madVR - high quality video renderer (GPU assisted)

[Warner306]

How many NGU users are using soften edges or add grain?

[seiyafan]

Interesting, I remember the days when Catmull-Rom LL was the most popular downscaler, so now it's bicubic 150 (do you also turn the LL on)?

[Hyllian]

Congrats, madshi! By looking at the comparison pictures, the envelope is pushed a little further. The results are amazing, mainly with good sources. I wonder how it would perform with retro games!

I suppose it'll stay closed source undefinitly, right? (I hope someday it'll be released.)

Is it totally based on shaders? Multipasses? Multialgorithms or is it just one algorithm?

[Sp00kyFox]

anyone else having issues with NNEDI3 lately? it gives me extreme graphical artifacts, doesn't matter which version I use. it's definitely not a heat or performance issue.

http://i.imgur.com/T323PcV.jpg

I'm using a GTX 760 (2GB VRAM) with driver version 375.95. don't know when it was the last time I tried this filter, so maybe it's a driver issue or a problem with this specific GPU?

[ryrynz]

anyone else having issues with NNEDI3 lately?

Not sure why you posted a screenshot with such extreme settings (500ms render time.. WTF) and then say it's not performance related..
Why not actually set some usable settings and see what happens, perhaps this is related in some way to using particular scalers together. So just try NNEDI3 by itself without killing your card slowly..
First step, disable NGU on chroma..

[thighhighs]

anyone else having issues with NNEDI3 lately? it gives me extreme graphical artifacts, doesn't matter which version I use. it's definitely not a heat or performance issue.

http://i.imgur.com/T323PcV.jpg

I'm using a GTX 760 (2GB VRAM) with driver version 375.95. don't know when it was the last time I tried this filter, so maybe it's a driver issue or a problem with this specific GPU?You use madvr\player x32 on windows x64? I have same artifacts when i use madvr x32 + nnedi image doubling on win10 x64. madvr x64 work fine. gtx 690, gtx660ti

[fedpul]

anyone else having issues with NNEDI3 lately? it gives me extreme graphical artifacts, doesn't matter which version I use. it's definitely not a heat or performance issue.

http://i.imgur.com/T323PcV.jpg

I'm using a GTX 760 (2GB VRAM) with driver version 375.95. don't know when it was the last time I tried this filter, so maybe it's a driver issue or a problem with this specific GPU?
Probably you are exceeding the VRAM limits with that settings. Try lower settings, those are extreme as already mentioned before.

Enviado desde mi MI 4LTE mediante Tapatalk

[Mikey2]

Even though my computer can process most of the awesome super advanced scaling algorithms, am I alone in still liking "SoftCubic" the most for Chroma Upscaling?

Image(Luma) is a different story; I like a super sharp, yet non-aliased luma algorithm. But that only makes the juxtaposition of the ultra-smooth colors all the more pleasing!

The same goes for 1080P source material, where only Chroma Upscaling is used!

[Sp00kyFox]

I wonder how it would perform with retro games!
with real pixelart, not good. with downscaled drawings or renderings it works great though as expected. here are some examples, the last one works best with this filter:
https://imgur.com/a/TnTFW

Not sure why you posted a screenshot with such extreme settings (500ms render time.. WTF) and then say it's not performance related..
I already tried minimal settings with NNEDI3 with the same result. my target was not to run the filter in realtime but to compare it with others. but if you insist:
http://i.imgur.com/uiFEUqm.jpg

to repeat myself, the problem is not performance related.

You use madvr\player x32 on windows x64? I have same artifacts when i use madvr x32 + nnedi image doubling on win10 x64. madvr x64 work fine. gtx 690, gtx660ti
yeah, maybe. I'm using ZoomPlayer which is only available as a 32bit application. probably time to use another software.

[ryrynz]

to repeat myself, the problem is not performance related.


Nobody else has this problem, it's related to your setup.
Try a proper removal and clean up of your drivers and reinstall, also disable any and every other application, ideally do a diagnostic startup through msconfig and try again.

[Sp00kyFox]

Nobody else has this problem, it's related to your setup.
I didn't state that this were the case. I actually asked if anyone else experienced this issue and as you can see someone did. so your claim that no one else has this problem is already false.

Try a proper removal and clean up of your drivers and reinstall, also disable any and every other application, ideally do a diagnostic startup through msconfig and try again.
no other applications were running at the time of capturing and if you've read my post you should know that I just installed a new GPU driver.
inb4: format c and reinstall windows :sly:

thighhighs got it right. I just tried it with MPC-HC and the problem is still there with the x32 version but gone with the x64 version.

[austonrush]

How many NGU users are using soften edges or add grain?

I'm using both :). At a level of one each. Since Add grain was fixed it has grown on me for real life video (I don't use it for animation). Soften edges is a must with NGU.

[austonrush]

I'm not liking any version of NGU for lower quality SD content. It significantly enhances artifacts and blocking. NNEDI3 64 with SR2 or SR3 performs much better in this space. Anyone else tested this. I don't have images to post right now but, it was not hard to see.

My current profiles are setup to use NGU image doubling for 720 and 1080 content. I'm sticking with NNEDI3 SR2 for low def content.

[Warner306]

I'm using both :). At a level of one each. Since Add grain was fixed it has grown on me for real life video (I don't use it for animation). Soften edges is a must with NGU.

So far, I agree soften edges is a must have. Haven't played with add grain. Neither option appears to hamper performance much.

[Scyna]

thighhighs got it right. I just tried it with MPC-HC and the problem is still there with the x32 version but gone with the x64 version.

I have the exact same problem you have except I get no video just audio. The queues also stay at 1/1. I'm on win 10 64x MSI Nvidia 1070 laptop 375.95 drivers.

[ryrynz]

I didn't stated that this were the case. I actually asked if anyone else experienced this issue and as you can see someone did. so your claim that no one else has this problem is already false.


I've not seen this listed as an issue. You found the one guy that experienced it.. Congrats, now lose the attitude.


no other applications were running at the time of capturing and if you've read my post you should know that I just installed a new GPU driver.
inb4: format c and reinstall windows :sly:


So? Installing a new driver doesn't always fix issues. What you're doing is just bypassing the issue with the driver working in 32 bit, so reverting or doing what I said may just fix it. inb4: I actually know what I'm doing but yet here I am using a 32 bit player :P

I have the exact same problem you have except I get no video just audio. The queues also stay at 1/1. I'm on win 10 64x MSI Nvidia 1070 laptop 375.95 drivers.

Try earlier drivers or completely remove the old set and reinstall. This probably isn't an issue across the entire set released by Nvidia and doesn't appear to be related to a specific family of cards.
I have no issues using the latest drivers on my 960 with a 32 bit player, so you shouldn't either..

[Sp00kyFox]

I've not seen this listed as an issue. You found the one guy that experienced it.. Congrats, now lose the attitude.
two guys, see Scynas post. and what attitude? I wasn't trying to anger or lecture you. maybe don't assume people are stupid before acting all so snippy and we can all help each other out in a friendly tone.

So? Installing a new driver doesn't always fix issues. What you're doing is just bypassing the issue with the driver working in 32 bit, so reverting or doing what I said may just fix it.
I didn't claim that installing a new driver would fix the issue. I just installed a new one recently but the problem was already there with the previous one. and did you read my posts? it does NOT work in 32bit but in 64bit.

inb4: I actually know what I'm doing but yet here I am using a 32 bit player :P
you're not the only one who knows how to properly handle a computer. yet here I am, a computer scientist who knows what he's doing and it's still not working in 32bit players. that's why I asked for help in the first place not because I'm too lazy to look for a solution.

I have no issues using the latest drivers on my 960 with a 32 bit player, so you shouldn't either..
there are more components to a system then the GPU and its driver. reminds me of gaming forums where "but it works fine here so you're the problem" is a common answer.
now, to cite you, lose the attitude. you could start by telling us what OS, what player and what render path in madvr you use and maybe we can figure out together what the issue is.

[huhn]

what decoder is used?

[Hyllian]

with real pixelart, not good. with downscaled drawings or renderings it works great though as expected. here are some examples, the last one works best with this filter:
https://imgur.com/a/TnTFW

Thanks for the images. Waifu2x presents the same behavior too. They have very similar outputs. It should be great with pre-rendered games.

[Sp00kyFox]

what decoder is used?
on my side it's LAVFilters v0.68.1-35. I usually use it with DXVA2 (copy-back) but I just tried all the other options with no success.

Thanks for the images. Waifu2x presents the same behavior too. They have very similar outputs. It should be great with pre-rendered games.
no problem. the challenge would be to seperate the image between pre-rendered elements and pixelart or 3D objects. maybe a similar approach like you once did with xBR does also work with NGU.

[Backflash]

to repeat myself, the problem is not performance related.


Do videogames run fine? If you didn't try probably VRAM is fried or slowly dying. Chip itself might be giving out such artifacts as well.
Does it crash your system? It soon will. I fried a few cards this way.
Which brand?

I think people need a disclaimer to stop running scalers that spin GPUs to max temperatures on lower end hardware or weak brands, it's really easy to ruin videocard with NNDI3 specially if you like to watch some series whole weekend non stop like me, games are much more tame temperature wise.

Also this is why I switched to NGU, it doesn't 100% my GPU(r9 390) all the time like NNDI3, I like when it's 55C' on a moderate load.

[burfadel]

I agrree Backflash, that looks like video memory corruption. It's one of the side affects you can get from overclocking too high, it might not necessarily crash. I'm not suggesting he was overclocking, just using that as an example. It could be something else as well of course, it's just one possibility that seems most likely at this stage. The settings used seems very overkill though.

[ryrynz]

Do videogames run fine? If you didn't try probably VRAM is fried or slowly dying. Chip itself might be giving out such artifacts as well.


Switching to a 64 bit media player fixes it so it's obviously driver related.


I agrree Backflash, that looks like video memory corruption.


Definitely corruption but I'm guessing he wanted to know the cause.


The settings used seems very overkill though.

It was, pretty silly to say it's not related to performance then post a screenshot with the most extreme settings.. but whatever.


two guys, see Scynas post. and what attitude? I wasn't trying to anger or lecture you. maybe don't assume people are stupid before acting all so snippy and we can all help each other out in a friendly tone.


Not posting like your stupid is a good start.


I didn't claim that installing a new driver would fix the issue. I just installed a new one recently but the problem was already there with the previous one. and did you read my posts? it does NOT work in 32bit but in 64bit.


Cool story bro. Tell me about the driver that came before that and the one before that.. same issue? Tell me about all the ones you tried?
Oh wait, that would've been the smart thing to do when first posting. Because troubleshooting is the obvious smart thing to do before doing so, and you did none so...


you're not the only one who knows how to properly handle a computer. yet here I am, a computer scientist who knows what he's doing and it's still not working in 32bit players. that's why I asked for help in the first place not because I'm too lazy to look for a solution.


Obviously.. you just needed to take a few minutes to troubleshoot, which you didn't. If that's not lazy...


there are more components to a system then the GPU and its driver. reminds me of gaming forums where "but it works fine here so you're the problem" is a common answer.


When it's only related to NNEDI3? When it's only happening on 32 bit players? When hardly anyone else is having the same issue?
C'mon now.. where are all the arrows pointing?


now, to cite you, lose the attitude. you could start by telling us what OS, what player and what render path in madvr you use and maybe we can figure out together what the issue is.

That's rich considering how you chose to repond first, but whatever.
375.95, Windows 10 14393.447, GTX 960 madVR 0.91.1

Now just do what I said and uninstall your drivers properly and try at least a few different previous versions if not more if you care.
Thing is you shouldn't really be using a 32 bit player without good reason anyway which I'll assume you're not now and hence no more issue.. but if you were interested in resolving it, that's where I'd start.
I'd probably reset madVR to defaults too if previous drivers versions changed nothing.

[MariaX9]

How many NGU users are using soften edges or add grain?

I use soften edges 2 atm.

[Smithy]

DXVA2 Copy Back, ED2 + colored + change, Fullscreen Window Mode (8 backbuffer) D3D9 old path, CPU64/GPU24
no General Settings Checked, no Image Enhancements Checked and no Trade Quality for Performence Checked
reduce Banding High/High, reduce Ringing only
Chroma ReconSoft SR2, Image Upscaling Jinc AR SL, Image Downscaling Bicubic 150 AR LL

SD 720x404 25p -> 1080p60 Smoothmotion

Jinc AR SL http://pix.defcon5.biz/images/79967370398173641729.png -> 9ms
Jinc AR SL - SR2 AR LL http://pix.defcon5.biz/images/19049767906798837580.png -> 14,5ms
SXBR AB25 -> SR2 AR LL http://pix.defcon5.biz/images/37724967688991784662.png -> 16ms
SXBR AB25 Quadruple -> Down http://pix.defcon5.biz/images/22247513489246501043.png -> 18,5ms
NGU Low/Low se1 -> http://pix.defcon5.biz/images/02571542679210162091.png -> 13,5ms
NGU Med/Med se1 -> http://pix.defcon5.biz/images/49239113436890867062.png -> 16ms
NGU High se1 / Jinc AR SL -> ~20,5ms
NGU High/Low se1 -> ~20,5ms
NGU High/Med se1 -> ~23ms
NGU High/High se1 -> http://pix.defcon5.biz/images/73322545513825247537.png -> ~37-62ms
NGU Quadruple High/High se1 -> Down http://pix.defcon5.biz/images/20754926886660168998.png -> 160ms

[ts1]

Congrats, madshi! By looking at the comparison pictures, the envelope is pushed a little further. The results are amazing, mainly with good sources. I wonder how it would perform with retro games!

I suppose it'll stay closed source undefinitly, right? (I hope someday it'll be released.)

Is it totally based on shaders? Multipasses? Multialgorithms or is it just one algorithm?

I'm pretty sure it is based on this (https://arxiv.org/abs/1609.05158) paper. It is implemented in python here (https://github.com/Tetrachrome/subpixel) and partly here (https://github.com/alexjc/neural-enhance).

[leeperry]

If I set the display to 23 or 24 fps for a 24 fps movie then I still get judder. Same problem with 59 and 60 fps. Only bluesky fps solves this. Any idea what the problem can be?

It takes at least 72 actual fps to fool the visual cortex FWIR and fighter pilots would be able to see up to 100fps so don't expect miracles but IME judder becomes a lot more bearable if it comes with perfect timing using Reclock + mVR's 60Hz FRC + both chroma and luma SR. Both kinds of SR together sharpen up motion blur and much improve motion smoothness to my eyes, like the LSF Avisynth plugin did IME ages ago. I can't always afford luma SR on 720p but chroma SR alone still helps.

The higher the refresh rate the better eventually, I really liked 100Hz FRC on CRT's and can hardly wait for 144Hz FRC on the BenQ EX3200R http://forum-images.hardware.fr/images/perso/drasche.gif

[sauma144]

@madshi
NGU luma doubling is just amazing!
I think add grain and soften edges should be always on.
However for 30hz/60hz (mostly from youtube) video files, my laptop can't handle NGU low luma doubling.

Are you working on a NGU luma upscaler (not doubler)?
Is there any chance to get a NGU as fast as Jinc? Spline3?

[madshi]

Either I am doing something totally wrong or things are not supposed to work the way I expect them to.

I ripped countless of times my Avatar 3D BluRay disc with latest MakeMKV v1.10.2 and yet the 3D subtitles are not working properly, for that matter they are not working properly even when I directly play the movie from my disc.

The 3D effects of the movie most of the time cut through the subtitles, and it doesn't matter which subtitle language I am using.
Maybe you can create a small sample which demonstrates the problem, using the latest MakeMKV version?

Congrats, madshi! By looking at the comparison pictures, the envelope is pushed a little further. The results are amazing, mainly with good sources. I wonder how it would perform with retro games!

I suppose it'll stay closed source undefinitly, right? (I hope someday it'll be released.)
Thanks! :)

I wanted an algo which reconstructs lines correctedly without any bloating and there simply wasn't anything available (except very slow neural networks like waifu2x), so I finally created my own.

Although I wanted to have a rudimentary support for "decimated" sources (pixel art, or images downscaled with nearest neighbor sampling), my main target was to get the best out of high-quality sources. So I suppose pixel art / games probably isn't where NGU shines.

Is it totally based on shaders? Multipasses? Multialgorithms or is it just one algorithm?
It's just one PS3.0 shader algorithm with multiple passes (number of passes depends on quality level).

No plans to open source it any time soon, unfortunately. I'm sorry about that. If I could just share it with you, for game playback, and could be 100% sure nobody else (except me) would ever use it for any other purpose, I would share it in a heartbeat. But I don't think sharing is possible without my "competition" getting access to it, too.

Waifu2x presents the same behavior too. They have very similar outputs.
Really? I thought waifu2x would do much better with pixel art! :eek:

I'm not liking any version of NGU for lower quality SD content. It significantly enhances artifacts and blocking. NNEDI3 64 with SR2 or SR3 performs much better in this space. Anyone else tested this. I don't have images to post right now but, it was not hard to see.
Some screenshots would be nice. With SR4, if you don't mind, so I can see if SR really handles this better than NGU.

But I do agree, if you have sources with high compression artifacts, NGU will try to interpret the artifacts as important image detail. The current NGU version really needs high quality sources to shine.

I'm pretty sure it is based on this (https://arxiv.org/abs/1609.05158) paper.
No, not at all. If my algo were based on that paper, it would be named after it instead of "NGU". I wouldn't touch that paper with a ten-foot pole, considering that it was written by MagicPony and then bought by Twitter and could potentially be burdened with patents. I specifically excluded that paper from my research, just to be safe.

My algo is not based on any one specific paper. As I said before, I combined ideas from several papers, and added a couple ideas of my own.

[cork_OS]

Since NGU doesn't benefit from SuperRes, is possible/useful to apply SuperRes only to the last upscale step (Jinc/C-R etc.), after NGU?
Maybe similar <to waifu2x> results/quality could be possible with NGU, too, when doing proper noise/artifact reduction before upscaling. I don't know, but something like this not likely to come to madVR soon.
Unfortunately, I have a lot of noisy/blocky SD stuff. Which AviSynth deblocker/denoiser could be used for real-time NGU preprocessing?

[ryrynz]

The current NGU version really needs high quality sources to shine.


I know NNEDI3 might be hard to beat in the lower source quality arena but still.. I await what may be cooked up from NGU to possibly replace it in future.


Unfortunately, I have a lot of noisy/blocky SD stuff. Which AviSynth deblocker/denoiser could be used for real-time NGU preprocessing?

I wouldn't be surprised if this was perhaps a little higher on madshi's todo list. I have no doubt we'll see something in future.
I've always been quite partial to Didee's double Fluxsmooth on noisier sources.

FluxSmoothST(12,4).FluxSmoothT(4).Merge(last,0.49)

The thing is though with how NGU picks things up it'd be hard to avoid the issues it creates with low quality sources without killing what little there is of the detail in the picture.
Might need a combination of filters to combat this. Certainly the best thing you can do right now is soften edges and use a soft scaler and avoid extra sharpening.

[Matching_Mole]

Unfortunately, I have a lot of noisy/blocky SD stuff. Which AviSynth deblocker/denoiser could be used for real-time NGU preprocessing?

I agree with you, a good deblocker (either from avisynth or integrated in Madvr), is now the last missing part to achieve a perfect playback chain.

[Kotik]

Maybe you can create a small sample which demonstrates the problem, using the latest MakeMKV version?


Here is a small sample (http://www.filedropper.com/avatar), notice how the 3D effects of the movie cut through the 3D subtitles, this is happening during the whole movie and it doesn't matter whether I use xySubFilter or the MPC-HC internal sub engine.

Sample was done using latest MakeMKV and then splitting it with latest MKVToolNIX.

[Sp00kyFox]

Do videogames run fine?
yes. I already excluded the possibility of a performance or heat problem before I even posted here. see my inital response to ryrynz answer.

Not posting like your stupid is a good start.
not assuming that people don't know what they talking about would be. you're making one wrong assumption after the other instead of conceding that your discussion partner may be as competent as you are and taking him by the word.

Tell me about the driver that came before that and the one before that.. same issue? Tell me about all the ones you tried?
the previous version I had was v375.63. before I posted here I did try all the driver versions from the nvidia archive beginning with v359.06 (1.12.2015) with no avail.

Oh wait, that would've been the smart thing to do when first posting. Because troubleshooting is the obvious smart thing to do before doing so, and you did none so...

Obviously.. you just needed to take a few minutes to troubleshoot, which you didn't. If that's not lazy...
your assumption is wrong again. I did put several hours into it before I decided to ask if someone else knew this problem. which did help. thanks to thighhighs now we know how to avoid the problem but we still haven't figured out the cause for it in the first place.

When it's only related to NNEDI3? When it's only happening on 32 bit players? When hardly anyone else is having the same issue? C'mon now.. where are all the arrows pointing?
if you have figured it out, tell me because I didn't. otherwise I wouldn't have asked here and just solved the problem. thanks to Scyna we can probably rule out a GPU architecture related problem (it appears both on Keplar and Pascal). and with your information it doesn't seem driver but player related. since at least one user (you) doesn't have this problem with the same driver version despite using a 32bit player.

That's rich considering how you chose to repond first, but whatever.
I don't see what's wrong with my initial answer. but english is not my native language so if I happen to came across the wrong way I apologize.

375.95, Windows 10 14393.447, GTX 960 madVR 0.91.1
so far the only difference between us is the GPU. you said it works for you with a 32bit player. can you tell us which one? I would give it a try. and also which render patch you're using, DX9 or DX11? the problem occurs with both here.

Now just do what I said and uninstall your drivers properly and try at least a few different previous versions if not more if you care.
already did that and with your information we know that it should work correctly with the recent version.

Thing is you shouldn't really be using a 32 bit player without good reason anyway which I'll assume you're not now and hence no more issue.. but if you were interested in resolving it, that's where I'd start.
I just happen to like ZoomPlayer which is only available as a 32bit player. so far this is the only issue I ran into. we already know that it works fine with 64bit players. question is why doesn't it work in 32bit.

I'd probably reset madVR to defaults too if previous drivers versions changed nothing.
already tried that with no success.

[huhn]

it is not cuvid so it should be this problem:
http://bugs.madshi.net/view.php?id=396

[Sp00kyFox]

thanks for the info, so this is indeed a known issue. still strange why it works for ryrynz though.

[huhn]

well a 960 is maxwell 2.0 and a 760 is kepler. nvidia made huge improvements with maxwell in term of openCL performence. so they are quite different.

[Sp00kyFox]

if this is the cause then the issue must have been reintroduced with the Pascal architecture. because Scyna is using a GTX 1070 and has confirmed this problem as well.

[huhn]

try the workaround posted by madshi in the bug tracker or
use 64 bit
or/AND report the issue to nvidia.

there is nothing else i can say about this.

[Sp00kyFox]

like I said, it works fine with a 64bit player. just curious why it doesn't with 32bit when according to ryrynz it should.
well maybe in the end it is a NVIDIA driver problem which just happen to be non-existent on Maxwell GPUs only.

[huhn]

you can use 64 bit openCL in 32 bit mpc-hc try the workaround.

delete the reg key start a 64 bit player and use nnedi3 and than try the 32 bit player again.

madVR is only creating one openCL kernel if the first openCL kernal is 64 bit this may work.

i can't remember a change to this behavior so it should work. give it a try...

[Sp00kyFox]

nice idea but that didn't help either. I just think that it doesn't work with nvidia in 32bit applications. I'm gonna ask a friend to try out, I know he has a GTX 980 TI (Maxwell). and I'm eager to hear what 32bit player ryrynz was using.

[StinDaWg]

Not posting like your stupid is a good start.
You're :stupid:

Lose the attitude.

[Pat357]

just tried the LOW setting of NGU, but still no dice. my hardware is to weak i guess..(gtx670)
@ madshi

greetz

You should be able to use NGU, because my my GTX-680 does even medium NGU !

[VHT]

Hello,

Either I am doing something totally wrong or things are not supposed to work the way I expect them to.

I ripped countless of times my Avatar 3D BluRay disc with latest MakeMKV v1.10.2 and yet the 3D subtitles are not working properly, for that matter they are not working properly even when I directly play the movie from my disc.

The 3D effects of the movie most of the time cut through the subtitles, and it doesn't matter which subtitle language I am using.

I am using latest madVR build + latest LAV, I tried with both xySubFilter and with MPC-HC internal subtitle engine.

I remember a couple of months ago madshi mentioned that 3D subtitles are supposed to work with xySubFilter.

FYI I got a passive 3D TV and madVR 3D setting set to auto, the movies themselves do work fine it is just the subtitles in both the bluray disc and the .mkv rip are not really working properly.

So am I the only one with this issue? am I doing something wrong? or 3D subtitles were never meant to work with madVR + LAV?

Thanx in advance.


You are not alone. I've had the same problem since the day one when I started usind madVR+LAV with 3d movies. Movie is fine but subtitles are in wrong depth. I've also tried different subtitle engines, original BluRay discs etc. but with no luck. :(

[pose]

RX 470 user here. Win7 64bit, 16.11.4 driver. Ive tested a few different scenarios and basically when using the new algo as image doubler for luma NGU-med = NNEDI3 64 and NGU-high = NNEDI3 128 performance wise for me. Changing upscaling from Jinc to CR gave me -1ms of rendering time. Anime with NGU looks really good to me. Thanks!

Just tried a movie as video source and it looks hecking "natural". Well done, madshi!

[pirlouy]

Not posting like your stupid is a good start.You're :stupid:

Lose the attitude.
Totally agree.
Sp00kyFox made a good description, is polite etc. If you don't want to help, just don't post, instead of being that aggressive.

[Sp00kyFox]

okay my friend with the GTX 980 (v375.95) just reported back and to my suprise the NNEDI3 problem doesn't occur on his system. he tried it with the 32bit version of MPC-HC. so it seems as if this bug indeed only happens with Kepler and Pascal GPUs but not on Maxwell ones. how weird.

[Asmodian]

okay my friend with the GTX 980 (v375.95) just reported back and to my suprise the NNEDI3 problem doesn't occur on his system. he tried it with the 32bit version of MPC-HC. so it seems as if this bug indeed only happens with Kepler and Pascal GPUs but not on Maxwell ones. how weird.

I have a Pascal GPU (Titan X Pascal) and I don't have any issues with NNEDI3 and 32-bit Zoom Player. :confused:

[Sp00kyFox]

then it must be another determining factor. well, maybe someone will figure it out. I'm out of ideas at the moment.