Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Hardware & Software > Software players

Reply
 
Thread Tools Search this Thread Display Modes
Old 5th February 2017, 02:34   #42241  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by plasma View Post
I liked to use upscaling refinement such as adaptive sharpen and line thinning but now as I upgraded to a 4k tv it uses way too much performance. Using "image enhancements" does not give similar results, albeit much better perf.

I assume that upscaling refinements uses the upscaled 4k images and then applies to refinements while image refinements does it at the source resolution, whatever it is. Is there any way to apply these refinements at a half upscaled image (aka 1080p) so you get better balance in terms of performance penalty.
Have you tried using NGU upscaling? It's much sharper than the other upscalers, so maybe you don't need any upscaling refinement, anymore?

Quote:
Originally Posted by Cinemancave View Post
I get screen tearing when using fullscreen exclusive mode on my 60 Hz monitor (with or without smooth motion). Fullscreen windowed works just fine. On my TV and projector at 24 and 60Hz respectively, I have no tearing in exlusive mode. And ideas? :S
That's very weird! Have you tried rebooting your PC? I have this problem once in a while and rebooting always helps. If it doesn't help in your case, try a different GPU driver version. If that still doesn't help, I'm out of ideas. Fullscreen windowed is ok, too, though, if it runs smoothly and reliably for you.

Quote:
Originally Posted by dansrfe View Post
For some reason if I leave MPC-HC/madVR paused and the mouse/keyboard inactive long enough then either the gfx card or motherboard makes a long beep sound and reboots. The screen is set to turn off after 3 mins, a couple mins after this I hear the beep and it reboots.
Probably a GPU driver bug. You could try a different GPU driver version.

Quote:
Originally Posted by Warner306 View Post
Check out this post by a new user using the new image doubling menus. It is revealing what someone with basic knowledge and a good graphics card would do in these menus.

Note the choices he made (set everything on high) and what he didn't understand (downscaling after doubling and supersampling).

http://forum.kodi.tv/showthread.php?...009#pid2478009

Maybe downscaling after doubling could be called reduce overshoot of target resolution with image downscaling.
Ouch.

Quote:
Originally Posted by anta777 View Post
Сomparison algorithms chroma upscaling:
BrownFox.png(original) 0.0000
nguveryhighsr4 0.0024
nguhighsr4 0.0026
nguveryhigh 0.0028
ngumedsr4 0.0030
nguhigh 0.0032
sxbr150arsr4 0.0032
nnedi128sr4 0.0033
ngumed 0.0034
sxbr75arsr4 0.0036
sxbr100arsr4 0.0037
ngulow 0.0039
sxbr150ar 0.0048
reconsharp 0.0050
reconsharpar 0.0050
reconplacebo 0.0051
reconplaceboar 0.0052
nnedi128 0.0054
nnedi64 0.0057
sxbr100ar 0.0057
bilatoldsr4 0.0058
nnedi32 0.0059
bilatsoftsr4 0.0060
bicubic75ar 0.0064
sxbr75ar 0.0064
reconsoftsr4 0.0071
bilatold 0.0110
reconsoft 0.0117
bilatsharp 0.0134
bilatsoft 0.0135

I use dssim and brownfox.
Wow, that is a pretty good test - well done!!

How did you create the chroma subsampled image (4:2:0) from the original image, though? Did you take the chroma shift into account?

Quote:
Originally Posted by Georgel View Post
Can we get a force 200% scaling for Jinc in image upscaling?

After about 100 tests I came to the conclusion that anything but Jinc will add a bit of grain to edges while watching anime.

Jinc seems the most rounded, cleanest algorithm, but I want to force a 200% upscaling still using Jinc to get the most out of it. NNEDI and NGU work very well with human and life-like content, but not with anime... At least not in my tests / setup...
Everybody else seems to agree that NNEDI3 is the best algorithm for (aliased) Anime, myself included.

Quote:
Originally Posted by Warner306 View Post
I've come across a user trying to create profiles for 3D and non-3D content. This is his rule:

else if (3D) and ((srcWidth > 1280) and (srcWidth <= 1920)) "1080p 3D"
else if (3D) and ((srcWidth <= 1280) and (srcHeight > 720) and (srcHeight <= 1080)) "1080p 3D"

else if (not 3D) and ((srcWidth > 1280) and (srcWidth <= 1920)) "1080p"
else if (not 3D) and ((srcWidth <= 1280) and (srcHeight > 720) and (srcHeight <= 1080)) "1080p"

Can you tell me if anything is wrong with this rule?
Is that the whole rule, or just a part of it?

It seems to me the user just wants to switch between 2 different profiles, one for 2D and one for 3D. In that case why not simply using "if (3D) "1080p 3D" else "1080p""?

Quote:
Originally Posted by oddball View Post
Should I forget about sending 1080p to my UHD TV and let MadVR scale it to 2160p?
Yes, please! madVR will scale in much higher quality, anyway!

Quote:
Originally Posted by WolframRhodium View Post
There's a new downscaling algorithm presented in "Rapid, Detail-Preserving Image Downscaling", which seems to have better performance and result compared to SSIM.

Will it be in madvr's downscaler list in the future?
Interesting! They compared to the original SSIM paper, though, while the algorithm used by madVR is a lot different to that. I'm not sure if I'm going to implement this new algorithm. Can't say right now, but thanks for letting me know!

Quote:
Originally Posted by Neo-XP View Post
Strangely, NGU-High (former NGU-Med) is a lot sharper in v0.91.1 VS v0.91.4
Nothing strange about it. We already discussed earlier that I'm using strict downscaling now, which I might change to relaxed in the next official build.

Quote:
Originally Posted by 70MM View Post
Im using the 1080 card with madvr...
In the 1080 settings under "Adjust Desktop Color Settings" there is a dropdown box "content type reported to display" If I set it to (Auto-select (recommended) the image has terrible banding, if I set it to "Full screen videos" the image as NO banding and the blacks are more enhanced. The only problem is sometimes when i turn on the PC it has reverted back to Auto-select and Im back to the banding. Even though I save the setting why would it revert back? I only use this card and PC in my HT.

What do others have this setting selected to?
It seems weird that this has an effect at all. Are you using anything DXVA related (scaling, deinterlacing, decoding)?

Quote:
Originally Posted by 70MM View Post
Here you go....
resolution is set the same as yours I believe.
As long as I have Content type reported by display set to Full screen videos i have no banding. On auto I get banding.
Shouldn't "Digital Vibrance" be set to 0%?

Quote:
Originally Posted by Neo-XP View Post
That was it ! because there is no difference when I test NGU-veryHigh on both versions.

I'm not certain that it is an improvement though. It sure hides the artifacts better, but it is really soft.

NGU-veryHigh on the other hand is extremely impressive : razor-sharp with practically no visible artifacts. I don't how madshi managed to do that, it is almost magical
It's actually the downscaling algo, which makes the difference. VeryHigh uses SSIM. Lower versions use strict Bicubic150 which can be a lot softer. See my previous comment on that.

Quote:
Originally Posted by Neo-XP View Post
@madshi Do you plan to enable the "add grain" enhancement for other image doublers than NGU in the next version ?

It is a really neat feature which subjectively increase the details of the image when upscaling SD content to FHD or 4K. I would like to use it with super-xbr.

Did you write the code yourself or is it a script/shader that you can share ?
Maybe in a future version. It's a very very simple TPDF noise which madVR also uses for dithering.

Quote:
Originally Posted by cokeefe View Post
Madshi, thank you again for your work. NGU is the best thing that's happened to my HTPC in quite some time.


Quote:
Originally Posted by Neo-XP View Post
Is the anti-bloating gone for good for super-xbr doubling ? I use it at only 25% and it does a pretty good job in reducing ringing and dark halos.
It's gone for now. Of course I could bring it back, but instead I would really like to get rid of super-xbr altogether (for image doubling at least). Why using super-xbr AB instead of NNEDI3 or NGU?

Quote:
Originally Posted by ausvette View Post
I have a issue I haven't been able to solve with scaling from 2k to 4k.

Normal playback is fine. rendering time around 29ms. But when I pause and resume playback or seek forwards and backwards, quite often the picture disappears and I get a screen which is mainly white with sometimes a vague noise outline of the major contrast areas. Sound remains fine.
This seems like a decoder issue to me. But in order to make sure I'd suggest that you check if playback is still running (e.g. display the madVR seekbar by moving the mouse cursor to the bottom of the screen, if you're in fullscreen exclusive mode), also check the OSD (Ctrl+J) to see if the queues are all full and no frames are being dropped.

Have you tried software decoding?

Quote:
Originally Posted by Arm3nian View Post
Is it recommended to use "process HDR content" without an HDR display?
No. The display would still receive HDR data (PQ transfer curve), which it can't properly display.

Quote:
Originally Posted by Arm3nian View Post
When I specify a 3DLUT, for any of the 2 options on the hdr page, it says wrong input/format for madvr.

Do I make a bt.709 lut or a bt.2020 lut for bt.2020 content on a normal gamut display (100% srgb)?
If your display can't handle HDR on its own, then do a normal SDR calibration on the calibration tab. That calibration will also cover HDR content, as long as you allow madVR to convert HDR to SDR by using pixel shader math.
madshi is offline   Reply With Quote
Old 5th February 2017, 02:35   #42242  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by CruNcher View Post
Did somebody compared MadVR vs Sonys latest released X-Realtiy PP especially the Super Resolution Upscaler, compared to a relative price point optimized consumer Product like the XD80 series ?
What about Samsungs KS7000/8000 SAIT R&D improvements, they currently seem to win a lot of prices visualy not only from the Panel part itself but their FRC conversion seems to have reached the most stable state including the BFI could be even intersting comparing vs SVP and AMD Fluid Motion by now.
Feedback from users on AVSForum suggests that they like NGU better than what the latest Sony and JVC projectors do internally. But it's hard to compare this in detail because we can't do screenshots of what the TV/projectors do internally.

Quote:
Originally Posted by CruNcher View Post
Panasonic is also doing some crazy stuff creating for each Frame a Histogram and adapt that to get better contrast and depth from inputs, also Samsung seems todo something similiar since some time now
That's nothing new, has been done by some CE devices for decades. I hate that kind of stuff. E.g. movies with low-contrast misty scenes suddenly become high-contrast? Way to destroy artistic intent!

Quote:
Originally Posted by Georgel View Post
Seems I got a random Crash with this error

exception number : 1
exception class : Exception
exception message : Zugriffsverletzung bei Adresse $61a07805 in Modul 'XySubFilter.dll'. Lesen von Adresse $0.

No idea what could had caused it (?)
It clearly says that the crash is in XySubFilter.dll.

Quote:
Originally Posted by Patrik G View Post
NGU goes overboard and crash and burns why do you want to compare it?
but maybe its hard to see on a LCD/OLED tv with only 300 lines of motion resolution lol
its still blurry as hell.

for plasma users NGU looks crap
i get the same effect if i bump up sharpness +20.
Hello to flickering chainsaw edges and morie patterns all over the image.

NNEDI3 produces an accurate and natural image NGU does not.
Back to the drawing table madshi
and one tip: use a proper tv to evaluate your scaling algorithms
One tip to you: Train your eyes before you try to judge scaling algorithms. It seems you don't really know what you're talking about. And btw, while using artifical test patterns can be beneficial if taken with a pinch of salt, testing with real world material is much more important.

In all my scientific tests (using groundtruths and measurements like PSNR and SSIM), NGU runs circles around NNEDI3.

Fact is: NNEDI3 has a specific purpose (originally: deinterlacing; somewhat unintended: upscaling of highly aliased sources) and works great for that specific purpose. But it's rather bad for upscaling high-quality video/photo sources. NGU also has a specific purpose (mainly upscaling of high quality video/photo sources) and works great for that specific purpose. NGU fails totally if you feed it with crappy sources. NGU doesn't like aliased sources, or sources with lots of visible compression artifacts.

Quote:
Originally Posted by CruNcher View Post
No noise at all the clear Signal that is what you aiming for neither Temporal Noise nor Spatial Noise
You might aim for that if you want to build a "pure passthrough" source device and do all processing in the display. But that's the one and only situation where you aim for "no noise".

Already YCbCr -> RGB conversion means you need to apply dithering (noise), otherwise you get much more dramatic artifacts then noise.

Quote:
Originally Posted by kevbo View Post
I was wondering if madVR has ever considered including DLP Checkerboard as an output format?
I don't think it makes sense for me to implement something without being able to actually test it.

Quote:
Originally Posted by CruNcher View Post
Anyway Debanding is not a solution but a workaround and im pretty shocked that some find debanding the whole frames efficient same they like to enable dithering (i hate both and want it to finally die)
Debanding is useful if (and only if) the source has banding artifacts baked into it. Some Blu-Rays have this problem. For those madVR debanding can help *a lot*. Of course with 10bit UHD Blu-Rays the hope is that it will not be needed, anymore. We will see...

Dithering is a whole different topic. There's a law in digital processing: Whenver you reduce bitdepth, you *HAVE* to apply dithering. It's good, it's necessary, and if you don't do it, very bad things will happen to image quality.

Quote:
Originally Posted by Neo-XP View Post
The 4x algorithm is unfortunately too soft compared to the 2x2x method, but I guess some users will like it because of this
It's softer for some edges, but sharper for others. I'll post a detailed image comparison when all the work is done. Will be while, though.

Quote:
Originally Posted by JackCY View Post
On the other hand I have to agree that soft DVD sources do need a bit of edge presharpen so that NGU can do a clean edge instead of trying to "vectorize" the thin soft edge gradient with a "single color" vector area, this happens more with the higher quality settings on some samples. With a sharp 4k recording downscaled to DVD resolution it wouldn't be an issue but these older DVDs mastered from film are often soft, if not then they are ruined to hell with a global sharpener and halos by who ever was attempting to convert the film to DVD.

It would be nice to have a variable in the GUI that would change how sensitive NGU is to these soft edges so that one can dial it in.
Will see later when the other improvements are released if it's needed.
Overall I do postedge sharpen and thin edges as before without NGU, looks better to me. Sure NGU is sharper but I guess as other have experienced previous versions, it was even more sharp before and these postprocess filters wouldn't be needed.
Can you post more information about this? E.g. some screenshots, maybe a short sample?

Quote:
Originally Posted by huhn View Post
the first thing is why am i able to use NGU very high x4 on a RX 480? and how should i compare them when chroma high is using NGU low?

i have to switch madVR versions to do a proper comparison.
Yeah, comparing performance requires switching versions. Sorry about that. The direct quadruplers are much more efficient than running the doublers twice, which is why performance is better.

Quote:
Originally Posted by Asmodian View Post
Wow, I am very impressed by NGU x4 for SD content! Much of the very bad amplification of artifacts I noticed when upscaling 640x480 -> 4K is gone!


Quote:
Originally Posted by wysiwygbg View Post
Yeah, I've already tried all of the above and none of them work. So is there going to be frame interleave support in the foreseeable future? Does this mean that all TV sets made before 2012, and working with HDMI 1.3, are currently unsupported by madVR for watching 3D?
I'm using the 3D API offered by Windows 8.1 and Windows 10. This API doesn't let me control the 3D format. I can only activate 3D or not. If I activate it, the OS only seems to support frame packed. It's not a limitation I have any control over.

I suppose that PowerDVD etc use private APIs from AMD, NVidia and Intel to make other 3D formats work. I don't really have much interest in implementing custom support for 3 different GPU driver APIs.

Quote:
Originally Posted by pose View Post
Is it normal to have multiple madHcCtrl.exe running while playing a video?
Yes. madVR sometimes starts a 2nd madHcCtrl.exe to watch over the media player, so the original display mode can be restored in case the media player crashes or something like that.

Quote:
Originally Posted by alexnt View Post
Just one more question. I used to watch movies with letterboxing. So now that madvr can handle scalings in a better way than the player itself I want to know which is the best setting to upscale the non true 1080p files(like 1920x800) or the 720p files to 1080p to cover the whole screen.
If I use in mpc (right click)video frame-> zoom 1 I loose a bit from the edges of the film. If zoom 2 a little more. If I use touch window from outside I loose a bit more from zoom 2.
If I use scale to 16:9 TV people in the movie seem more streched (I don't how to describe it) and if I use zoom to widescreen I loose again part from the edges of the movie.
What is the best setting in your opinion or what is the most mainstream setting.
Make sure you use the latest MPC-HC version, then choose "touch window from inside". Then setup madVR's "zoom control" settings to your liking.

Quote:
Originally Posted by aufkrawall View Post
Smooth edges seems to be broken with the 4x NGU step, doesn't show any effect then.
I think it should still work, but it's only applied on the final 4x upscale, while if you double twice, soften edges is applied after each 2x upscale. And the first time it's applied, it's more effective because the 2nd 2x upscale already upscales an already softened image. Same with "add grain". I might have a solution for that for a future build.

Quote:
Originally Posted by CruNcher View Post
Ouch i looked at old issues virtually all of those are fixed (options added) in 0.91.5 and performance (stability) pretty awesome as well on low overhead config.
Glad to hear that!

Quote:
Originally Posted by Mistery 73 View Post
Experts do you advise me to create the 3d lut for madVR as calman calibration software or displaycal?
Both should do fine (I hope). If you have a Calman license you can try both and compare. Let me know the results!

Quote:
Originally Posted by 70MM View Post
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%?
Which exact setting are you talking about?

Quote:
Originally Posted by dipje View Post
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.
In the madVR "calibration" tab, is the "disable GPU gamma ramps" option checked? Does unchecking that option solve the problem?

Quote:
Originally Posted by CruNcher View Post
If you configure MadVR in the lowest overhead way possible i would say now it can reach EVR Native Performance


Quote:
Originally Posted by CruNcher View Post
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 .
Yeah, refresh rates can be a bitch. I had to do ugly system level hacks to make Direct3D use the refresh rates I want, but still things like Suspend can screw things up, unfortunately.
madshi is offline   Reply With Quote
Old 5th February 2017, 02:36   #42243  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by CruNcher View Post
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 ?
Is the stats timer too slow for your taste? What kind of information do you need to be updated more often in the OSD?

Sure, a nice graph would be lovely, but it also costs development time, which I'd rather spend elsewhere right now, to be honest.

Quote:
Originally Posted by Seedmanc View Post
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 (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 (the chroma shift to the right is the capture card's doing).
And here's how Reconstruction (sharp) looks: image. 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.
Great test - well done!!!

And I'm glad to hear that madVR proved to be the most accurate renderer - that's what I was always aiming for!

From your screenshots it looks as if the chroma placement keeps getting worse, though. Is it possible that your capture card is responsible for that? I'm pretty sure that madVR does the chroma placement correctly. Well, at least for MPEG2, VC-1 and h264 content.

Quote:
Originally Posted by oddball View Post
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 do think that at least 120Hz is needed for that, maybe more.

Quote:
Originally Posted by ryrynz View Post
Madshi, I'd like to be able to cycle through selected profiles using the same key combination.
This would make it easy to review during playback what profile may be viewed as best for the playing content without needing to remember various key combinations for each suitable profile.
Can you add support for this please?
Isn't that already working? You can assign key shortcuts to either one profile, or to the profile group. The profile group shortcut should toggle/cycle through the profiles.

Quote:
Originally Posted by Mistery 73 View Post
curiosity ... madVR using the settings of the VPR / Tv affecting the image detail keep them to zero?
Sorry, I don't understand the question.

Quote:
Originally Posted by koravei View Post
Hello Madshi, is it possible to implement HLG (Hybrid Log-Gamma/ARIB STD-B67) in your HDR to SDR conversion?

Here is one demo in HDR - HLG: http://demo-uhd3d.com/fiche.php?cat=uhd&id=154
Yes, it's possible, but not very high priority right now.

Quote:
Originally Posted by steakhutzeee View Post
EDIT: Red line solved, problem was related to "Radeon Settings". I disabled 'Demo Mode' and problem solved, but why problem disappears if i disable Madvr too? Just curious...
Might depend on whether DXVA processing is used or not.

Quote:
Originally Posted by hannes69 View Post
I further investigated the problem. The videos have a resolution of 640x360. There seems to be a problem with display aspect ratio, stream aspect ratio, container aspect ratio. I´m a little bit confused with the different areas where aspect ratios come into play. But I found out that one movie´s aspect ratio was declared as 1.772 in one place. When doing the math the other way leads to 361 lines when fixing the horizontal resolution to 640 (640 / 1.772 ~= 361). So there is no cropping needed, but a fix of the aspect ratio. I tried the setting in LAV video concerning the stream / container aspect ratio, but this didn´t change anything. Then I tried the settings in MPC HC -> View -> Video Frame -> Aspect Ratio and there I got success
The default DAR aspect ratio setting doesn´t work here. So I tried 16:9 and Assume square pixels (SAR). Both seem to work.
16:9 I assume is a kind of force, and the SAR setting uses the stream aspect ratio forced. I don´t understand what the difference is from the corresponding LAV video settings though...
I suppose the weird AR information probably comes from the video container? I've no idea, though.

Quote:
Originally Posted by flashmozzg View Post
I have the problem with latest (though not sure for how long, maybe a year) versions of madVR: in MPC-HC (latest stable release) if I go fullscreen and right click (for options) the screen goes black and mpc starts hogging up the memory (~4 GB) and using 60% of the CPU until I shut it down from Task Manager. Using EVR, ir right-clicking in windowed mode is fine. Is this a known issue with madVR/mpc-hc?
Win 10 x64, r9 290
Try using "use D3D11 for presentation". D3D9 seems to be partially broken in newer AMD drivers.

Quote:
Originally Posted by derpycat View Post
I'm also having a huge issue with Nvidia (GTX 1070). In FS exclusive I'm getting no actual dropped or delayed frames or presentation glitches, but I'm getting extremely noticeable stutter. Rendering time is something like 35ms, not the ~41 at which you start seeing problems. Windowed mode actually reduces rendering time by about 10ms and eliminates the stuttering.

I've got my card at full power for mpc-hc, and it's happening in hi10p videos.

However, this only happens with NGU Very High. With NGU set to High instead, rendering times are ~15.5ms for both exclusive and fullscreen, and no stutter.
Does the OSD (Ctrl+J) report dropped frames when the stutter occurs? FWIW, NGU Very High is only slightly better quality than NGU High, while being a lot slower. I'd recommend that you simply stick to NGU High.

Quote:
Originally Posted by fuzz! View Post
1. Change the "reduce bar size" option from a drop-down to a numeric where we can enter any percentage (25% is often a bit low whereas 50% can sometimes be severe)
Originally I had planned to only offer "zoom away" or "keep". Then I thought some intermediate values might be a good idea, so I added 25%, 50% and 75%. Do we really need more than that? It's not so easy to fit all that into a settings page. The current drop down menu is easy to understand and just one control. If I replace that with a numeric edit field, it becomes harder to understand, and what happens to the "zoom the bars away completely" option then? Sure, it could be interpreted as 100%, but again, doing that makes it less intuitive. I'd rather add a couple of % numbers to the drop down, if it's really needed.

Quote:
Originally Posted by fuzz! View Post
2. If it's possible to detect scene changes (not sure - don't think a simple wait for an I-frame would do) it would be great to initiate zoom on a scene change so it's not noticeable to the user. This way movies that change aspect ratio (eg. IMAX titles like Interstellar/The Dark Knight) could become a seemless blend of full-frame and zoomed content.
The first detection of a changed AR can be either fast or slow, depending on how bright the scene is. But once the 2 different IMAX ARs were correctly identified, switching back and forth should be *very* fast afterwards. E.g. in Interstellar in the scene where the father talks to his daughter through time, AR changes a lot back and forth. When playing this, AR scene changes are pretty much instant for me. I'm not sure if there's much potential for improvement there, to be honest!

Quote:
Originally Posted by Anima123 View Post
Somehow, the new NGU 4x + show comic like effect on very low-res videos, not as good as NGU 2x to me.
Screenshots?

Quote:
Originally Posted by baii View Post
What is the difference hdr -> sdr via shader math or 3dlut? and how to make such 3dlut?
The difference is that shader math is more flexible and easier to activate, but eats more GPU power. You can create a 3dlut with DisplayCAL. Questions about how to create the 3dlut please ask here:

http://www.avsforum.com/forum/139-di...argyllcms.html

Quote:
Originally Posted by BluesFanUK View Post
Quick question - i'm using PotPlayer and it can render VR videos without the need for additional software. It doesn't work when MadVR is set as the video renderer though. Is this a MadVR issue or is there a way to disable MadVR when playing a VR vid?
I've no idea about VR right now. Haven't looked into it at all yet.

Quote:
Originally Posted by Animus120 View Post
So I just found out that using madVR 0.89.6 + XySubFilter beta3 actually shows the subtitle in screenshot, but any newer madVR than 0.89.6 will make subtitle not appear in screenshot. I wonder what cause this problem though
You can always use PrintScreen, of course. Screenshotting functionality will be revisited in some future version, will take a while, though.

Quote:
Originally Posted by ryrynz View Post
could you please add everything in the upscaling refinement page into the keyboard shortcut list?
I don't like the idea of doing that now. I'm not even sure the page will stay as it is for long.

Quote:
Originally Posted by Xaurus View Post
Ok so I have a strange problem recently. Getting a sudden bunch of frame drops that last something like 1 second. Typically 10-15 frame drops at the same time.
Then it can be perfectly fine again for 30 minutes or so, then it happens again.

I'm using the GPU to about 75% usage (roughly 30 ms rendering time for 23p content) and I have tried with DX 11 fullscreen and windowed, DX 9 fullscreen and windowed and DX 9 overlay.

I checked GPU-z and I noticed something weird, the GPU usage would hover around 65-75% all of the time, but at those intervals where I get these frame drops the GPU usage is 99%.
So something is suddently eating all of the GPU resources.

This only started to happen recently. I used DDU and removed the latest nvidia drivers and tried an older version, but the problem persists.

Does anyone have an idea what this can be? I am on Win 10 x64 with that big update released in the fall (clean install).
See my signature for hardware details.
Do you have any 3rd party software installed, other than the media player, LAV and madVR? Notorious software for this kind of problem is GPU-Z, f.lux, and maybe your browser. Or any other software which might access the GPU.

Quote:
Originally Posted by Damien147 View Post
...oh...no fun.I can hear this crackling noise while movie playing.
Could be DPC latency issues. Google for "dpc latency checker".
madshi is offline   Reply With Quote
Old 5th February 2017, 02:40   #42244  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Nvm. Will look when I'm home.
ryrynz is offline   Reply With Quote
Old 5th February 2017, 02:48   #42245  |  Link
flashmozzg
Registered User
 
Join Date: May 2013
Posts: 77
Quote:
Originally Posted by madshi View Post
Try using "use D3D11 for presentation". D3D9 seems to be partially broken in newer AMD drivers.
I've managed to fix my issue by setting the presentation queue size to 6 instead of 8.
I'm not sure why there was no issue before, maybe it never worked with 8 but updating madVR reset that setting somehow (when it introduced new settings UI), maybe it worked before but something triggered that bug. Anyway, 6 is fine too)
flashmozzg is offline   Reply With Quote
Old 5th February 2017, 02:49   #42246  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
So, here's a little test build. Simply replace the v0.91.5 files with these:

http://madshi.net/madVRpixart.zip

I've created a new NGU variant which I'm going to call maybe "NGU pixart" or "NGU smooth", I'm not sure yet. This NGU variant will be optionally available in addition to the already existing NGU algorithm. The "pixart" variant has a look that is quite similar to NNEDI3. Which means it's very smooth and works great to reduce aliasing artifacts in the source. However, just as NNEDI3, "NGU pixart" is quite soft. My development goal was to create a NNEDI3 successor which has similar properties, but achieves a better "quality per watt" ratio.

Currently available are NGU pixart in "low", "medium" and "very high" quality settings. The "high" quality setting internally uses the "medium" algorithm for now. There'll be a separate "high" algo in the next official build.

In order to keep my development cost low, I haven't changed the settings dialog yet. So this test build replaces the main NGU algorithm with "NGU pixart". Which means this test build does not do the main NGU algorithm at all. The one and only purpose of this test build is so you guys can directly compare "NGU pixart" to NNEDI3, for both luma doubling and chroma upscaling.

So now I need your FEEDBACK:

1) Do you think NNEDI3 looks better? Or the new NGU pixart variant?
2) How does performance compare for you?
3) Can I replace NNEDI3 with NGU pixart in the next official build (pretty please)?
4) Can I replace super-xbr (for luma doubling, only) with NGU pixart in the next official build?

If you prefer NNEDI3, please let me know why. If you think NNEDI3 looks better, please provide comparison screenshots (for luma doubling comparisons always include the original unscaled source, please).

Thanks!
madshi is offline   Reply With Quote
Old 5th February 2017, 03:17   #42247  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Here's a little comparison comparing various algorithms when upscaling pixel art (highly aliased sources):

low-res original -- | -- Jinc -- | -- waifu2x -- | -- NGU very high -- | -- super-xbr 100 -- | -- NNEDI3-16 -- | -- NNEDI3-256 -- | -- NGU pixart low -- | -- NGU pixart med -- | -- NGU pixart very high

As you can see, with sources like this, linear resamplers like Jinc fail. Also the standard "NGU Very High" and even the mighty waifu2x fail here. super-xbr does fine, after all that's exactly what it was made for. But best in class are NNEDI3 and NGU pixart.

Which do you prefer?

(Zooming a little into the images will reveal more.)
madshi is offline   Reply With Quote
Old 5th February 2017, 03:26   #42248  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
Quote:
Originally Posted by madshi View Post
Yeah, comparing performance requires switching versions. Sorry about that. The direct quadruplers are much more efficient than running the doublers twice, which is why performance is better.
well more than x3 times faster was more than unexpected for me.

there are no in deep test for x4 because i didn't had card to use it.

Quote:
I've created a new NGU variant which I'm going to call maybe "NGU pixart" or "NGU smooth", I'm not sure yet. This NGU variant will be optionally available in addition to the already existing NGU algorithm. The "pixart" variant has a look that is quite similar to NNEDI3. Which means it's very smooth and works great to reduce aliasing artifacts in the source. However, just as NNEDI3, "NGU pixart" is quite soft. My development goal was to create a NNEDI3 successor which has similar properties, but achieves a better "quality per watt" ratio.
let's see if it isn't highlighting source artifacts.
i personally prefer "set and forget" scaler like super XBR.

Quote:
4) Can I replace super-xbr (for luma doubling, only) with NGU pixart in the next official build?
just for the polaris user you shouldn't do this.
i would even say a luma only super XBR version would be very nice for iGPUs at least i don't see NGU low to be faster.
huhn is offline   Reply With Quote
Old 5th February 2017, 03:31   #42249  |  Link
AngelGraves13
Registered User
 
Join Date: Dec 2010
Posts: 254
Ok, will test.

1) Does the pixart also apply to NGU Chroma or Upsampling only?

2) Is it possible for Chroma to scale in 1-step like Super-XBR when doubling? Like If I've selected NGU High for Chroma Scaling and selected NGU High for Luma Doubling, to just have the option under it for Chroma Doubling grey out and have it scale in a single step without the need for doing Bicubic 60?

NGU-VeryHigh still performs quite slow on my 1080 GTX and drops frames down to 0-2 render queue, so it's still unuseable. Maybe in another 2 years when there's video cards fast enough. Same for Chroma. High works ok, but cuts it pretty close for 24p content. 30p and higher content is not possible with NGU because it can't render fast enough with any current graphics card. I never figured out how to make profiles, so I just make one setting for all content.

Last edited by AngelGraves13; 5th February 2017 at 03:50.
AngelGraves13 is offline   Reply With Quote
Old 5th February 2017, 03:38   #42250  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by huhn View Post
just for the polaris user you shouldn't do this.
i would even say a luma only super XBR version would be very nice for iGPUs at least i don't see NGU low to be faster.
Luma only wouldn't be much (if any) faster.

NGU low will probably be slower on Polaris, but on other GPUs it's nearly the same speed as super-xbr.
madshi is offline   Reply With Quote
Old 5th February 2017, 04:01   #42251  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
Quote:
Originally Posted by madshi View Post
Here's a little comparison comparing various algorithms when upscaling pixel art (highly aliased sources):

low-res original -- | -- Jinc -- | -- waifu2x -- | -- NGU very high -- | -- super-xbr 100 -- | -- NNEDI3-16 -- | -- NNEDI3-256 -- | -- NGU pixart low -- | -- NGU pixart med -- | -- NGU pixart very high

As you can see, with sources like this, linear resamplers like Jinc fail. Also the standard "NGU Very High" and even the mighty waifu2x fail here. super-xbr does fine, after all that's exactly what it was made for. But best in class are NNEDI3 and NGU pixart.

Which do you prefer?

(Zooming a little into the images will reveal more.)

i think the general lines on NGU pixart low are to thin it get's better with medium but with very high the background lines are really strange/uneven and nnedi3 wins there hands down.
even NGU veryhigh wins there most of the part the line thickness is really good! but there or other obvious issues with NGU very high.
mario him self looks "best with pixart very high. his eyes are to small but compared to super XBR they are pretty good.

i have a huge problem selecting a winner here...


Quote:
Originally Posted by madshi View Post
Luma only wouldn't be much (if any) faster.

NGU low will probably be slower on Polaris, but on other GPUs it's nearly the same speed as super-xbr.
so my last standing argument is that NGU is "hit or miss". super XBR works with everything.
huhn is offline   Reply With Quote
Old 5th February 2017, 04:12   #42252  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
Quote:
Originally Posted by AngelGraves13 View Post
2) Is it possible for Chroma to scale in 1-step like Super-XBR when doubling? Like If I've selected NGU High for Chroma Scaling and selected NGU High for Luma Doubling, to just have the option under it for Chroma Doubling grey out and have it scale in a single step without the need for doing Bicubic 60?
it can't be done. NGU can only do x2 scaling (x4 too but what ever) and not all scaling operation are x2.
and on top of it quadrupling chroma is questionable...

Quote:
NGU-VeryHigh still performs quite slow on my 1080 GTX and drops frames down to 0-2 render queue, so it's still unuseable. Maybe in another 2 years when there's video cards fast enough. Same for Chroma. High works ok, but cuts it pretty close for 24p content. 30p and higher content is not possible with NGU because it can't render fast enough with any current graphics card. I never figured out how to make profiles, so I just make one setting for all content.
my 1060 can do 1080p23 -> 2160p23 NGU very high luma only without problems.
huhn is offline   Reply With Quote
Old 5th February 2017, 04:19   #42253  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Quote:
Originally Posted by madshi View Post
Which do you prefer?
NICE. The low preset is terrible though IMO, I would remove it as the improvement from low to medium is quite considerable. I need to play with it more to state which I prefer, it's going to take some time..
I still do prefer the slightly softer more natural look of NNEDI3, but a softer downscaler or soften edges 1 might make it comparable.
Any possibility for control of sharpness within the algorithm or do we have to rely on additional resizers and postprocessing?

Last edited by ryrynz; 5th February 2017 at 04:26.
ryrynz is offline   Reply With Quote
Old 5th February 2017, 04:26   #42254  |  Link
AngelGraves13
Registered User
 
Join Date: Dec 2010
Posts: 254
Quote:
Originally Posted by huhn View Post
my 1060 can do 1080p23 -> 2160p23 NGU very high luma only without problems.
I have a 1440p screen, so it has to scale back down and the render times are about +-35ms, which is far too high for 1080p content, as it can sometimes spike past 42ms, throwing off audio sync.

Guess I'll stick with either Jinc or Super-XBR.

Looking forward to getting that 27 inch 4K HDR monitor this year from Asus

Last edited by AngelGraves13; 5th February 2017 at 04:44.
AngelGraves13 is offline   Reply With Quote
Old 5th February 2017, 04:46   #42255  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
still no problem for my 1060 and quite some headroom left.
huhn is offline   Reply With Quote
Old 5th February 2017, 10:23   #42256  |  Link
Sunset1982
Registered User
 
Join Date: Sep 2014
Posts: 280
Is there a way to set Hotkeys for different profiles?

If there will be a new NGUsmooth variant for lower quality sources, it might make sence to be able to change the profile ota when watching high or low quality material.
__________________
Intel i5 6600, 16 GB DDR4, AMD Vega RX56 8 GB, Windows 10 x64, Kodi DS Player 17.6, MadVR (x64), LAV Filters (x64), XySubfilter .746 (x64)
LG 4K OLED (65C8D), Denon X-4200 AVR, Dali Zensor 5.1 Set
Sunset1982 is offline   Reply With Quote
Old 5th February 2017, 10:44   #42257  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,342
Quote:
Originally Posted by Sunset1982 View Post
Is there a way to set Hotkeys for different profiles?
Check the settings of the profile groups, you can assign a keyboard shortcut there.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders

Last edited by nevcairiel; 5th February 2017 at 10:57.
nevcairiel is offline   Reply With Quote
Old 5th February 2017, 11:17   #42258  |  Link
Sunset1982
Registered User
 
Join Date: Sep 2014
Posts: 280
Quote:
Originally Posted by nevcairiel View Post
Check the settings of the profile groups, you can assign a keyboard shortcut there.
ah ok, thx. Is there a madvr key table anywhere, to see what keys are free and which are used?
__________________
Intel i5 6600, 16 GB DDR4, AMD Vega RX56 8 GB, Windows 10 x64, Kodi DS Player 17.6, MadVR (x64), LAV Filters (x64), XySubfilter .746 (x64)
LG 4K OLED (65C8D), Denon X-4200 AVR, Dali Zensor 5.1 Set
Sunset1982 is offline   Reply With Quote
Old 5th February 2017, 12:07   #42259  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by huhn View Post
i think the general lines on NGU pixart low are to thin it get's better with medium but with very high the background lines are really strange/uneven and nnedi3 wins there hands down.
even NGU veryhigh wins there most of the part the line thickness is really good! but there or other obvious issues with NGU very high.
mario him self looks "best with pixart very high. his eyes are to small but compared to super XBR they are pretty good.

i have a huge problem selecting a winner here...
Quote:
Originally Posted by ryrynz View Post
NICE. [...] I need to play with it more to state which I prefer, it's going to take some time..
I still do prefer the slightly softer more natural look of NNEDI3, but a softer downscaler or soften edges 1 might make it comparable.
Any possibility for control of sharpness within the algorithm or do we have to rely on additional resizers and postprocessing?
Have you tried zooming into the images to see the differences more clearly?

For example, here are NNEDI3-256 and NGU pixart med at 300% zoom. Where does NNEDI3 look more natural here? The "M" in "TIME" looks better with NNEDI3-256. In every other part of the image, I personally prefer NGU pixart med. NNEDI3 also has a lot of very noticeable artifacts. And btw, NNEDI3-256 is about 3000% slower compared to NGU pixart med (!!!).

Or let's look at a photo instead of a game screenshot. The low-res photo was downscaled using a box filter so it has more aliasing than usual. Here's again how NNEDI3-256 and NGU pixart med compare.

I suppose I could offer even softer NGU pixart variants, but is it really needed? I posted a pixart image comparison with a very old very aliased game screenshot because that's really the hardest image type to get upscaled nicely. I think with real world images NGU and NNEDI3 look very similar.

Would be great if you could do some real world comparisons.

Quote:
Originally Posted by huhn View Post
so my last standing argument is that NGU is "hit or miss". super XBR works with everything.
Are NNEDI3 and NGU pixart "his or miss", too? I think they beat super-xbr pretty much all the time.

I do not want to leave Polaris users behind, though, so I might keep super-xbr for the time being, just because it's probably still noticeably faster than NGU pixart for Polaris users.

For comparison sake, here are the super-xbr 300% Mario zoomed image, and the clown zoomed by super-xbr. Look at the front wheel in the clown image. It's still quite aliased when using super-xbr. Less so with NNEDI3-256, and even less so with NGU pixart med.

Quote:
Originally Posted by ryrynz View Post
The low preset is terrible though IMO, I would remove it as the improvement from low to medium is quite considerable.
It's true that "medium" is a noticeable improvement over "low", but it's also a bit slower. There may be users whose GPU can perform "low" but not "medium". So why remove "low"? Is it really *that* terrible? If I compare it to NNEDI-16, IMHO NGU pixart "low" competes just fine. Or try comparing "low" to the regular (non-pixart) NGU variant!
madshi is offline   Reply With Quote
Old 5th February 2017, 12:07   #42260  |  Link
cork_OS
Registered User
 
cork_OS's Avatar
 
Join Date: Mar 2016
Posts: 160
New NGU med+ is definitely outperform sXBR in terms of quality and could replace it completely. NGU smooth also looks even a bit softer than NNEDI3.
However, right now it's hard to talk about speed because NGU smooth a) doesn't have high mode; b) seems doesn't have 4x mode.
Personally I rarely use NNEDI3 and I have nothing to say in it's favor. But why it's so important to drop out NNEDI3 instead of drop out something like Spline?
Additionally, it's still a big difference in sharpness between sXBR/NNEDI3/NGU smooth and plain NGU. Maybe there is a room for some intermediate NGU variant?
__________________
I'm infected with poor sources.
cork_OS is offline   Reply With Quote
Reply

Tags
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 16:34.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.