madVR - high quality video renderer (GPU assisted)

[nekromantik]

hi all
would any of MadVRs upscaling work well with Intel HD 620 graphics on a NUC?
or would it be limited to bicubic and would need a dedicated card like 1050 nvidia for anything more?
thanks

[huhn]

depends on the target and source resolution.

i used a HD 4400 for month and there was no problem at all.

[nekromantik]

source will be 720p and 1080p to 2160p

[huhn]

a nuc should be very limited for that.

they usually can hold the max clock speed to because they overheat or reach there power target.

a small ITX sized PC on the other hand should have any of these problems but i can't give details about that i'm not using hardware like that.

[nekromantik]

a nuc should be very limited for that.

they usually can hold the max clock speed to because they overheat or reach there power target.

a small ITX sized PC on the other hand should have any of these problems but i can't give details about that i'm not using hardware like that.

was afraid of that :(
guess I shall leave my TV to upscale then.

[CruNcher]

a nuc should be very limited for that.

they usually can hold the max clock speed to because they overheat or reach there power target.

a small ITX sized PC on the other hand should have any of these problems but i can't give details about that i'm not using hardware like that.

How about a Super efficient Sony X1 Processor with "double database" instead tumbles down this year into the Premium PS4 class TVs (from the Ultra Premium Class) ;)

https://www.youtube.com/watch?v=zb8UH0liZW8

Now that more users gonna get access to it i wonder if we get more out of it how it works and what it really is :D

[Asmodian]

lol, you keep promoting non-madVR solutions in this thread. :p

[nekromantik]

How about a Super efficient Sony X1 Processor with "double database" instead tumbles down this year into the Premium PS4 class TVs (from the Ultra Premium Class) ;)

https://www.youtube.com/watch?v=zb8UH0liZW8

Now that more users gonna get access to it i wonder if we get more out of it how it works and what it really is :D

As good as the X1 is, I doubt it can beat MadVR for processing quality.

[nussman]

lol, you keep promoting non-madVR solutions in this thread. :p
+1

It's just annoying and I think we dont need a Sony X1 Processor discussion atm in this thread. :(

[aduwind]

lol, you keep promoting non-madVR solutions in this thread. :p

I agree and remember at all a madshi advice:

MODERATOR Note: This thread has had a tendency to generate long off-topic discussions. madshi has stated that the thread is for madVR support/development and not for related general discussion. Please confine this thread to direct madVR-related support/development issues. If in doubt, start a new thread and put a simple link to it if needed. If you put madVR in the thread title, people will find it. The moderating policy being enforced, including deletion of off-topic posts and material, has been discussed with and approved by madshi. Thank you for your cooperation in keeping the thread manageable for madshi and his users.

[CruNcher]

As good as the X1 is, I doubt it can beat MadVR for processing quality.

I wonder why you are so sure about that anyways yeah it gets off topic and since no one really does a foundational compare of MadVRs Shader core vs Sony Dedicated Hardware IP it's wild speculation ;)

Especially as iffy that marketing around that chip is many things shown there aren't the chips function at all it seems TRILUMINOS is hardware for example (WCG, quantum dot) and can be excluded and gets also split up from it in the Product line as we seen with the X800D.
X-Reallity (Pro) seems to be the Generic Upscaling and that is included as well so overall it becomes hard to decipher what X1 is actually and if it's really a Chip at all or only IP marketed as such ;)

One database is used to clean the picture, reducing on-screen noise. The other is used to upscale the resolution, improving clarity. These two powerful image improvement databases work together, dynamically improving pixels in real time. Each database has tens of thousands of references, amassed from our experience creating content for TV and movies over the years.

This describes nothing else then triticals NEEDI3 a trained algorithm through a dataset and they speak of 2 datasets one the denoiser was trained on and 1 the upscaling algorithm used was trained on.

So this and

14-bit signal processing breaks up the solid bands of color, effectively smoothing out the gradation between shades — displaying a 14-bit equivalent gradation, even from an 8-bit source, showing 64 times2 the color levels of a conventional display.

Could be the 2 features that are actually missing on non X1 Processor sold TVs

The whole thing becomes even more confusing if you actually understand that they also market 3 different Chips itself with this :D

4K Processor X1
4K HDR Processor X1
4K Processor X1 Extreme


This is how you confuse people correctly ;)

[pirlouy]

To avoid off-topic, if you're not logged, you can block boring users (multiple posts _questions or off-topic_) if you use uBlock:
forum.doom9.org##.tborder:has([href$="u=9759"])
Hallelujah !

[hannes69]

I have detected a strange behavior with image doubling. I don´t quite remember, but I think somebody already mentioned the problem and maybe an explanation / solution / workaround, but I don´t find it at the moment in this thread, so:
I´m using madVR 0.91.3 and I am watching a 640 x 360 movie on a 720p (1280 x 720) projector. If my math is correct ;) this means simple doubling. But it isn´t working as that - see following madVR OSD screenshot:

https://s28.postimg.org/sza1n838p/doubling_problem.jpg (https://postimg.org/image/sza1n838p/)

The problem is indicated with "scale 0,0,640,360 -> 0,1,1280,720".
If I interpret this line correctly, this means: movie has a source rectangle with left top corner 0,0 and right bottom corner 640,360 (seems right) which has to be drawn onto a target rectangle of left top corner 0,1 and right bottom corner 1280,720.
I don´t have an explanation why the target left top corner isn´t 0,0. I have checked all settings like cropping black borders and so on which could possibly influence that but didn´t find anything.
Maybe someone knows something useful about that... Seems a liitle bit strange to me to see doubling and then downscaling here when simple doubling is needed. And I don´t know what´s technically happening when scaling 720 lines down to 719...:D
Hannes.

[foxyshadis]

I wonder why you are so sure about that anyways yeah it gets off topic and since no one really does a foundational compare of MadVRs Shader core vs Sony Dedicated Hardware IP it's wild speculation ;)

Start a new thread if you want to talk about anything that isn't MadVR. Even if you're comparing MadVR to something else, start a new thread, except if it's to highlight a bug.

[JackCY]

I've had no problem using MadVR even at times when I've had no dGPU and ran only the poor and slow Intel iGPU HD4600, meaning MadVR can be set low enough for anyone to use it and it doesn't consume much power if you don't want it to. There are tons of settings and you just gotta know what to set and see what impact it has if you want to, it's the same with setting anything else in any program, customization is good. If you want a fast and dumb solution use the renderers available from the OS.

[cork_OS]

It would be interesting to see photo comparison madVR vs. competitors.

[QBhd]

It would be interesting to see photo comparison madVR vs. competitors.

In another thread...maybe

QB

[SweetLow]

The problem is indicated with "scale 0,0,640,360 -> 0,1,1280,720".

No, you have another problem (probably):
movie 640*360, 1.78:1

Precise 16:9 (640:360==16:9) have to view as:
movie 640*360, 16:9

Looks like you have wrong AR embedded in your stream or wrong decoder.

[cork_OS]

Add grain don't work (or almost don't work) with 4x NGU.

[Mistery 73]

a greeting to the forum,
Experts do you advise me to create the 3d lut for madVR as calman calibration software or displaycal?

[AngelGraves13]

Add grain don't work (or almost don't work) with 4x NGU.

It works, but you were getting double grain in the previous build. Now it applies grain once after 4x scaling.

[hannes69]

No, you have another problem (probably):
movie 640*360, 1.78:1

Precise 16:9 (640:360==16:9) have to view as:
movie 640*360, 16:9

Looks like you have wrong AR embedded in your stream or wrong decoder.
Probably ... NOT ;)
I watched another movie yesterday, now the behavior has changed (now with movie marked as 16:9).
So once again 2 different OSD screenshots for 2 different movies:

https://s28.postimg.org/sza1n838p/doubling_problem.jpg (https://postimg.org/image/sza1n838p/)

https://s28.postimg.org/sggcj2e95/doubling_problem2.jpg (https://postimg.org/image/sggcj2e95/)

Seems that my projector can dynamically change its vertical resolution:cool:
Screenshot 1 indicates 719 lines, screenshot 2 indicates 722 lines. Negative numbers for target rectangle seem a little bit strange to me. Why display the first movie line in Nirwana?:cool::confused:
Hannes.

[JackCY]

It would be interesting to see photo comparison madVR vs. competitors.
Easy for anyone who can use the keyboard and press printscreen. Do it yourself.
Add grain don't work (or almost don't work) with 4x NGU.
Works as intended for me, it's no longer applied twice as it was in 2x+2x or it was exaggerated in some other way.

[dipje]

Maybe I've missed something but..

... starting MPC-HC with MadVR causes the DisplayCal Profile loader to switch to a 'disabled' state.
Is it possible to disable this behavior? I know there is a special madVR calibration-lut mode, but what if I don't want to? Just leave my windows calibration alone (specially on the other screen while MadVR is playing). Is that possible somehow? Can't seem to get it to work.

[ryrynz]

Easy for anyone who can use the keyboard and press printscreen. Do it yourself.


If you followed the conversation you'd have noticed
"competitors" was generally referring to hardware solutions.

[huhn]

are you using a 3D LUT?

[nevcairiel]

Screenshot 1 indicates 719 lines, screenshot 2 indicates 722 lines. Negative numbers for target rectangle seem a little bit strange to me. Why display the first movie line in Nirwana?:cool::confused:


Do you have any of the black bar options enabled? maybe madVR is detecting a black bar of 1-2px at the bottom and zooming it away.

[SweetLow]

>Do you have any of the black bar options enabled?
I have checked all settings like cropping black borders and so on which could possibly influence that but didn´t find anything.


now the behavior has changed
Ok. As madVR detect "touch windows from inside" try enable "disable scaling if image size changes" to some high value (and completely disable hard coded black bars detection to debug). And then try to view movie in natural size.

[cork_OS]

Easy for anyone who can use the keyboard and press printscreen. Do it yourself.
As soon as I buy Sony X1 Processor TV, I will use keyboard and press printscreen combo.

It works, but you were getting double grain in the previous build. Now it applies grain once after 4x scaling.
Works as intended for me, it's no longer applied twice as it was in 2x+2x or it was exaggerated in some other way.
Well, OK, but after NGU 4x added grain is definitely less visible than after NGU 2x (EDIT: once).

[Selur]

Is there some way to control which gpu madvr is using?
Got a Geforce GTX 980 Ti, a Radeon RX 460 and an Intel HD Graphics 4600 in my system and I suspect that MadVR has some problem with the Radeon.
Since I added the Radeon my whole system crashes from time to time when I play content in MPC-HC while using MadVR.
Tried en-/disabling DX11 rending in MadVR but that didn't help.
So I was wondering if I could tell MadVR somehow to now use the Radeon?

[70MM]

I find that just a few of my older titles dont fit the left and right sides of my motorised 4x3 masking. 90% are fine, the other 10% are too small on the left and right only.
I would like to use the zoom control on madvr but the 25% is too big!
How can I get a smaller zoom stepup thats less than 25%?
It would be nice if there was control for 5% increments.....

I have tried all the different combinations under ZOOM, black bars etc etc but I cant find anything in there that will bring these last 10% 4x3 films up to the true aspect of 4X3.

I have no issues at all on scope or 16:9 with my masking which was setup from a test disc for ratios.

[dipje]

are you using a 3D LUT?

No, .icc / .icm profile in Displaycal Profile Loader.

I just want MadVR to NOT touch my Windows calibration for both my monitors, I don't even care about calibrated MadVR to be honest, just don't suddenly turn my Photoshop into 'uncalibrated photoshop' that's on another monitor while MadVR is running :S...

[huhn]

with default calibration setting it isn't unloading anything for my. you may have a look at this: http://bugs.madshi.net/view.php?id=290

[nevcairiel]

Is there some way to control which gpu madvr is using?
Got a Geforce GTX 980 Ti, a Radeon RX 460 and an Intel HD Graphics 4600 in my system and I suspect that MadVR has some problem with the Radeon.
Since I added the Radeon my whole system crashes from time to time when I play content in MPC-HC while using MadVR.
Tried en-/disabling DX11 rending in MadVR but that didn't help.
So I was wondering if I could tell MadVR somehow to now use the Radeon?

madVR has to use the GPU the screen its on is connected to, everything else would be extremely inefficient.

[CruNcher]

If you configure MadVR in the lowest overhead way possible i would say now it can reach EVR Native Performance :)

With MadVR and it's Statistics i also could find out that Nvidias Driver is doing something very oddly with it's Refreshrates 59Hz is shifting away after suspend got active here to 60Hz (Control Panel still shows 59Hz)

After the System goes into Suspend i have to reset the Refreshrate manually back to 59Hz with the used Driver, odd behavior here on my System, that i most probably without MadVR wouldn't be gotten aware off :D.

The very precise Refreshrate Display can be handy :)


http://i1.sendpic.org/t/5R/5R4SHNenFpnBrFxV8WR6RgpkEfC.jpg (http://sendpic.org/view/1/i/68ZiZ41QR8XlIT3mFg6dTGmHGmS.png)

Lowest overhead so far i could push out of MadVR, which results in a perfect stutter free playback result even on the CPU :)

Sad that the Tearing Test is not available for MadVR but at that Dance Motion you should realize any Motion Flow interruption in a ms most probably even before the timers actually catch them in that buffer and DWM Playout ;).


@madshi

could you make the Statistics timer pooling interval configurable (in a hidden advanced option) ?
And could you add Gothsyncs (EVR Sync Renderer) like Graph as a second Statistical visualization output ?

http://www.ostrogothia.com/video/wp-content/uploads/2009/07/nojudderatvsync.png
http://i1.sendpic.org/t/pT/pTycPIXrKtuPMP5iKbDUeqNQfS2.jpg (http://sendpic.org/view/1/i/lP6Vg8AIH9oVJxbuDSUJAjDfRBJ.png)

http://www.ostrogothia.com/video/?page_id=1218

[Selur]

madVR has to use the GPU the screen its on is connected to, everything else would be extremely inefficient.
okay, so if I move the window from on monitor to the other the gpu used during processing is changed?

[Seedmanc]

So I got myself a capture card and as I was messing around with it, capturing the output of my videocard and sending it back to a videoplayer on the same screen, I noticed, that using different renderers for video output progressively degraded video quality as it was sent through the capture card and back to the player multiple times. Pretty much every renderer except for madVR affected quality in one way or another, ruining colors, making image too dark or too bright and so on. For example, this is how it looks for a generic "Video rendered" in Potplayer: image (http://i.imgur.com/Q0DbMLe.jpg) (leftmost image is the original, the further to the right - the more times the image went through the pipeline). Props to madshi for making quality software I guess.

Another thing I wanted to note is that since my capture card converts the video to YUY2 with horizontal chroma resolution halved, I thought it would be an interesting setup for testing chroma upscaling capabilities of madVR. Again, utilizing multiple passes to enhance the discrepancies, if any.
For example, here's how it looks without any resampling, just nearest neighbor: image (http://i.imgur.com/T60EWCI.jpg) (the chroma shift to the right is the capture card's doing).
And here's how Reconstruction (sharp) looks: image (http://i.imgur.com/WqgQGcB.jpg). This was most odd looking image out of all other methods. Also I couldn't quite see the difference between NNEDI 16 and NNEDI 256.

I'm sharing this so that maybe someone would find this method useful or develop a similar one for evaluation of madVR filtering.

[CruNcher]

I guess MadVR could even beat Furmark and become Reviewers choice of Stability testing :D

i really wonder what happens if i use NGU now instead :D

http://i1.sendpic.org/t/9y/9ygHxtZBypmyujHlmyJyQBk7LzC.jpg (http://sendpic.org/view/1/i/qkJ2tnwTOlLVN28mH4UlWoqAiF4.png)

NGU is actually more stable then NEEDI3 in this specific CPU Decoding case

chroma/image dxva = 150W
chroma/image Lanczos4 = 300W
chroma NGU-Low = 300W

[oddball]

Did anyone ever come up with a way to 'fake' black frame insertion @ 60Hz without crazy flickering and other weirdness? From what I understand proper BFI requires 24/120 and only inserting a black frame for a few miliseconds (like a projector) whilst repeating a frame. Simply inserting a black frame for 1 single frame can cause flickering? I'm sure this could be achieved if the input supported 120Hz but for 60Hz is another matter entirely.

I have a feeling it's just one of those things that can only be done on a 120Hz display though (in the display itself).

I have an LG OLED BTW.

[aufkrawall]

If you configure MadVR in the lowest overhead way possible i would say now it can reach EVR Native Performance :)

madVR works different than any other video renderer in terms of queuing/drawing frames by really telling driver (or DWM?) to actually render ahead. As long as queues are filled, observed "overhead" is mostly just a theoretical observation which is not very interesting.
No one cares about a few percent more or less CPU usage when watching a video. Even if madVR would have a higher CPU usage, playback is still more stable than with other renderers, as long as you don't push anything to impractical limits.

[CruNcher]

Yes sure but the chroma and scaling part can be a really heavy burden on the System depending on the actual input and goal but everyone needs to decide this actually himself and MadVR gives those option to decide it but getting this really avg user friendly is a pretty daunting task.


DXVA Chroma = 150W

http://i1.sendpic.org/t/tD/tDDbQToDhJx6o0ya2DVvI4DE9QZ.jpg (http://sendpic.org/view/1/i/oOnrBNPaag9cTk8jBjq2poHVOTi.png)

NGU-Low = 300W

http://i1.sendpic.org/t/kY/kYsYMO98x9SSznAj5iNirFVtghZ.jpg (http://sendpic.org/view/1/i/3gDEIRTw6baAnD3kPTMxqz6QUYC.png)


In motion detecting the difference is like impossible, but the consequences are huge and there is nothing that makes aware of them in the GUI there is no indication about Performance and GPU Shader efficiency at all.

So it doesn't surprise me at all that many users actually experience dropped frames overall in some configuration overtasking their system for a very questionable gain.

The small CPU overhead compared to EVR-CP or EVR-Sync is pretty unproblematic as you said at least on Nvidias Driver ;)

[huhn]

and these power consumption number come from what?

even with a titan X it is going to be hard to reach 300 watt...

[cork_OS]

and these power consumption number come from what?
I bet it's overall system consumption measured by PSU or external power meter.
But high CPU usage with madVR is weird thing. Looks like software decoding.

[CruNcher]

130W
290W

What my Energy Provider is going to calculate with,not far off actually ;)

And yes total system power still 2x as much is counted with Ngu-Low and the Shader usage also indicates it

[aufkrawall]

Even in Germany power prices aren't so high that you'd have to worry about that, unless your media habits are quiet unhealthy...
And you are not even forced to do NGU chroma for 4k content. Jinc or eben bicubic might be well enough.
DXVA2 scaling and decoding are poison for image quality, we had this several times. If you don't care about quality, then, well, don't.

[CruNcher]

Yeah i see though madshi did already improvements on the GUI logic Nearest Neighbour and Bilinear seem to be completely avoided and DXVA prefered only after that you can higher the Shader overhead like you wish and also the Ordering is actually in the Overhead order though NEEDi3 falls off somewhat being Compute and behaving differently.

Seeing that Lanczos4 overhead and knowing you get that on Intels GPU for free lets you also wonder why Nvidia isn't moving on.

It's pretty nice at just 140W roughly 10W above DXVA in it's MadVR Shader incarnation.

http://i1.sendpic.org/t/bE/bEpCr0DIeGRcHvIueDUdczXOzMJ.jpg (http://sendpic.org/view/1/i/bg1SUjZpugy7bULauo3lGwSSOl6.png)


Ok so if image upscaling is DXVA2

changing chroma upscaling to Nearest Neighbour or Bilinear has no effect and stays on DXVA but if image upscaling is not DXVA2 it can be changed to either, interesting.

[YxP]

unless your media habits are quiet unhealthy...

Hahah :D Yes yes spam and everything.

[Buckster]

300W - what GPU is that with ?

I think I measured about 110W (will check later) base unit - from the wall power meter NGU-High (Xeon 1225v3 + overclocked 1050 Ti)

[Arm3nian]

4k playback is broken for me. GPU usage is at 99% but the card stays at low clock levels. More upsampling just increases render times. 1080 video plays fine. Decoder queue is at 128 but render queue is at 0-2.

It was working fine a couple of days ago. I don't remember for sure but I think it started after new nvidia drivers.

[huhn]

did you change the power settings back to adaptive.

you have to change this with every driver it is simply like that.