madVR - high quality video renderer (GPU assisted)

[Razoola]

Not here, in fact, in my Win7 Ult dual-mon situation w/FSE on the secondary mon, it is the most stable and performant video playback setup that I can muster amongst Linux or Windows. Hands down, jRiver with madVR (FSE) is untouchable.

Perhaps detail your hw/sw combination a bit more.
I have to agree

[pie1394]

Backlight scanning (or strobing) really does make a huge difference to perceived motion.


That is why I love my 65HX920's MotionFlow Clear+. The 960Hz + averaged 50% time of black-bar light-strobing has done a good job. :p

I hate the interpolated images on this 240Hz panel set since the object's motion feeling is still quite weird with them.


Anyway there is a good news from Sony that Full-array local dimming set comes back again. You might want to consider XBR-65/84X950B KDL-65/84X9500B this year -- if the price is afforable to you. :D

Sony engineers seemed to notice your complaint about the local dimming system on HX920/HX950. It is heard the full-screen unified control algorithm is used again. But I don't know if it is based on HX909's design.

This set equips a new back-lit module system, which the light-emitting efficiency is 100% more than tranditional one used on most 2013 TV sets.

Sony also claims even better HDR on the new dimming system. Peak light should be at least 1000 cd/m^2 (I guess ... since it is claimed 3x as bright as conventional sets). Not quite sure how dark it can improve from 65HX920.

About the panel supplier, it can be either SLCD S-PVA or Innolux S-MVA since both companys can provide 65" 84" 4K 120Hz panel. If it is a 240Hz 4K panel design, it has higher chance to be SLCD S-PVA.

[Plutotype]

Guys,
Can somebody please advise, why is madVR and Reclock detecting suddenly 112Hz instead of 60Hz on my setup? The weirdest thing is, that after some PC restarts, the detected Hz went out ok with 60Hz and also the display rr switcher worked well. But next restarts showed again 110/111/112/113.xxx Hz? Obviously you cant play back any video when refresh rate is like this. What could be messed up?
http://s1281.photobucket.com/user/Plutotype/media/rrissue_zps4ea455a6.png.html?sort=3&o=0

Thanks for any hint

[turbojet]

Plutotype: this occasionally happens to me when switching displays. Restarting aero fixes it.

net stop uxsms
net start uxsms

or restart desktop windows manager in services.msc.

I don't know why this happens, it never used to, it started happening shortly after I reinstalled windows 7 but there was also a madvr update around the same time.

[Plutotype]

Thanks Turbojet, restarted Aero, but the problem did not disappear. I still see 112.43121 Hz...weeeeird...
Should I buy new graphics card or you think its a "Windows" problem? Why is the display running at 1920x1080 @112Hz if it has a cap at 60Hz?

[noee]

Have you tried to reset your Reclock timings database?

[tp4tissue]

Try reduce the present frame in advance or queue, had similar problem on 1440p.

:D

That worked.. seems my gfx card is weak sauce, 7970 can only handle 12 frames...

[6233638]

That is why I love my 65HX920's MotionFlow Clear+. The 960Hz + averaged 50% time of black-bar light-strobing has done a good job. :p

I hate the interpolated images on this 240Hz panel set since the object's motion feeling is still quite weird with them. It's funny you should say that, because the "Clear" modes are still using intepolation. Only the "Impulse" mode does not.

However, the Clear/Clear Plus modes are tuned for film playback, and I find that they rarely produce artificial looking motion compared to the other MotionFlow modes. (or interpolation systems from other manufacturers)

It's a good balance between interpolation (smoothing out film judder) and backlight scanning. (improving motion clarity)
The Clear modes aren't perfect, but I prefer to watch films with the judder reduction that interpolation brings. "Pure" 24fps is simply too low a framerate.

[6233638]

Going back to a much earlier discussion, I think madVR should be displaying 16-240 by default, rather than 16-235.
Perhaps it is a coincidence, but I've noticed a couple of films recently which have exhibited clipping.

There are actually values all the way up to 255 in this example, but nothing I would consider important picture information.

http://www.abload.de/thumb/235bqf0r.png (http://www.abload.de/img/235bqf0r.png) http://www.abload.de/thumb/2405hi2f.png (http://www.abload.de/img/2405hi2f.png) http://www.abload.de/thumb/2543cdaj.png (http://www.abload.de/img/2543cdaj.png)

http://www.abload.de/thumb/16-235w8dzo.png (http://www.abload.de/img/16-235w8dzo.png) http://www.abload.de/thumb/16-2400titg.png (http://www.abload.de/img/16-2400titg.png) http://www.abload.de/thumb/16-2554ce3d.png (http://www.abload.de/img/16-2554ce3d.png)

Considering that valid luma ranges are 16-235, and valid chroma ranges are 16-240, I think 16-240 might be a good idea as the default.

If anyone wants to use 16-240 with the current builds of madVR, setting the output to 0-250 rather than 0-255 will do this. (though I had to use 0-249 with my 3DLUT enabled)


This film also has very visible compression artifacts in the black bars at times, so here's a sample if you're still looking for them.

http://www.sendspace.com/file/3jirbq

[pie1394]

It's funny you should say that, because the "Clear" modes are still using intepolation. Only the "Impulse" mode does not.

However, the Clear/Clear Plus modes are tuned for film playback, and I find that they rarely produce artificial looking motion compared to the other MotionFlow modes. (or interpolation systems from other manufacturers)



I am curious if it is actually processed with "Sharpening" instead of interpolation with Clear and Clear+ modes ? For the 24 fps contents, I don't really feel the motion is as smoothy as the native 60 fps or 24->60 FRC contents.

[EDIT]
You are right about frame interpolation engagement in Clear and Clear+ modes. But do you know what kind of complex 24fps film examples can make the HX920 show the disrupted object on motion?

[huhn]

would a tv normally clip that too ? i mean they always treat input as 16-235

Considering that valid luma ranges are 16-235, and valid chroma ranges are 16-240

were did you get the part about chroma?

and if a 16-235 source is encoded isn't it normal that the encoder produce artifacts in 0-16 and 235-255 i mean there is no rule to not use theses bits in encoding and if that helps in compression why not ?

[Werewolfy]

I am curious if it is actually processed with "Sharpening" instead of interpolation with Clear and Clear+ modes ? For the 24 fps contents, I don't really feel the motion is as smoothy as the native 60 fps or 24->60 FRC contents.

[EDIT]
You are right about frame interpolation engagement in Clear and Clear+ modes. But do you know what kind of complex 24fps film examples can make the HX920 show the disrupted object on motion?

Almost all movies with subtitles. It's the reason why I only use "True Cinema" on my Sony W905 even if it doesn't change anything. The stuttering doesn't bother me, I almost never notice it and when it's the case I would have to set Motionflow on Smooth to have a perfect motion.

Except with subtitles, I've also noticed a lot of interpolation artefacts in the TV show Teen Wolf (season one) where you have a scene where light pass through a fan. The result is really bad with Motonflow on. It doesn't works well either with a lot of anime.

I asked for help in the Avisynth forum for writing a BFI (black frame insertion) script and finally got results.
link: http://forum.doom9.org/showthread.php?t=170056

I just have to put my experience in the madvr board too.




Now I know why when I watch movies in the cinema it looks a lot clearer/sharper and less blurred.

I don't feel the necessity to have a deblur feature on 24 fps movies, the frame rate is so low that the the motion blur is pretty minimal and you will see double imaging instead. And yes, I tested it with the site you provided.

You spoke about the Hobbit but it's a 48 fps film that has always been shot with that in mind that you're looking in 24 fps. I don't think you will find a 24 fps movie where you can't read a text in motion because the directors know 24 fps is too low and keep that limitation in mind.

I'm pretty sure too that BFI has to work with backlight to be effective and bearable. Impulse mode works well, it doesn't flicker too much but the light output dipped significantly even with backlight and contrast bumped up to maximum.

[James Freeman]

@Werewolfy
The BFI was just an experiment not something I actually use.
Yes, a 24 movie looks more like image doubling than blur.
Yes, Backlight strobing will do a way better job.
The Hobbit was shot at 48, but convertet to 24 to look like 24. (blur added).

IMO, Double Imaging looks worse than blur (its wider),
the BFI was just to see how it would look like without the LCD "sample-and-hold" side effect, but with a lot of flicker (ie: not something I actually use).
Leaving the flickering aside, the BFI motion looks AMAZING as I already stated, clean and clear.
Most of the motion blur that I saw all those years and though was part of the movie (camera shutter speed), was actually ugly "sample-and-hold" LCD side effect.
For what Its worth, I'm waiting for a backlight strobing IPS panel, which isn't so far according to Mark Rejhon form BlurBusters.com.

@6233638
One of my favorite reference blu ray "Samsara" has native blacks at 7 and there is a lot of information there.
I never checked this until now.
Strange... probably a bad (non standard) transfer.

On the other hand most of the movies (that I tested) rarely reach 235 on the histogram.
Besides, you eyes are a lot less sensitive at brighter scenes compared to darker, ie: natural gamma curve of the eye.

[MokrySedeS]

The Hobbit was shot at 48, but convertet to 24 to look like 24. (blur added).

"Converting 48fps to 24fps was done simply by discarding every other frame, a process called ‘disentangling’ (which is also what it is called when you extract a 2D video file from a 3D one where the image for both eyes is recorded one after the other in the same stream. Because it is effectively the same thing)."

“Many frame blending tests were done with different systems, but we tended to like the results we got from simply discarding every other frame.”

Source: http://www.definitionmagazine.com/journal/2013/2/19/winning-the-frame-game.html
Sorry for being off topic...

[huhn]

@Werewolfy
The BFI was just an experiment not something I actually use.
Yes, a 24 movie looks more like image doubling than blur.
Yes, Backlight strobing will do a way better job.
The Hobbit was shot at 48, but convertet to 24 to look like 24. (blur added).

IMO, Double Imaging looks worse than blur (its wider),
the BFI was just to see how it would look like without the LCD "sample-and-hold" side effect, but with a lot of flicker (ie: not something I actually use).
Leaving the flickering aside, the BFI motion looks AMAZING as I already stated, clean and clear.
Most of the motion blur that I saw all those years and though was part of the movie (camera shutter speed), was actually ugly "sample-and-hold" LCD side effect.
For what Its worth, I'm waiting for a backlight strobing IPS panel, which isn't so far according to Mark Rejhon form BlurBusters.com.


so you like tons of over drive artifacts?

and there is a backlight strobing vs panel out which should be far better than any ips 8 bit panel. the eizo foris fg2421.

btw you can see flicker with a 120 hz monitor and bfi but ... i stop om this topic i mean this is not even worth calling a picture...

[iSunrise]

madshi, since your very first deband test build (compared to 0.86.7 and 0.86.11), there is something wrong with the display refresh rate detection.

0.86.11 and all versions before that, have an extremely stable detection and only minimal deviation, while something changed in madVRdeband.rar (madVR.ax from 09/30/2013), where the display refresh rate shown in the OSD is going nuts on my system.

I can reproduce this, if I just install 0.86.7 or 0.86.11 and replace only the madVR.ax included in the first madVRdeband.rar file. If I revert that change back to 0.86.7, everything is fine again.

And now the weird part, this seems to only affect the windowed modes (windowed and windowed fullscreen) and NOT exclusive fullscreen.

Since these are all inofficial releases, I am not sure if you need me to also use bug tracker for this. But I do if you want.

EDIT: You can ignore that madshi, since you already stated (http://forum.doom9.org/showthread.php?p=1661244#post1661244) that this is already fixed in your latest sources.

[flashmozzg]

It was stated by madshi that this problem is already fixed in latest sources few pages ago.

[madshi]

madshi: Before I forget, when you have some time would you be able to look at interlaced detection and reducing cadence breaks or handling them better. Both of these should be detected as interlaced, but aren't, vlc and potplayer auto detect correctly. Also both have cadence breaks, this one (https://www.mediafire.com/?mntpfnsgaalms97) has 1 that's noticeable whereas tivtc and decomb handle it fine. I'm guessing it's an issue with near static frames. Here's a more extreme sample (https://www.mediafire.com/?mntpfnsgaalms97) that tivtc, decomb don't handle all that well but it would be nice if madvr was able to handle it as well as they do.
I'm confused by your links. They seem to be identical, but you seem to be talking about different clips. Also you're saying they're not detected as interlaced, but they are on my PC. I can see some combing artifacts. Could be a cadence misdetection, but really, the sample is very very short. Almost too short to be useful. Can you provide me with a longer sample? Ideally where cadence detection first works well and then at some point stops working?

I've see where you want us to use the bug tracker, and I was halfway through that process already, but don't want to fill up the tracker unless I know for sure what I have is a bug :o

all my bluray rips are from my own real blurays and all ripped with my one default dvdfab profile, that is to say they're all pretty much identical. some videos are starting paused and some don't. it's always the same videos that start paused and the same ones that don't. as an example Game of Thrones season 2 episode 1 starts normally but season 2 episode 9 starts paused. other shows and movies do this too but what they are is probly insignificant.

so I dunno, should I fill out the bug tracker with this?
The bug tracker is for issues which are likely to be bugs in madVR, and which occur for more than just one user. If you're not sure, it makes sense to ask here first. In your specific case we should first check a few things, because I don't remember anyone else reporting your issue recently. So it could be a misconfiguration or some other filter making problems.

Could the problem have to do with the subtitles somehow? E.g. playback starting if there are no subtitles and not starting when there are some?

This problem is probably related to the "delay playback start until queues are full" option in the madVR settings.

madVR's exclusive mode does not work well on dual monitors. The exclusive mode keeps focus on itself and is impossible to use the other monitor as any click, scroll operations goers straight to MPC-HC. Sometimes it freezes the other monitor.
What happens if you start a fullscreen exclusive game on the secondary monitor? Can you then use the primary monitor just fine, without any issues?

Since there seems to be quite some talk about the new Oculus Rift HD prototype, called Crystal Cove and a new developer SDK, I am getting very eager and excited to buy that or an even better version, as soon as it is available. Now, what I have in mind is not just playing games, but also watching movies or other stuff on it.

What exactly would have to be done to make movies watchable on such a device?
I've no idea. Maybe standard 3D playback will automatically work with the Oculus Rift. If so, you'll still have to wait for whenever madVR gets 3D playback support. If not, you might be out of luck, because I'm not willing to work with a custom SDK for some custom display device I don't even own myself.

why is madVR and Reclock detecting suddenly 112Hz instead of 60Hz on my setup? The weirdest thing is, that after some PC restarts, the detected Hz went out ok with 60Hz and also the display rr switcher worked well. But next restarts showed again 110/111/112/113.xxx Hz? Obviously you cant play back any video when refresh rate is like this. What could be messed up?
Reclock misdetects the refresh rate, too? That sounds weird. Might be a problem with your GPU hardware or drivers or BIOS or whatever. I've no idea. FWIW, the latest test build (for debanding) has some problems with reliable refresh rate detection. But the current official version has no known issues with identifying refresh rates properly.

Going back to a much earlier discussion, I think madVR should be displaying 16-240 by default, rather than 16-235.
Perhaps it is a coincidence, but I've noticed a couple of films recently which have exhibited clipping.

There are actually values all the way up to 255 in this example, but nothing I would consider important picture information.

Considering that valid luma ranges are 16-235, and valid chroma ranges are 16-240, I think 16-240 might be a good idea as the default.
Chroma using 16-240 has nothing to do with this. The chroma channels should only affect color, but not brightness. The official standards clearly state how to interpret the data and how to display it properly. Darkest black is 16. Peak white is 235. If there's content out there using WTW then it's not encoded correctly.

That said, madVR does not clip anything. BTB and WTW are passed to the display, if you configure madVR to output video levels. So you can calibrate the display to show some of the WTW range, if you prefer. Or of course you can use custom levels in madVR, as you mentioned yourself.

I'm not willing to default to non-standard behaviour, just because some content isn't encoded properly. If *ALL* content were encoded with clearly exactly 16-240, that would be a different topic. But it is not. If I changed madVR to display 16-240 by default, we would lose brightness and contrast for properly encoded material. And madVR's output would then be different to what a properly working video player or external hardware Blu-Ray player would produce. That does not sound like a good idea to me.

This film also has very visible compression artifacts in the black bars at times, so here's a sample if you're still looking for them.
Thanks, will have a look.

[iSunrise]

It was stated by madshi that this problem is already fixed in latest sources few pages ago.
Thanks for mentioning it, I´ve edited my post.

[Plutotype]

Reclock misdetects the refresh rate, too? That sounds weird. Might be a problem with your GPU hardware or drivers or BIOS or whatever. I've no idea. FWIW, the latest test build (for debanding) has some problems with reliable refresh rate detection. But the current official version has no known issues with identifying refresh rates properly.


Its more crazy than that. Today I started the PC and got this

http://s1281.photobucket.com/user/Plutotype/media/RRproblemReclock_zps0410cb96.jpg.html

Then I restarted the PC again and got this:
http://s1281.photobucket.com/user/Plutotype/media/RRiisue2_zps4fd78601.png.html

Tried to re-flash BIOS but it didnt help. Also I tried to clean-up the Reclock databases, no success. Will try to buy a new card probably.

[flashmozzg]

Its more crazy than that. Today I started the PC and got this

http://s1281.photobucket.com/user/Plutotype/media/RRproblemReclock_zps0410cb96.jpg.html

Then I restarted the PC again and got this:
http://s1281.photobucket.com/user/Plutotype/media/RRiisue2_zps4fd78601.png.html

Tried to re-flash BIOS but it didnt help. Also I tried to clean-up the Reclock databases, no success. Will try to buy a new card probably.
Has this always been like this? Have you tried updating your drivers?

[Plutotype]

Has this always been like this? Have you tried updating your drivers?

Im on CCC 13.12, tried 13.4, but it didnt help. It started a couple of weeks ago, but it was like 1 out of 10 PC starts the refresh rate was incorrect. I didnt pay attention to it, but after Christmas ( PC stayed off for 3 weeks during the holidays ) it turned out to be screwed 9 out of 10 PC starts. The PC has been in operation more than 1000 days now ( 3.5year ).

[madshi]

Since both Reclock and madVR show the same problem, it's likely to be a real driver or hardware problem. The code used by Reclock and madVR to calculate the refresh rate is totally different. If both come to the same conclusions then probably your GPU is doing something weird. You could try using a different output port (e.g. try the other DVI port, or a HDMI -> DVI converter or anything else like that). If you're lucky maybe just one output port is fried or something...

[turbojet]

Clipboard issue, extreme cadence sample (http://www.mediafire.com/download/t5i0gbt4246u5i5/cadence_extreme.mpg) perfect ivtc is probably impossible but tivtc shows it can be much better.

There's wrong frames removed after the third scene change in the cadence clip. It usually happens during/after near-static frames. Decomb has issues with this in more extreme cases while tivtc doesn't. This explanation (http://forum.doom9.org/showthread.php?p=1658871#post1658871) might be why tivtc doesn't. How long of a clip is needed?

Both of these clips are 'in doubt' interlaced to madvr. When enabled it deinterlaces but also deinterlaces already deinterlaced 30fps. When disabled it doesn't deinterlace either.

EDIT: Here's a clip (http://www.mediafire.com/download/6arvdbr58irhbjd/combing.mpg) that uses duplicate frames to give a slow motion effect. Some interlaced frames get through and most of the time it's an unknown cadence when it should be 3:2. Here's what tivtc did with it (http://www.mediafire.com/download/q9mpll8pvi8u2fl/combing-tivtc.mkv), nevermind the ugly frames, it's a broadcast problem. Would it make sense for madvr to always look for 3:2 pulldown when ivtcing with an optional mode to look for different patterns? Except for some old silent movies, the rare 24 sped up to 25 by inserting dupe frames and pal converted to 30fps using dupe frames. 3:2 should be every other case I can think of and much more common.

[Plutotype]

Since both Reclock and madVR show the same problem, it's likely to be a real driver or hardware problem. The code used by Reclock and madVR to calculate the refresh rate is totally different. If both come to the same conclusions then probably your GPU is doing something weird. You could try using a different output port (e.g. try the other DVI port, or a HDMI -> DVI converter or anything else like that). If you're lucky maybe just one output port is fried or something...

Thanks for the input, tried DisplayPort to DisplayPort and the issue persists. Also confirmed on my SONY TV per HDMI. Will buy a new GPU.

[turbojet]

Plutotype: Have you considered installing windows on a different partition (to save current) to see if it has the issue?

While your issue is a bit different then mine, it goes lower vs higher refresh rate. I'm pretty sure windows install did it for me.

[Plutotype]

Plutotype: Have you considered installing windows on a different partition (to save current) to see if it has the issue?

While your issue is a bit different then mine, it goes lower vs higher refresh rate. I'm pretty sure windows install did it for me.

No I didnt, but I could try to boot from a backup SSD ( 2012 image ).

[noee]

No I didnt, but I could try to boot from a backup SSD ( 2012 image ).

You could also boot a live USB image of some linux distro and see what xrandr or DisplayConfig say.

[Werewolfy]

Thanks for the input, tried DisplayPort to DisplayPort and the issue persists. Also confirmed on my SONY TV per HDMI. Will buy a new GPU.

Have you overclocked your CPU or GPU? I had once similar problems with Reclock, it was due to the way I overclocked my CPU.

[flashmozzg]

Im on CCC 13.12, tried 13.4, but it didnt help. It started a couple of weeks ago, but it was like 1 out of 10 PC starts the refresh rate was incorrect. I didnt pay attention to it, but after Christmas ( PC stayed off for 3 weeks during the holidays ) it turned out to be screwed 9 out of 10 PC starts. The PC has been in operation more than 1000 days now ( 3.5year ).

Try testing this on clean install of windows. Sometimes some programs (like desktop viewers/remote desktop/skype or others) can mess things up.

[Plutotype]

Have you overclocked your CPU or GPU? I had once similar problems with Reclock, it was due to the way I overclocked my CPU.
Thanks its stock 3.33Ghz i7-980X, no overclocking on CPU/GPU.

[Plutotype]

Try testing this on clean install of windows. Sometimes some programs (like desktop viewers/remote desktop/skype or others) can mess things up.

I will try to do repair install first. Thanks.

[6233638]

Chroma using 16-240 has nothing to do with this. The chroma channels should only affect color, but not brightness. The official standards clearly state how to interpret the data and how to display it properly. Darkest black is 16. Peak white is 235. If there's content out there using WTW then it's not encoded correctly.Well it's only the red channel which is clipping, which is why I thought it was related to chroma. Expanding the visible range from 16-235 to 16-240 fixes most of it. I agree that content should not have video information in the BTB/WTW region.

I'm not willing to default to non-standard behaviour, just because some content isn't encoded properly. If *ALL* content were encoded with clearly exactly 16-240, that would be a different topic. But it is not. If I changed madVR to display 16-240 by default, we would lose brightness and contrast for properly encoded material.Perhaps I was misunderstanding how the 16-240 chroma range affects madVR's RGB output then.
Should values of 240 in chroma still end up in the 16-235 range when converted to RGB?

[pie1394]

I will try to do repair install first. Thanks.

Just want to know if you set the VSync always ON if (3D) application is not specified...

If I don't remember it wrong, madshi had mentioned those VSync OFF or adaptive-frame-rate mechanisms are actually only good for gaming purpose.. but bad for video playback...

[Ver Greeneyes]

If I don't remember it wrong, madshi had mentioned those VSync OFF or adaptive-frame-rate mechanisms are actually only good for gaming purpose.. but bad for video playback...
I think it depends on whether you're running in windowed or fullscreen exclusive mode. In windowed mode for Windows (Vista and up), the compositor handles VSync for you, so you don't want to end up in a race condition. If you only present during VBlank in windowed mode with the compositor enabled, you might be just in time for composition one frame and miss it the next, causing lots of duplicate and skipped frames. One way to deal with this is to present after a call to DwmFlush instead - then there should always be enough time for the compositor to pick up on your frame. If you know you're going to present with stable frame rate, another way is to use DwmSetPresentParameters to make the compositor aware of it, though that might increase the latency. In their infinite wisdom, Microsoft also dropped support for this in Windows 8.

[madshi]

Perhaps I was misunderstanding how the 16-240 chroma range affects madVR's RGB output then.
Should values of 240 in chroma still end up in the 16-235 range when converted to RGB?
It depends on the Y value, too. With the right combination of Y and chroma values even a chroma value of 240 should still end up in the 16-235 RGB range.

But now that you mention it, there might actually be a small bug in the madVR math, I'm not totally sure right now, I'll double check.

[PetitDragon]

IIRC, when Y: 16-240, CbCr should be 13-243.

[bacondither]

@madshi

Do you have any plans to incorporate a sharpening algorithm pre/post image resizing like LimitedSharpenFaster(mod) or any other algorithm that can improve the MTF (http://www.normankoren.com/Tutorials/Summary_intro.gif)without introducing too much under or overshoot.

[Stereodude]

Does anyone know what Nvidia card has equivalent support for all the advanced madVR processing that the ATI HD 7790 is capable of?

The nice charts at AVSforum put together by renethx (http://www.avsforum.com/t/1477339/so-youve-built-your-htpc-now-what-is-next-how-to-get-the-ultimate-picture-and-sound-quality-from-your-htpc-madvr-svp-xbmc-mediabrowser-jriver/60#post_23512510) don't go above the GTX 650 Ti and it doesn't quite have the capability of the HD 7790. I guess that means the GTX 650 Ti Boost or GTX 660 / GTX 660 Ti would be the equivalent, but that's speculation on my part.

[6233638]

Does anyone know what Nvidia card has equivalent support for all the advanced madVR processingThe problem is that madVR is a moving target. What might be the highest quality scaling options today (Jinc 3+AR) may be replaced in the future with much more demanding algorithms.
New features may be introduced (such as Debanding) which will also increase the workload for the GPU.

So what might be sufficient for the highest quality settings in madVR today, may not be tomorrow.

It seems like a 650Ti should be similar to (or better than) a 7790, but I don't know how madVR performance compares to the gaming tests most review sites use.
http://www.anandtech.com/bench/product/1035?vs=1040

[TheElix]

@madshi

Do you have any plans to incorporate a sharpening algorithm pre/post image resizing like LimitedSharpenFaster(mod) or any other algorithm that can improve the MTF (http://www.normankoren.com/Tutorials/Summary_intro.gif)without introducing too much under or overshoot.You mean like Darbee does?

[Stereodude]

It seems like a 650Ti should be similar to (or better than) a 7790, but I don't know how madVR performance compares to the gaming tests most review sites use.
http://www.anandtech.com/bench/product/1035?vs=1040The madVR specific test results I linked to in my post show the 650 Ti isn't quite equal to the 7790. There is an * in the level 5 column for FHD Video for the 650 Ti that isn't there for the 7790. In my experience Nvidia cards have better HW video processing mainly in terms of IVTC not being thrown off by station logos / commercials or other crap faded on and off the screen (broadcast TV). However, if the 650 Ti Boost or 660 Ti is the equivalent card to the HD7790 you take a fairly substantial price and power hit to go Nvidia.

I do realize it's a moving target at madshi could introduce new modes that need more processing power, but I'm just looking for relative parity between them.

[bacondither]

You mean like Darbee does?
Well... yes and no, Darbees example images from their webpage looks just lika a wide pixel radius unsharp mask. I applied a unsharp mask myself and it looked almost identical. Whats more funny is that the darbee example images are not soft they are almost aliasied and the wide pixel area unsharp mask acts almost like a contrast enhancer and not improving soft image transitions (http://docs.gimp.org/en/images/filters/examples/unsharp-mask-2.png).(green=original, black=prefered, red=ugly)

It would be fun to have a darbee to play with some testpatterns, but i would never buy one. Not when you can do the same with some lines of code in a PC for "free".

[turbojet]

Stereodude: Madvr's ivtc is no different on nvidia or intel, no amd to test but I think the way madvr avoids issues with animated overlays is it looks only in the center of the picture (by default). This is mostly good but when there's motion only on the edges of the picture it's guessing and leaves interlaced and duplicate frames, example (http://www.mediafire.com/download/14gc1pg9p3jlbc1/edge_motion.mpg). It also has problems deinterlacing these animated overlays, like this one (http://www.mediafire.com/download/8d1vdxmhjgk4k63/combed_bottom.mpg) While most avisynth filters have problems removing the wrong frames during animated overlays, IT.dll (http://www.avisynth.nl/users/warpenterprises/files/it_5F25_dll_20030712.zip) handles them well while using the whole picture but it has other problems, mainly during scene changes. Maybe some ideas can come from the source code.

bacondither: Why not try shaders like lumasharpen (http://dropcanvas.com/l46x9) it runs on gpu and while LSF does a good job at preventing aliasing, lumasharpen does just as well ime with the option to sharpen a lot more than lsf. It mainly sharpens lower contrast edges while not touching high contrast edges where aliasing would be noticeable but it does emphasize edge artifacts in the source.

[AlexCeed]

What happens if you start a fullscreen exclusive game on the secondary monitor? Can you then use the primary monitor just fine, without any issues?
It works well this way.

[bacondither]

bacondither: Why not try shaders like lumasharpen (http://dropcanvas.com/l46x9)...
Thanks! i'll try it. :)
I've played around with sharpen complex v2 and found the default to strong with edge artifacts, but with "CoefBlur 1" and "Sharpen_val0 1.2" i think i'ts a good compromise.

[huhn]

Stereodude: Madvr's ivtc is no different on nvidia or intel, no amd to test

ivtc runs on the cpu so yeah the gpu doesn't matter at all for ivtc.

[madshi]

Do you have any plans to incorporate a sharpening algorithm pre/post image resizing like LimitedSharpenFaster(mod) or any other algorithm that can improve the MTF (http://www.normankoren.com/Tutorials/Summary_intro.gif)without introducing too much under or overshoot.
At some point in the future, probably.

It works well this way.
Ok, that means that it's probably not the fullscreen exclusive mode itself which causes the problems on your PC. In that case it could be madVR's fault, or it could be the media player's fault. Which media player are you using? And does the problem also occur if you e.g. use EVR in fullscreen exclusive mode?

[Stereodude]

ivtc runs on the cpu so yeah the gpu doesn't matter at all for ivtc.I wasn't referring to madVR's IVTC. I was referring to the IVTC that can be done by the video card driver(s). I've found the Nvidia implementation superior to the AMD/ATI one.

[turbojet]

Stereodude: How do you plan on using nvidia's ivtc? Program, renderer, decoder, etc.