madVR - high quality video renderer (GPU assisted)

[6233638]

any thoughts about the impact of dithering on the motion interpolation algorithms in modern HDTVs?It shouldn't make a difference.

So, to the people who are reporting 0 frame drops even on long playbacks with smooth motion on, here's the one million dollar question: are you doing any special configuration?I have a GTX 570 and have done nothing special to avoid dropped frames.

Are you using Windowed, Overlay, or Full-screen Exclusive mode? I mostly use Windowed mode - though FSE works just as well. Overlay seems susceptible to interruption from other processes more than Windowed/FSE.


Are you using ReClock, JRiver Video Clock, or SmoothMotion?
If you are not using one of these, you are likely to always have dropped frames during playback.

Considering movies are usually two hours long, so you drop on the average 4-5 frames for one entire movie? :p That's about 0.2 seconds of frame dropping per ENTIRE movie.

On the other hand, blink of an eye takes 0.2 to 0.3 seconds. The average person blinks some 15-20 times per minute, or 1800-2400 times per a movie. According to the theory of probability, you chance of see a frame drop is pretty low.And yet I somehow manage to see it every time there is a frame drop, and I find even a single dropped frame very noticeable.

[Guest]

On the other hand, blink of an eye takes 0.2 to 0.3 seconds. The average person blinks some 15-20 times per minute, or 1800-2400 times per a movie. According to the theory of probability, you chance of see a frame drop is pretty low. Actually, the chance of seeing a dropped frame is very high, because your eyes are open much longer than they are closed. The chances that your blinks exactly match all the drops is minuscule.

[seiyafan]

i see. :D

[Asmodian]

About NNEDI: There's nothing wrong with high gpu usage, you missed my point :). I was saying that 60 fps with high NNEDI-settings don't work for me, so i need a profile, that chooses the settings according to the source video frame rate - which madvr can't display right.

This is SVP's fault not madVR's. SVP does not create or change the flags to tell madVR what the new frame rate is.


Are you using ReClock, JRiver Video Clock, or SmoothMotion?
If you are not using one of these, you are likely to always have dropped frames during playback.

Good point! I reported I see no dropped frames at completely default madVR settings (except with smooth motion on) but I am using ReClock. e-t172 is using smooth motion though.

*Every* single hardware and software component has been swapped. Completely fresh hardware with fresh Windows install and testing multiple driver versions. These frame drops keep haunting me everywhere I go. At this point I'm starting to wonder if I'm not becoming insane.

A long shot but how about PCI-E settings or SATA settings? Make sure you are using PCI-E 3.0 and SATA is set to AHCI or RAID. I have nasty system wide pauses if I leave the default IDE SATA mode on my x58 system.

Edit: It looks like your Gigabyte GA-Z87X-UD3H defaults to AHCI so that probably isn't it. :(

[StinDaWg]

Tuning for maximum image quality (only IMO of course) for 720p24 -> 2560x1440 I ended up with NNEDI3 128 Luma doubling which left room for Jin3AR chroma + image resizing. Going to 64 neuron luma doubling for 32 neuron chroma scaling or 16 neuron chroma doubling is not worth it. I do like NNEDI3 chroma; I think if I could run 128 luma doubling with NNEDI3 for the chroma too I would but then again I haven't tested 256 neuron luma doubling. ;)
What card is this with?

[Asmodian]

GTX Titan, with a bios mod to allow up to 350W before throttling and water cooled, runs at 1097 MHz.

[omarank]

@ Madshi: When playing a DCI video, I can see a colour cast in the image. Even when playing a Bluray video, if primaries are forced as DCI-P3 or "this display is calibrated to the following primaries/ gamut: DCI-P3" is selected, the colour cast can be seen in the image. I think it's due to a shift in the white point, if there is no bug in madVR (this doesn't happen with BT.2020). So I was wondering if madVR could also do white point matching of the source with the display white point while performing gamut conversion so that DCI content can be viewed on regular monitors without any colour cast in the image.

BTW, does 3dlut take the source white point as input or it always assume the source white point to be D65?

[nevcairiel]

So, to the people who are reporting 0 frame drops even on long playbacks with smooth motion on, here's the one million dollar question: are you doing any special configuration?

No, no special drivers, pretty much default madVR settings except the usual quality related changes (scalers, dithering, calibration,...), no power management tweaks.
It "just works".

[e-t172]

Considering movies are usually two hours long, so you drop on the average 4-5 frames for one entire movie? :p That's about 0.2 seconds of frame dropping per ENTIRE movie.

On the other hand, blink of an eye takes 0.2 to 0.3 seconds. The average person blinks some 15-20 times per minute, or 1800-2400 times per a movie. According to the theory of probability, you chance of see a frame drop is pretty low.

I can notice a frame drop very easily if it happens while something is moving on the screen. If it was not noticeable I wouldn't spend this much time trying to fix it.

Not sure if you've already tried or if it could make any difference whatsoever, but can you try with another display?

Yeah, that's what I'm going to try as soon as I get back from work.

Are you using Windowed, Overlay, or Full-screen Exclusive mode? I mostly use Windowed mode - though FSE works just as well. Overlay seems susceptible to interruption from other processes more than Windowed/FSE.

I'm always testing with the madVR defaults (except smooth motion). That includes FSE.

Are you using ReClock, JRiver Video Clock, or SmoothMotion?
If you are not using one of these, you are likely to always have dropped frames during playback.

I am always testing with smooth motion enabled precisely to avoid this problem.

Good point! I reported I see no dropped frames at completely default madVR settings (except with smooth motion on) but I am using ReClock.

Wait... so you're saying that if you enable smooth motion, you get frame drops?

A long shot but how about PCI-E settings or SATA settings? Make sure you are using PCI-E 3.0 and SATA is set to AHCI or RAID. I have nasty system wide pauses if I leave the default IDE SATA mode on my x58 system.

GPU-Z indicates PCI-E 3.0 x16. SATA has always been set to AHCI.

No, no special drivers, pretty much default madVR settings except the usual quality related changes (scalers, dithering, calibration,...), no power management tweaks.
It "just works".

You're seeing presentation glitches though. (http://forum.doom9.org/showthread.php?p=1674569#post1674569) Strangely enough, I never get presentation glitches except in one and only one configuration: 780 Ti combined with drivers >= 334.89 (and they happen quite often - every 5 minutes or so - in this configuration, with visible effects). With any other combination (different GPU or drivers <= 332.21), I still get one frame drop every 20-30 minutes, but never any presentation glitches. So if you want to get rid of your glitches you might want to try 332.21. My tests show quite clearly that presentation glitches are a regression introduced in 334.89.

[nevcairiel]

You're seeing presentation glitches though. (http://forum.doom9.org/showthread.php?p=1674569#post1674569) Strangely enough, I never get presentation glitches except in one and only one configuration: 780 Ti combined with drivers >= 334.89 (and they happen quite often - every 5 minutes or so - in this configuration, with visible effects). With any other combination (different GPU or drivers <= 332.21), I still get one frame drop every 20-30 minutes, but never any presentation glitches. So if you want to get rid of your glitches you might want to try 332.21. My tests show quite clearly that presentation glitches are a regression introduced in 334.89.

The glitches are only on my desktop PC with a multi-monitor setup, who knows whats happening there. My HTPC is glitch free and drop free for whole movies. I don't usually watch stuff on the desktop.
The glitches also only happend with smooth motion apparently, which I don't use usually, since my TV supports proper 24p.

(The glitches are also not visible to me, I wouldn't have noticed if I didn't look at the OSD)

[Asmodian]

Wait... so you're saying that if you enable smooth motion, you get frame drops?

No, simply that I was using defaults. Except smooth motion was on. No frame drops. Shouldn't have used "with". ;)

[cyberbeing]

@e-t172

Have you tried forcing HPET on in Windows?

Ensure HPET (High Performance Event Timer) is enabled in BIOS
bcdedit /set useplatformclock true
Reboot

Have you tried using Reclock + Smooth Motion together?

Have you tried installing your NVIDIA driver with "Message Signaled Interrupt (nv_msiSupport_addreg)" enabled instead of traditional IRQ interrupts?

Have you tried tweaking your monitor refresh rate slightly?


I don't have much first-hand experience with Smooth Motion over long durations, but the following are the settings I use without framedrops with exact multiple refresh rates on Win7 SP1 x64 on my GTX770 with software decoding and "Prefer Maximum Performance" enabled for mpc-hc.exe in driver profiles:

General (http://i.imgbox.com/WBVTygJ5.png) | Windowed (http://i.imgbox.com/iHhH9ERO.png) | Exclusive (http://i.imgbox.com/PUt4B5la.png)

[e-t172]

Thanks for the additional tricks, cyberbeing. I'll try them if swapping monitors doesn't work. Out of curiosity, are you using any of these tricks right now? Also, what driver version are you using?

[6233638]

I am always testing with smooth motion enabled precisely to avoid this problem.Does your display only support 60Hz?

If it supports 24Hz, try installing JRiver and using their VideoClock feature instead of SmoothMotion. (nothing to configure, unlike ReClock - you just tick a box)

Have you tried forcing HPET on in Windows?I would only do this as a last resort. It should not be necessary to do this on modern CPUs with Invariant TSC support. (I think this requires Windows 7 or newer?)

[wolfman2791]

You can pretty easily overclock a 60Hz monitor to 75Hz... i use this with my MSI GeForce 660 to overclock my monitor: http://3dvision-blog.com/wp-content/files/evgapixelclockoc.exe.

As far as other tweaks. I did the following:
bcdedit /set useplatformclock true
bcdedit /set tscsyncpolicy Enhanced
bcdedit /set disabledynamictick yes

Just type those on the command prompt (admin rights) separately.

useplatformclock forces the use of the platform clock as the system’s performance counter.
tscsyncpolicy controls the times stamp counter synchronization policy. The option are "default", "legacy"and "enhanced". I set it to "enhanced".
disabledynamictick enables and disables dynamic timer tick feature.

EDIT: I'll add that, if you play games on your PC, you use steam AND you're running Windows 8, you shouldn't overlock your monitor unless there's a setting in the games that allow you to run it at the monitor's native rate. Otherwise, there's a glitch that will cause the games to crash. Just a FYI.

[cyberbeing]

I would only do this as a last resort. It should not be necessary to do this on modern CPUs with Invariant TSC support.

While I would agree, e-t172 seems to quickly be reaching his last resort at this point. Invariant TSC isn't perfect, and can have minor +/- skew between reboots, while the HPET should always be perfectly stable if properly implemented on the motherboard. FWIW, I do force HPET on my i5-3570K which supports Invariant TSC because of such timer instability issues between reboots, HPET only lowered my DPC & Interrupt latency, as well as the HPET only timer being measurably more accurate in madVR. YMMV, since some systems seem to have the exact opposite experience with HPET for whatever reason.

By default Windows uses a hybrid TSC/HPET timer with syncing, while madVR always set a 0.5ms timer when available which requires HPET anyway. One frame drop every 30mins on 23.976 video, is equivalent to a +0.555ms skew, which essentially means e-t172's system is either missing a timer tick, or otherwise ends up with a 0.002% timer skew over time. Such a small deviation could be the fault of hybrid TSC/HPET, where forcing a single global HPET timer may be immune.

[e-t172]

That's quite interesting and I will definitely try that, thanks. Keep in mind, though, that this issue persists even after a complete motherboard and CPU swap (from completely different generations: I moved from Nehalem to Haswell). It would be very surprising that a timer skew issue would resist such drastic changes. However at this point anything would surprise me anyway, so I'll just add that to my list of weird tricks to try out.

One frame drop every 30mins on 23.976 video, is equivalent to a +0.555ms skew

I don't think your calculation is correct considering that I'm using smooth motion on 60Hz. The fact that the original source is 24p is probably irrelevant, I guess.

[cyberbeing]

I don't think your calculation is correct considering that I'm using smooth motion on 60Hz. The fact that the original source is 24p is probably irrelevant, I guess.

Yeah, I'm not exactly sure how madVR's frame drop logic works with Smooth Motion, since theoretically the blending should prevent frame drops. The original video frame rate is still relevant though, since those timestamps are used for Smooth Motion calculations along with your refresh rate measurements.

[GREG1292]

e-t172 : I had the same problems with evga precision x with a Gtx-770
but switched to MSI afterburner and problem solved. Only reason I figured
this out is because I switched from afterburner to precision x and it bonked
my playback. Of course if you are not using one of these programs it may not
help. Windows 8.1 pro no dropped frames with nnedi3 32. Good luck

[Mangix]

You can pretty easily overclock a 60Hz monitor to 75Hz... i use this with my MSI GeForce 660 to overclock my monitor: http://3dvision-blog.com/wp-content/files/evgapixelclockoc.exe.

EDIT: I'll add that, if you play games on your PC, you use steam AND you're running Windows 8, you shouldn't overlock your monitor unless there's a setting in the games that allow you to run it at the monitor's native rate. Otherwise, there's a glitch that will cause the games to crash. Just a FYI.

Wrong link. You'd get better results by using CRU since you can adjust the timings: http://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU

That's just wrong and sounds like a symptom of not using CRU. The EVGA Pixel Clock utility works by creating a custom resolution through the Nvidia Control Panel. It's just another GUI for it. What this means is that Windows treats this new higher refresh rate resolution as non-standard and unsupported by the monitor. This is why there are problems with video games.

CRU on the other hand edits the EDID and makes Windows think that 75Hz is a built-in resolution and thereforce problem solved. The only thing to make sure of if you want 75Hz in game is to launch the game when your desktop is at 75Hz. Works for me with Source games.

[wolfman2791]

Wrong link. You'd get better results by using CRU since you can adjust the timings: http://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU

That's just wrong and sounds like a symptom of not using CRU. The EVGA Pixel Clock utility works by creating a custom resolution through the Nvidia Control Panel. It's just another GUI for it. What this means is that Windows treats this new higher refresh rate resolution as non-standard and unsupported by the monitor. This is why there are problems with video games.

CRU on the other hand edits the EDID and makes Windows think that 75Hz is a built-in resolution and thereforce problem solved. The only thing to make sure of if you want 75Hz in game is to launch the game when your desktop is at 75Hz. Works for me with Source games. Brilliant!
Thank you.

[kalston]

You can pretty easily overclock a 60Hz monitor to 75Hz...

Easily if the monitor supports it yes. Not all LCDs do (I still find that a good number do but I haven't extensively tested all of them for framedrops or fake numbers) Some monitors also have issues other than framedrops when doing that, such as a corrupted display, weird scan lines etc. It's not THAT simple :)

It's really easy to test whether it works or not though. Fire up an old game (or well any game that runs at a super high framerate on your system and preferably a FPS) with v-sync on, run around with fraps and/or some other reliable fps meter... see what happens. There's other methods now but that's the one I'm used to (since I've been "overclocking" LCDs for quite a while).

And ye, use CRU or do a good old EDID override if you want to use it in games and such. My LCD for example is running at 76.5hz & 1080p* (I made Windows treat that as his native resolution) which works with absolutely everything, even applications that give you no options to change any advanced settings (most just run with your desktop resolution and refresh rate by default anyway).
*the max you can fit in DVI single link with a 165mhz pixel clock and tight timings.

And I added 71.928hz and 75hz for madVR of course.


I must say I haven't had any issues with video playback for a long time now (well, ever since I got my Nvidia GTX 670), using madvr/lav/JRiver MC. It's all been perfection thanks to madshi and nevcairiel (and the JRiver team for the audio part) :)

Oh, and the upscaling is absolutely fantastic (using settings mostly similar to 6233638) and it makes some old films very enjoyable to watch on my 22" LCD monitor, which is quite a feat considering how close to it I am sitting.

I haven't really tried the new NNEDI3 thingy though, maybe I should. Been using error diffusion option 2 (the one that says it produces less noise) but I'm not sure I can see much of a difference between the dithering settings anyway (my eyes may not be good enough for that kind of thing though)

[wolfman2791]

Yeah... i guess mine does :lol: I played Skyrim at 75 fps without a problem when it was totally wigging out without using CRU. I had to push it down to 60 fps to make it work before.

Have you noticed that the brightness increases with the higher refresh rates? i had to bring down my brightness a tad.

[seiyafan]

I think I found a bug. When I enable BOTH NNEDI3 chroma upscaling and luma doubling, I get frame drops (2 frames per second) when the rendering time is less than the dropping theshold (~41ms), however if I choose a chroma upscaling other than the NNEDI3, I don't get frame drop even if the rendering time is longer than that of the above, as long as it's less than the threshold.

[Mangix]

Not all LCDs do (I still find that a good number do but I haven't extensively tested all of them for framedrops or fake numbers)

In my experience, most LCD monitors do support 75Hz but not past that. I have two such monitors from ~2004 and ~2007 that are like that.

Modern LCDs *should* be able to go higher. CRU was developed as a tool for the cheap Korean monitors on eBay to increase the refresh rate to 120Hz.

[bugmen0t]

@cyberbeing
You use 'don't flush' for everything in FSE. I can hardly remember but didn't madshi said something needs 'flush & wait' to work properly, e.g. reporting dropped frames in the OSD or something else?

[Cinemancave]

Since I've been using NNEDI3 and Error Diffusion I have really noticed a jump in quality in my front projection setup. I have just purchased an Oppo 103D and plan to use its Darbee processing in combination with MadVR. Now I've just read that the Darbee in the Oppo automatically converts any signal to YcBcR 4:2:2, where it does its internal processing (I earlier thought that the processing was done in the color space that was sent into the unit, but appareantly only the standalone Darblet does this).

So my question is if using MadVR to output RGB to the 103D is worthless because of the 103D's internal color space conversion? Or am I still going to take advantage of some of MadVR's greatness? :confused:

[leeperry]

Darbee can't do anything PS scripts can't provide for free IMHO (http://darbeevision.com/darblet), and the latter work in much higher resolution(16bit RGB).

[Cinemancave]

Darbee can't do anything PS scripts can't provide for free IMHO (http://darbeevision.com/darblet), and the latter work in much higher resolution(16bit RGB).

What scripts are you referring to exactly? I have often seen these kinds of comments but never actually seen any references of them providing a better result than the Darbee. But I am all ears.

Otherwise, does anyone know how much the 103D reconversion is going to "damage" MadVR's results?

[cyberbeing]

@cyberbeing
You use 'don't flush' for everything in FSE. I can hardly remember but didn't madshi said something needs 'flush & wait' to work properly, e.g. reporting dropped frames in the OSD or something else?

Flush & Wait is needed if you want the madVR OSD to report Render & Present times. madVR doesn't use this information itself. Setting everything to 'don't flush' is equivalent to madVR only flushing in pre-determined critical sections, instead of being arbitrarily forced to flush at additional locations by end-user settings. NVIDIA GPUs with desktop composition enabled do not require any additional flushing to run smoothly, as of the r304 driver release from a couple years ago which contained numerous vsync stuttering fixes. The settings exist in madVR mainly for troubleshooting, with less flushing being preferred if your GPU can handle it without stuttering, glitches, or GPU queue abnormalities.

[*Touche*]

Darbee can't do anything PS scripts can't provide for free IMHO (http://darbeevision.com/darblet), and the latter work in much higher resolution(16bit RGB).

Ouch, their examples look terrible. I would never want to do that to my picture.

[leeperry]

What scripts are you referring to exactly? [..]
Otherwise, does anyone know how much the 103D reconversion is going to "damage" MadVR's results?
Considering what Darbee does to the picture, the RGB/YUV conversions aren't your biggest problem IMO.

Not too sure, when I want subtle EE I simply enable Gamma Light Dithering in mVR.....for strong halo-based EE you could try your luck with those 2 PS scripts packs: 1 (http://forum.doom9.org/showthread.php?t=170357) (LumaSharpen might be right up your alley) & 2 (http://forum.doom9.org/showthread.php?t=157634)

TMT went through the same kind of commercial fluff with their CUDA based sharpener (http://www.arcsoft.com/topics/media-player/what-is-simhd.html), long story short it didn't quite meet their promises.

What we all crave is Super Resolution (http://link.springer.com/chapter/10.1007%2F978-3-319-00951-3_5) and at this point NNEDI is as close as it's gonna get. Of course there were impressive software doing exactly that (https://www.google.com/search?q=vReveal+cuda&tbm=vid), but they still don't run in realtime: Ikena Forensic’s super-resolution reconstruction algorithm (http://www.motiondsp.com/products/ikena/super-resolution)

If you're willing to go Avisynth, this might help (http://forum.doom9.org/showthread.php?t=139250) too.

You can rest assured that the reviewers raving about Darbee were either on their payroll or very easy to impress.

[Cinemancave]

Considering what Darbee does to the picture, the RGB/YUV conversions aren't your biggest problem IMO.

Not too sure, when I want subtle EE I simply enable Gamma Light Dithering in mVR.....for strong halo-based EE you could try your luck with those 2 PS scripts packs: 1 (http://forum.doom9.org/showthread.php?t=170357) (LumaSharpen might be right up your alley) & 2 (http://forum.doom9.org/showthread.php?t=157634)

TMT went through the same kind of commercial fluff with their CUDA based sharpener (http://www.arcsoft.com/topics/media-player/what-is-simhd.html), long story short it didn't quite meet their promises.

What we all crave is Super Resolution (http://link.springer.com/chapter/10.1007%2F978-3-319-00951-3_5) and at this point NNEDI is as close as it's gonna get. Of course there were impressive software doing exactly that (https://www.google.com/search?q=vReveal+cuda&tbm=vid), but they still don't run in realtime: Ikena Forensic’s super-resolution reconstruction algorithm (http://www.motiondsp.com/products/ikena/super-resolution)

If you're willing to go Avisynth, this might help (http://forum.doom9.org/showthread.php?t=139250) too.

You can rest assured that the reviewers raving about Darbee were either on their payroll or very easy to impress.

Thanks for the tips, I'll look into some of them.

As for rest of the Darbee comments, no one serious about AV uses the thing at 100% effect. DO NOT look at their example pictures and judge it by them, they all look awful. Most seem to use it at 20-30% which doesn't introduce many artifacts at all, but enhances "perceived depth" in the image. I have a JVC X90 (RS65) and it's suppose to work well with it's MPC settings combined with the Darbee at lower levels (according to numerous people over at AVS, who take this hobby very seriously). The Darbee really is a way for me to try and squeeze every last bit out of my setup, even if it means only using it at a very low level.

[kalston]

Yeah... i guess mine does :lol: I played Skyrim at 75 fps without a problem when it was totally wigging out without using CRU. I had to push it down to 60 fps to make it work before.

Have you noticed that the brightness increases with the higher refresh rates? i had to bring down my brightness a tad.

Nope, mine doesn't do that. Like I said different monitors will behave differently when you overclock them.

Mine has no side effects besides a slight buzzing noise when displaying some content (such as this forum). It's very subtle though, I need to have my head close to it to hear it, but it is quite a bit louder at 75+hz compared to 60hz. Still too quiet to be an issue.

Btw don't play Skyrim at more than 60fps or you'll get messed up physics and the game's internal clock will go mad and bugs will appear. People playing on 120hz monitors with vsync found out the hard way :p 75 isn't a big increase from 60 but over long gaming sessions you'll still end up having issues because of how silly this game engine is (thank god TESO doesn't have that issue)

[leeperry]

Most seem to use it at 20-30% which doesn't introduce many artifacts at all, but enhances "perceived depth" in the image. I have a JVC X90 [..] The Darbee really is a way for me to try to squeeze every last bit out of my setup, even if it means only using it at a very low level.
NNEDI+Lanczos is pretty darn sharp, turn on Gamma Light dithering on top of it and that'll look plenty "deep" IMHO.

Overcoming the LCD panels poor convergence in your JVC projector by oversharpening the picture might not be your best option IMHO...maybe there are setttings to finetune their alignment, possibly in a service menu.

I've read that the latest 2160p projectors from SONY are still misconverged OOTB and you have to do your homework and go through the tedious task of aligning them by hand duh...and if the misalignment is not linear good luck with that.

mVR should provide R/G/B aligments soon or later, OTOH if they look fine and you're actually looking for slight sharpening then I kinda rest my case that mVR can already look very sharp on its own and PS scripts would nail it down.

[Cinemancave]

NNEDI+Lanczos is pretty darn sharp, turn on Gamma Light dithering on top of it and that'll look plenty "deep" IMHO.

Overcoming the LCD panels poor convergence in your JVC projector by oversharpening the picture might not be your best option IMHO...maybe there are setttings to finetune their alignment, possibly in a service menu.

I've read that the latest 2160p projectors from SONY are still misconverged OOTB and you have to do your homework and go through the tedious task of aligning them by hand duh...and if the misalignment is not linear good luck with that.

mVR should provide R/G/B aligments soon or later, OTOH if they look fine and you're actually looking for slight sharpening then I kinda rest my case that mVR can already look very sharp on its own and PS scripts would nail it down.

I'll give the Gamma light dithering a try, thanks.

Actually, the convergence on my unit is pretty darn good, especially after some alignment adjustments. But when MadVR lets me do it in software I'll be the first to try it and put the JVC at its default setting.

I am actually perfectly happy with the sharpness right now, I am more interested in the Darbee's ability to slightly adjust luminance levels to increase perceived contrast in the image. Is there any filter/script that I can use with MadVR to emulate that effect? (This goes against being true to the source material but I am fine with that, however this does vary from movie to movie though)

[Niyawa]

Unfortunately in the meanwhile user reports show the OpenCL <-> D3D9 interop cost seems to vary greatly from one PC to the next, even with the same GPU. Which means that I fear that your tests might only be true for your PC. It's possible that NNEDI3 will run faster or slower on somebody else's PC... :(
Interesting. I guess no one here has any sort of idea where we can set a deviation to that change...

I think such "levels" are not such a great idea when it comes to NNEDI, there is just so many different options that spreading such pre-defined levels can cause more confusion then it helps to solve.
Before, defining the performance levels of the scalers alone was OK, as most people agreed on the order of quality/performance, and if they didn't they usually knew why, but now..

There is so many cases where you might be able to use NNEDI on SD with a "worse" image scaler, but not on 720p, where you want a better image scaler. "Levels" are not going to cut it.
Not to mention the huge differences between NVIDIA and AMD at this time (in performance, and with NVIDIA even only working with certain drivers)
At the very least it's good to see people are parsing data. Even if at the time it's useless to label them, It's important to keep testing and looking for more details of how to handle NNEDI3 in majority of situations. Just having an idea that a R9 270x has the potential to be enough for lowest doubling (16 neurons) will let us get closer to a standard.

And yet I somehow manage to see it every time there is a frame drop, and I find even a single dropped frame very noticeable.
This should vary from people to people. My own experience tells me that without some sort of eye training most of us don't realize such things.

New madVR levels :D

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/400_100#post_24521720

http://cdn.avsforum.com/5/55/553cafaf_madVRlevels.PNG
Wow, thanks for sharing. I'm not sure about his choice of bilinear though. I noticed he's using DXVA as well, wonder if the results would differ in any worthwhile difference with software decoding instead.

Also, question for madshi (or anyone else that's able to answer this). Have you ever thought of implementing 'on-demand rendering algorithms' option for chroma upscaling in madVR? To be clear, they would be some sort of automatic-scaling filter that would activate/deactivate depending of the rendering situation madVR is in. Say we want to use Jinc as chroma upscaling, but it doesn't make much sense to have that enabled when there's no upscaling being done, so it would only be used when madVR actually upscales anything. I'm not a developer so I have no idea about the flexibility of this option, but it could help within a margin of reason.

One can say that the shortcut keys already do this, but I'd like an answer to make sure if the idea is a realistic approach or not.

[Guest]

My own experience tells me that without some sort of eye training most of us don't realize such things. I respectfully disagree. Your own experience tells you about only...your own experience. And what is "eye training" anyway? 6233638 says he can easily see single frame drops, and I can too. I think it would be the exceptional case to find someone that cannot. Why do you think people complain about telecine judder? It's only a single field jerk compared to a full frame drop, and yet it is easily detected. Human vision is finely tuned for tracking motion; it's not surprising that we can detect gross discontinuities like frame drops.

[leeperry]

I am more interested in the Darbee's ability to slightly adjust luminance levels to increase perceived contrast in the image. Is there any filter/script that I can use with MadVR to emulate that effect?
Maybe Jan would have something for you but dynamic contrast usually goes along with "pumping" effects during high contrast scenes transitions.

Darbee is prolly a nice toy when you run a standalone BD player but mVR's far more versatile than anything it could ever do IMHO, NNEDI and error diffusion dithering being just two things that will leave Darbee in the dust.

[Boltron]

I for one can't see half the detail many here argue about for days but I definitely do see judder and frame drops even just one or two.

[e-t172]

One thing to note is that to be able to see one lone frame drop (my current issue) you need to have reasonably accurate video playback in the first place (i.e. using smooth motion or matching refresh rate). If your video playback is not smooth to begin with (e.g. "raw" 24p@60Hz) then it is unlikely you will notice a frame drop, because it will be buried into the overall judder.

[seiyafan]

Besides nearest neighbor which is the lowest quality setting, all the other settings from bilinear to NNEDI3 I cannot tell any difference between them. To make things worse I can't tell nearest neighbor from NNEDI3 in chroma upscaling, lol! How....? Do you really need a projector to see the difference?

[QBhd]

One thing I have noticed about trying to tell the differences between algorithms... watching stills/frozen frames/same scene over and over doesn't seem to help. But I DO notice the differences if I watch for extended periods of time. It's more about NOT noticing things.

QB

[GREG1292]

Had a darbee hooked up for a short time before needi32 and
was ok with a panny 7000. For me it is not needed with my current
Dlp projector and latest Madvr.
Tried to do a smaller image not red on the projector just trying to post
an image on sharpness with Madvr removed for not posting correctly:o

[QBhd]

Image far far too large

And why is Gandalf's skin so pink!

QB

[renethx]

Interesting. I guess no one here has any sort of idea where we can set a deviation to that change...

In case you haven't looked at

http://www.avsforum.com/t/1477339/450#post_24514253

Here is a summary:

- PCI Express version (1.1 vs 2.0 vs 3.0) is the greatest factor that affects OpenCL <-> D3D9 interoperation in AMD GPU. madshi speculates that AMD internally does the interop by using some sort of copyback between CPU RAM and GPU RAM.
- Haswell PCIe 3.0 is the fastest, then IVB PCIe 3.0 (3.0 is supported by only Core i5 and higher), then AMD (Kaveri) in the benchmark. However, the difference between them is smaller in the real-world video playback (measured by rendering time).
- The lower the PCIe link speed is, the higher the GPU utilization (meausured with HWiNFO64) is. I am not sure how to interpret it.
- If you insert a device in the second or third PCIe 3.0 x16 slot of a Z87 chipset motherboard, the graphics card works only in PCIe 3.0 x8 or even x4 link, that is of the same bandwidth as PCIe 2.0 x16 or PCIe 1.1 x16. So be careful. In H87 / B85 chipset motherboards, there is no such problem, because the second PCIe x16 slot (if exists) is connected to the chipset (and works at 2.0 x4).

http://cdn.avsforum.com/7/78/7866d767_OpenCLSpeedTest.PNG

BTW using Bilinear with NNEDI3 is to fill the gap between Jinc3 and N16 x BC75AR, and the gap between N16 x BC75AR and N32 x BC75AR. Sometimes GPU may not be powerful enough for N16 x BC75AR, but enough for N16 x Bilinear. PQ is still better than Jinc3 in most cases.

Lanczos is very close to Bicubic in quality and I don't think it's worth its own level. Maybe level 2a = Bicubic, level 2b = Lanczos. There is a huge gap in both quality and speed between level 1, 2, 3, 4 and 5, while the difference in quality between a and b is very small (that's why I chose "a", "b"; "c' could be Jinc, but the difference between b and c is very small and NNEDI3 x c is too slow for most cases).

[aufkrawall]

Now I got a GTX 780 Ti@1.25Ghz and still I can't make use of Error Diffusion, with 4k GPU load at 100% and frame drops...

[Kalanoch]

I have nnedi3 working on GTX660/GK106 and Nvidia 335.23 drivers (and, it's 20% faster). I blocked deleting of OpenCL/D3D interop images between frames (madvr gets existing instances mapped to IDirect3DTexture9*) and it seems to fix the problem entirely.

This is huge, if this is legit and doesn't negatively affect any operations/image quality. (Hopefully it is because I'd like to use the latest nvidia drivers!)

[Ver Greeneyes]

Edit: Hah! It seems Kalanoch had the same idea - I guess I should have refreshed.

I have nnedi3 working on GTX660/GK106 and Nvidia 335.23 drivers (and, it's 20% faster). I blocked deleting of OpenCL/D3D interop images between frames (madvr gets existing instances mapped to IDirect3DTexture9*) and it seems to fix the problem entirely.I'm curious what madshi will think of this. He's usually pretty good at responding to everything but just so it doesn't get lost in the noise.. ;)

[nand chan]

How does madVR's support for BT.2020 work? madVR does not support constant-luminance encoding at all currently? How/when will it do so in the future?

How does color management work in the presence of BT.2020? How is BT.2020 content displayed on BT.709 monitors and vice versa? Do you intend to rely on the .3dlut being created against BT.2020? How will you transform into this space? Does madVR detect and respect the --primaries tag? What does it do in the absence of a .3dlut?

Do you do so with constant luminance? Which gamma transfer function do you use to de/encode BT.2020 and BT.709? How will it work moving forwards to a constant-luminance environment? How do plan on dealing with rounding/clipping artifacts on BT.709 and BT.2020 source 3dluts, especially for wide gamut profiles?

When will madVR support .ICC profiles, when will madVR support changing .ICC profiles at runtime (eg. if you move between monitors)?