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

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

 

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

Reply
 
Thread Tools Search this Thread Display Modes
Old 23rd November 2016, 07:23   #40621  |  Link
Backflash
Registered User
 
Join Date: Jan 2016
Posts: 52
Quote:
Originally Posted by Sp00kyFox View Post

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.
Backflash is offline   Reply With Quote
Old 23rd November 2016, 07:29   #40622  |  Link
burfadel
Registered User
 
Join Date: Aug 2006
Posts: 2,229
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.
burfadel is offline   Reply With Quote
Old 23rd November 2016, 07:46   #40623  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Quote:
Originally Posted by Backflash View Post
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.


Quote:
Originally Posted by burfadel View Post
I agrree Backflash, that looks like video memory corruption.
Definitely corruption but I'm guessing he wanted to know the cause.

Quote:
Originally Posted by burfadel View Post
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.


Quote:
Originally Posted by Sp00kyFox View Post
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.

Quote:
Originally Posted by Sp00kyFox View Post
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...

Quote:
Originally Posted by Sp00kyFox View Post
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...

Quote:
Originally Posted by Sp00kyFox View Post
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?

Quote:
Originally Posted by Sp00kyFox View Post
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.

Last edited by ryrynz; 23rd November 2016 at 08:22.
ryrynz is offline   Reply With Quote
Old 23rd November 2016, 07:59   #40624  |  Link
MariaX9
Registered User
 
Join Date: May 2016
Posts: 27
Quote:
Originally Posted by Warner306 View Post
How many NGU users are using soften edges or add grain?
I use soften edges 2 atm.
MariaX9 is offline   Reply With Quote
Old 23rd November 2016, 08:09   #40625  |  Link
Smithy
Registered User
 
Join Date: Apr 2015
Location: Germany
Posts: 20
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
__________________
32bit Vista SP2 (DX11 Platform Update), Nvidia GTX960 4GB AMP (365.19), 1. 27' FHD Benq / 2. 65' UHD HDR Sony XE93, madVR 0.92.17, LAV 0.73.1, MPC-HC 1.8.3

Last edited by Smithy; 24th November 2016 at 06:20.
Smithy is offline   Reply With Quote
Old 23rd November 2016, 08:25   #40626  |  Link
ts1
Registered User
 
Join Date: Jan 2015
Posts: 71
Quote:
Originally Posted by Hyllian View Post
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 paper. It is implemented in python here and partly here.
ts1 is offline   Reply With Quote
Old 23rd November 2016, 09:07   #40627  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
Quote:
Originally Posted by HillieSan View Post
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

Last edited by leeperry; 23rd November 2016 at 09:11.
leeperry is offline   Reply With Quote
Old 23rd November 2016, 09:12   #40628  |  Link
sauma144
Registered User
 
Join Date: Sep 2016
Posts: 89
@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?
sauma144 is offline   Reply With Quote
Old 23rd November 2016, 10:02   #40629  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by Kotik View Post
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?

Quote:
Originally Posted by Hyllian View Post
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.

Quote:
Originally Posted by Hyllian View Post
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.

Quote:
Originally Posted by Hyllian View Post
Waifu2x presents the same behavior too. They have very similar outputs.
Really? I thought waifu2x would do much better with pixel art!

Quote:
Originally Posted by austonrush View Post
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.

Quote:
Originally Posted by ts1 View Post
I'm pretty sure it is based on this 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.
madshi is offline   Reply With Quote
Old 23rd November 2016, 10:46   #40630  |  Link
cork_OS
Registered User
 
cork_OS's Avatar
 
Join Date: Mar 2016
Posts: 160
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?
Quote:
Originally Posted by madshi View Post
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?
cork_OS is offline   Reply With Quote
Old 23rd November 2016, 10:58   #40631  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Quote:
Originally Posted by madshi View Post
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.

Quote:
Originally Posted by cork_OS View Post
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.

Last edited by ryrynz; 23rd November 2016 at 11:11.
ryrynz is offline   Reply With Quote
Old 23rd November 2016, 11:18   #40632  |  Link
Matching_Mole
Registered User
 
Join Date: Feb 2009
Posts: 99
Quote:
Originally Posted by cork_OS View Post
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.
Matching_Mole is offline   Reply With Quote
Old 23rd November 2016, 12:41   #40633  |  Link
Kotik
Registered User
 
Join Date: Jun 2009
Posts: 85
Quote:
Originally Posted by madshi View Post
Maybe you can create a small sample which demonstrates the problem, using the latest MakeMKV version?
Here is a small sample, 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.
Kotik is offline   Reply With Quote
Old 23rd November 2016, 14:13   #40634  |  Link
Sp00kyFox
Registered User
 
Sp00kyFox's Avatar
 
Join Date: Aug 2007
Posts: 79
Quote:
Originally Posted by Backflash View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
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.

Quote:
Originally Posted by ryrynz View Post
I'd probably reset madVR to defaults too if previous drivers versions changed nothing.
already tried that with no success.
Sp00kyFox is offline   Reply With Quote
Old 23rd November 2016, 14:22   #40635  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
it is not cuvid so it should be this problem:
http://bugs.madshi.net/view.php?id=396
huhn is offline   Reply With Quote
Old 23rd November 2016, 14:31   #40636  |  Link
Sp00kyFox
Registered User
 
Sp00kyFox's Avatar
 
Join Date: Aug 2007
Posts: 79
thanks for the info, so this is indeed a known issue. still strange why it works for ryrynz though.
Sp00kyFox is offline   Reply With Quote
Old 23rd November 2016, 14:34   #40637  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
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.
huhn is offline   Reply With Quote
Old 23rd November 2016, 14:38   #40638  |  Link
Sp00kyFox
Registered User
 
Sp00kyFox's Avatar
 
Join Date: Aug 2007
Posts: 79
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.
Sp00kyFox is offline   Reply With Quote
Old 23rd November 2016, 14:43   #40639  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
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.
huhn is offline   Reply With Quote
Old 23rd November 2016, 14:50   #40640  |  Link
Sp00kyFox
Registered User
 
Sp00kyFox's Avatar
 
Join Date: Aug 2007
Posts: 79
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.
Sp00kyFox is offline   Reply With Quote
Reply

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

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

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

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

Forum Jump


All times are GMT +1. The time now is 22:51.


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