Bug report: Once the interlaced flag is activated midstream MadVR sees it as interlaced for the remainder of the duration even if the stream switches back to reporting progressive.

[mcn]

Originally Posted by madshi

Can you show a screenshot from one typical situation where subtitles are where you think they should be, and another situation where subs are *not* where they're supposed to be? Please use videos with SRT subtitles for those screenshots. Thanks!

I'm attaching some screenshots regarding the "issue" with subtitles positioning.
The first screenshot is of the original video, 1916x1076, without black bars.
The other two screenshots are taken with the video shown in full screen at 1920x1200, where black bars are added.

Embedded subtitles are partially shown in the bottom black bar, while external ones aren't.
Both should be definitely rendered by XySubFilter since modifying the timing through its properties window affects both subs.

[bozokaydin]

MadVR Option High End PC DXVA2 Alternative Configuration
chroma upscaling NNEDI3 128 Neurons
Image Doubling Always use NNEDI3 to Double Luma and Chroma Resolution using 128 neurons
Image Upscaling DXVA2
Image Downscaling DXVA2
smooth motion Enable Smooth Motion only if judder without
dithering Error Diffusion Option 1
Video Codec CUVID with hardware deinterlacing

I found this suggestion on a forum. In this senario, by using dxva for upscaling and downscaling does image doubling work for example for 720p content? ı am not sure it will work while using dxva.

[huhn]

bilinear is used in this case.

do your self a favour and don't sue DXVA scaling with image doubling and forget chroma doubling.

DXVa scaling should only be used if the GPU can't handle anything else.

[aufkrawall]

Sorry to say, but these settings are simply bizarre (wth invent people stuff like that?).

NNEDI3 128 neurons for chroma are a huge waste of performance, even 32 are quite luxury (with filmed content, NNEDI3 will hardly ever look noticeably better than Jinc AR for chroma).

Always using NNEDI3 for image doubling is also pure nonsense in 99% of cases. Linear scaling algorithms like Jinc introduce more artifacts the higher the scaling factor is. With <1.5x scaling, there usually hardly is an advantage of NNEDI3 over Jinc AR.
And 128 neurons are an insane count and most content doesn't profit much by the additional neurons, compared to 64.

DXVA scaling is simply bad quality wise. Don't do it if you aren't forced by badly slow hardware.
Use Jinc AR and Catmull-Rom AR for downscaling instead.

Don't use Error Diffusion if you can't see a difference for yourself. Instead, stick to normal ordered dithering of madVR.

Don't use CUVID, as it's a blackbox to developers (Nvidia doesn't grant nevcairiel or any other dev any control over it). Instead, use DXVA2 Copyback. It achieves the same quality and speed, but it doesn't have any drawbacks (instead of worse compatibility compared to software decoding, which almost never is an issue).

[iSeries]

Looking at either a R7 260x or an R7 360 to replace my R7 250. They're pretty much the same price. Is either one preferred over the other for MadVR?

[Warner306]

Originally Posted by bozokaydin

MadVR Option High End PC DXVA2 Alternative Configuration
chroma upscaling NNEDI3 128 Neurons
Image Doubling Always use NNEDI3 to Double Luma and Chroma Resolution using 128 neurons
Image Upscaling DXVA2
Image Downscaling DXVA2
smooth motion Enable Smooth Motion only if judder without
dithering Error Diffusion Option 1
Video Codec CUVID with hardware deinterlacing

I found this suggestion on a forum. In this senario, by using dxva for upscaling and downscaling does image doubling work for example for 720p content? ı am not sure it will work while using dxva.

You would be better off using these settings:

• Chroma: super-xbr125 + AR
• Image: Lanczos3 + AR
• Double Luma: 1.5x or greater - NNEDI3 64 Neurons
• Double Chroma: Off
• Downscaling: Catmull-Rom + AR + LL
• Upscaling Refinement: SuperRes (strength: 1, sharpness: 2)
• Artifact Removal - Debanding: Medium/High
• Image Enhancements: Off
• Dithering: Ordered

Image doubling for 720p content can have an advantage when used with SuperRes. I find SuperRes makes fewer errors when used with sharp, artifact-free image scaling such as super-xbr or NNEDI3. The overall image appears sharper.

If you aren't using SuperRes, image doubling won't make much a difference at a 1.5x scaling factor.

[wanezhiling]

No, R7 series are all weak

[huhn]

Originally Posted by iSeries

Looking at either a R7 260x or an R7 360 to replace my R7 250. They're pretty much the same price. Is either one preferred over the other for MadVR?

the r7 360 is usually a bit faster thanks to the higher core clock of 50-100 mhz.

looks like they are both the same chip with a different name.

both have about double the speed of the r7 250

[XRyche]

Originally Posted by wanezhiling

No, R7 series are all weak

My MSI R7 265 is not bad either especially for it's price point. It's like having a R9 270. I know it is by no means top of the line but it does a decent job with madVR and NNEDI3 Image doubling.

[iSeries]

Seems I can get an R7 265 for the same price as ether an R7 260x or R7 360. It has more shaders, twice as many ROPs and greater memory bandwidth but a lower clock speed. What's more important for MadVR?

[aufkrawall]

In practice, the R7 265 is quite faster than the other mentioned graphics cards.
It's usually always better to go with the bigger GPU if they are more or less the same generation as the smaller ones.

[AngelGraves13]

Originally Posted by Warner306

You would be better off using these settings:

• Chroma: super-xbr125 + AR
• Image: Lanczos3 + AR
• Double Luma: 1.5x or greater - NNEDI3 64 Neurons
• Double Chroma: Off
• Downscaling: Catmull-Rom + AR + LL
• Upscaling Refinement: SuperRes (strength: 1, sharpness: 2)
• Artifact Removal - Debanding: Medium/High
• Image Enhancements: Off
• Dithering: Ordered

Image doubling for 720p content can have an advantage when used with SuperRes. I find SuperRes makes fewer errors when used with sharp, artifact-free image scaling such as super-xbr or NNEDI3. The overall image appears sharper.

If you aren't using SuperRes, image doubling won't make much a difference at a 1.5x scaling factor.

I prefer Jinc AR for Chroma, and Image Upscaling. SuperXBR can be too sharp and I don't have enough time in my life to test out the optimal sharpness.

SupeRes is pointless as well, as it likely just sharpens the image and I literally can't tell what it does after a good hour of viewing with/without it.