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 27th July 2015, 16:12   #32161  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
Quote:
Originally Posted by XMonarchY View Post
Is the "Low Latency Mode" some new option or is it enabled by default?
Thats not an option users need to worry about, its only for players that want to show an interactive OSD with madVR.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders
nevcairiel is offline   Reply With Quote
Old 27th July 2015, 16:52   #32162  |  Link
clsid
*****
 
Join Date: Feb 2005
Posts: 5,643
@madshi
Symantec keeps detecting mvrsettings32.dll after each new release. You can submit a false positive report here (if you haven't done so already): https://submit.symantec.com/false_positive

I have already done it for the current version.
__________________
MPC-HC 2.1.7.2
clsid is offline   Reply With Quote
Old 27th July 2015, 16:56   #32163  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by XMonarchY View Post
With 88.21, I get presentation glitches unless I lower Chroma Upscaling SR to Passes = 2 & Strength = 1 and also lower Upscaling Refinement SR to Strength = 2, Radius = 0.66 .

Not sure why - maybe presentation glitches were not properly reported in 88.20?
SuperRes in Upscaling Refinement got more expensive in v0.88.21 compared to v0.88.20, because that was the only way to make higher radius settings work without artifacts. SuperRes is still a work-in-progress. Performance might change from one version to the next. When all is said and done, maybe we can squeeze some more performance out of SuperRes. Makes no sense to do time consuming optimizations right now, when things are not in their final state yet.
madshi is offline   Reply With Quote
Old 27th July 2015, 16:56   #32164  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by clsid View Post
@madshi
Symantec keeps detecting mvrsettings32.dll after each new release. You can submit a false positive report here (if you haven't done so already): https://submit.symantec.com/false_positive

I have already done it for the current version.
Argh, when are they finally going to learn?
madshi is offline   Reply With Quote
Old 27th July 2015, 17:06   #32165  |  Link
Thunderbolt8
Registered User
 
Join Date: Sep 2006
Posts: 2,197
Quote:
Originally Posted by madshi View Post
The exact source and target resolution doesn't matter too much to any of the scaling algorithms. The scaling factor decides everything. 1080p -> 4k is exactly 2.00x scaling factor. Which is the same as 400 -> 800 pixels.
well then at least use the scaling factors which refer to the scalings of the resolutions I suggested. also, using resolutions with which you are still able to recognize changes of course.
__________________
Laptop Lenovo Legion 5 17IMH05: i5-10300H, 16 GB Ram, NVIDIA GTX 1650 Ti (+ Intel UHD 630), Windows 10 x64, madVR (x64), MPC-HC (x64), LAV Filter (x64), XySubfilter (x64) (K-lite codec pack)
Thunderbolt8 is offline   Reply With Quote
Old 27th July 2015, 17:08   #32166  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by Thunderbolt8 View Post
well then at least use the scaling factors which refer to the scalings of the resolutions I suggested.
I did exactly that. Using 2.0x scaling, same as 1080p -> 4K.
madshi is offline   Reply With Quote
Old 27th July 2015, 17:10   #32167  |  Link
aufkrawall
Registered User
 
Join Date: Dec 2011
Posts: 1,812
Quote:
Originally Posted by madshi View Post
Argh, when are they finally going to learn?
Probably never, since most likely there's not much work done by humans when false positive threshold isn't exceeded.
aufkrawall is offline   Reply With Quote
Old 27th July 2015, 17:30   #32168  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
Quote:
Originally Posted by madshi View Post
Might have made sense to mention that in your original post... It's quite confusing if you use the same syntax to describe different settings, without actually explaining that anywhere.
Well, these are the only knobs in both builds. I don't see what's confusing, 0.66 radius would even appear to be the default value:





I'll give you that I didn't mention "softness" in .15 but then again it shoulda been called "bluriness", this knob is beyond useless IME.

Quote:
Originally Posted by madshi View Post
I'm not totally sure what your goal with those images were.
1) Refute the following claims you made a few days ago:
Quote:
Originally Posted by madshi View Post
The reason for that is that LQ mode sucks. Big time.
Quote:
Originally Posted by madshi View Post
the direction LQ is going is away from the ground truth, while HQ is going nearer to the ground truth
It definitely does in .20 because even a strength of 1 is way oversharp, I don't see how you could say that 3@0.41LQ in .15 sucks this bad compared to HQ and that HQ is so much closer to the ground truth.

It's not quite the end-user's fault if you've changed the SR logic in a way that makes LQ useless in .20.

Again, it's the middle of summer and many ppl are on vacation......still, not every mVR tester has gone AWOL and someone else would appear to see things the same way I do(God forbid ^^):
Quote:
Originally Posted by tFWo View Post
It's hard to compare because strength 1 for nohq is too strong. (..)
HQ still has that almost softer than original look with slight ringing around numbers. White is spilled into black area, black is spilled into white area.

NoHQ doesn't have that problem even on higher strengths. But as I said even strength 1 is too strong for my taste.
I believe it to be rather undeniable in my comparison that HQ is way softer and that reasonable LQ settings in .15 look "better", as in "closer to the ground truth".

2) You've decided to only provide on single strength knob in .20, being your own very personal homebrew experimental recipe of a mix between the number of the passes and strength(as known in .15). This cannot work IME and it definitely doesn't in .20 as even a strength of 1 is way too sharp to be of any use huh(and again, I'm not the only one thinking so).

On one hand you refuse to provide knobs in order to finetune sxbr and on the other LQ SR with a strength of 1 is über-sharp and HQ is blurry as hell, major show stoppers at work here.

3) Show that sxbr125 with 3@0.41LQ in .15 looks better than anything HQ could ever provide in .15 or .20.

4) Allow me to see that NEDI has been superseded by sxbr125, once the SR settings are finetuned in order to counterbalance its sheer sharpness.

Quote:
Originally Posted by madshi View Post
A goal oriented test would have been to use v0.88.15 for all images, and then to compare 3 passes with 0.65 strength to 2 passes with 1.00 strength, to check if there's a noticeably difference in image quality. But you didn't do that, so you didn't put any weight to your wish that you want to be able to adjust passes and strength separately.
Right, well 3@0.65 was nice with NEDI but I now prefer 3@0.41 with the sharper sxbr125. That's the only sharpness knob I have access to(the sharpness setting in my TV looks horrific) so I make do with what I got and once all properly set it truly does wonders

I did use both builds with the sole features they provide, stop hiding knobs from the end-user and you'll get your fair comparison.

As I said, in .15 I don't like 4@0.41 or 3@0.42, I'm not sure such tiny differences would be visible on screenshots at all ....the same way most of the chroma suboptions would be mostly invisible too, yet users are allowed to pick one over another.......if they all look "roughly the same", you might as well ditch most of them and only keep one or two for all I know. I really wasn't aware that mVR was going towards a "more or less the same ya know the deal" route, good to know.

Quote:
Originally Posted by madshi View Post
So far nobody has shown any proof that 3 passes with lower strength looks better at all than 2 passes with higher strength. All the evidence so far suggests that both produce nearly identical results. Yes, there are tiny differences, but it's hard to say which is actually better. If there were proof that 3 passes with lower strength would objectively be better, I would consider allowing that, even if it costs more GPU power. But there's no evidence that points in that direction. And I'm not going to waste GPU power on something where nobody is able to produce any evidence of improvement with any test images.
Right, so you make subjective choices for the end-user because you know it will look "nearly" identical even if it doesn't at all to the end-user.

Lemme know what screenshots might possibly change your mind then? Would a comparison of 2@1.0LQ against 3@0.41LQ with the former being sharper and less detailed make your day? I should fairly easily be able to make that happen.

BTW, I also really like 3@0.41(0.00 softness eventually) SR on chroma with sxbr125+AR...PQ stunningness never ends hah!

Quote:
Originally Posted by madshi View Post
I've not changed anything there. But if your media player still runs, the DLL will still be loaded. Make sure your media player is really closed. It might still be lingering somewhere in your task manager.
I did check eventually and it is not, I can even rename PotP's folder without a itch. W7 is prolly keeping your DLL locked for some reason, this very DLL would appear to be related to networking.

Last edited by leeperry; 28th July 2015 at 01:04.
leeperry is offline   Reply With Quote
Old 27th July 2015, 17:44   #32169  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by leeperry View Post
I'm not sure such tiny differences would be visible on screenshots at all ....the same way most of the chroma suboptions would be mostly invisible too
As I said, when using the "right" test images/videos, different chroma scalers can look vastly different from each other. So what you're saying here (and already said in your previous post) is flat out incorrect.

Quote:
Originally Posted by leeperry View Post
Lemme know what screenshots might possibly change your mind then? Would a comparison of 2@1.0LQ against 3@0.41LQ with the former being sharper and less detailed make your day? I should fairly easily be able to make that happen.
I've told you many times already. 3 passes with 0.66 strength should be roughly identical to 2 passes with 1.00 strength. So make comparison screenshots for those two situations with v0.88.15. You can use any other combination of passes and strengths, too. Just make sure that they sum up to roughly the same value. E.g. 3 * 0.66 = 1.98. 2 * 1.00 = 2.00. And 1.98 is roughly identical to 2.00. If you insist that you have to use a strength of 0.41, then compare 3 passes with 0.41 with 2 passes with 0.61. Because 3*0.41 is roughly identical to 2*0.61.

Quote:
Originally Posted by leeperry View Post
I did check eventually and it is not, I can even rename PotP's folder without a itch. W7 is prolly keeping your DLL locked for some reason, this very DLL would appear to be related to networking.
Well, some process must still have it loaded. Maybe madHcCtrl.exe is still running? Maybe you've changed the madVR tray icon setting to have it always visible?
madshi is offline   Reply With Quote
Old 27th July 2015, 18:01   #32170  |  Link
aufkrawall
Registered User
 
Join Date: Dec 2011
Posts: 1,812
Something's still wrong with SuperRes, causing a load of aliasing (tested with that lighttower picture leeperry posted some days ago).
Doesn't happen with this SuperRes version (as long as softness isn't used).
aufkrawall is offline   Reply With Quote
Old 27th July 2015, 18:29   #32171  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by aufkrawall View Post
Something's still wrong with SuperRes, causing a load of aliasing (tested with that lighttower picture leeperry posted some days ago).
Doesn't happen with this SuperRes version (as long as softness isn't used).
I'm going to reenable linear light downscaling for SuperRes in the next build (currently I'm using gamma light downscaling), which may fix the problem with this specific image, but I'm not sure. You may also have to increase the radius to 1.00, or at least higher than 0.66. Shiandow is also still working on SuperRes. So neither the version in madVR nor the one in MPDN is "final" yet.
madshi is offline   Reply With Quote
Old 27th July 2015, 18:29   #32172  |  Link
Telion
Registered User
 
Join Date: Sep 2011
Posts: 78
Quote:
Originally Posted by madshi View Post
madVR now renders in paused and stopped mode, too. When seeking, the media player usually stops playback for a short time. So madVR might think "oh, I'm in stopped mode, so I'm going to render black now". There's a workaround in place which IIRC blocks rendering from occurring for 1 second after seeks. If your seek takes longer, madVR might be tempted to render black.
Would you please increase this to 2 sec at least? I have a rather slow rig where some seeks get into this interval without a black screen, whereas some don't and I see a black screen flashing. So it's a tad irritating to have such fickle seek experience.
Telion is offline   Reply With Quote
Old 27th July 2015, 19:07   #32173  |  Link
DragonQ
Registered User
 
Join Date: Mar 2007
Posts: 934
Quote:
Originally Posted by madshi View Post
Does this test build fix it?

http://madshi.net/madVR8821b.rar
Looks like it, yes. Thank you.

By the way, are there any known issues with "use Direct3D 11 for presentation" or "use separate device for presentation" with HD4000 IGPs? On my laptop using either of these options leads to major presentation issues (frames jumping about all over the place). Not a big deal since I can disable both options, just wondering if you knew.
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7
DragonQ is offline   Reply With Quote
Old 27th July 2015, 19:11   #32174  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
Quote:
Originally Posted by madshi View Post
As I said, when using the "right" test images/videos, different chroma scalers can look vastly different from each other. So what you're saying here (and already said in your previous post) is flat out incorrect.
Right, so you don't want to have two seperate SR knobs for the number of passes and strength in order to avoid clogging the mVR settings control panel(when they both make extremely obvious changes to the picture), but OTOH it'd make perfect sense to for instance provide as many as 11 chroma upscaling algorithms with three subsettings, that might very well sometimes make barely visible differences using 400% magnification on nasty test patterns.

Reality is that 99.9999999% of mVR users watch good ole SD/HD movies and will either go cheap coz they have to, J3AR due to its high bang/bucks ratio or NNEDI3/sxbr if they wanna go all the way sharpness-wise.

Seriously, if anything is bound to confuse new users it's for instance the ability to upscale chroma using that many different algorithms, I mean who would ever use Spline for this job? No one, yet this useless option is here confusing the hell out of newbies.

The whole point of two separate knobs for SR is the ability to finetune the final picture sharpness based on the end-user personal taste, display native sharpness(due to blurry/grainy anti-reflective coatings for instance, especially on TN computer screens that look hazy compared to mineral glass Plasma's), viewing distance(the farther the sharper), the MTF/OTF sharpness of the end-user prescription eyewear lenses(organic glass and even worse polycarbonate come with very poor constringence and have to be counterbalanced), how the TV applies its RGB/YcBcR back and forth conversions, how sharp the image doubling algorithm is and so on. Botttom line is that mVR is in dire need of a very subtle sharpness knob IMO and you are against it as of now.

Quote:
Originally Posted by madshi View Post
I've told you many times already. 3 passes with 0.66 strength should be roughly identical to 2 passes with 1.00 strength. So make comparison screenshots for those two situations with v0.88.15. You can use any other combination of passes and strengths, too. Just make sure that they sum up to roughly the same value. E.g. 3 * 0.66 = 1.98. 2 * 1.00 = 2.00. And 1.98 is roughly identical to 2.00. If you insist that you have to use a strength of 0.41, then compare 3 passes with 0.41 with 2 passes with 0.61. Because 3*0.41 is roughly identical to 2*0.61.
Fair enough, I'll give it a shot but that won't change anything to the fact that 2*0.61 is impossible to achieve in the newest mVR builds as you didn't make the logic of the new strength knob public and again, even a strength of 1 in LQ mode is way too sharp to be of any use in .20.

Quote:
Originally Posted by madshi View Post
Well, some process must still have it loaded. Maybe madHcCtrl.exe is still running? Maybe you've changed the madVR tray icon setting to have it always visible?
Nope I also checked and it's not running either, only that very DLL is locked. The mVR tray icon isn't visible either. I guess I could try this app and see what's locking it up but it'll be Explorer anyway.

Last edited by leeperry; 28th July 2015 at 12:53.
leeperry is offline   Reply With Quote
Old 27th July 2015, 20:10   #32175  |  Link
KoD
Registered User
 
Join Date: Mar 2006
Posts: 567
Quote:
There is a known issue if D3D11 does not officially support the refresh rate you want to use. In that case some GPU drivers run into trouble with queues not filling properly, especially when using a large number of prepresented frames. There's no solution to this. madVR is able to make D3D11 use the refresh rate you want, but with some GPU drivers these queues-not-filling problem is the consequence. Probably if you limit yourself to 1080p60, you'll not have this problem.
Hi madshi, the problem does not seem to be related to D3D11 supported refresh rates. It happens no matter the refresh rate, from 23.976 to 60Hz. As I said, the issue only happens when madVR sends 10bit per color channel output, and the TV receives it (well, it says it receives 12bit per channel). And that happens only when using D3D11 + fullscreen, and the TV set configured as "10bit or more" in the madVR settings.

As for me, it's not important if this gets fixed or not, I am not using fullscreen mode, as it's too annoying that lag which happens when I try to display the context menu in MPC-HC to switch the audio or subtitle track. With windowed mode, and therefore no 10 bit per channel output, there are no issues, even with D3D11 (and the framerate does not matter in this case, either).

As an additional note, the sudden frame skips that happen with 0.80.20 when turning the OSD on/off and that people reported, I have also seen them - and they also happen only when using D3D11 and fullscreen mode (did not see them when using windowed mode, or D3D9). I have not yet tested 0.88.21, so I don't know if these frame skips are still present.

Who knows, maybe the issue is in the way MPC-HC interacts with madVR, since the issue happens only after two or more switches between fullscreen and windowed mode. Or it could be something weird with the nvidia driver.

Last edited by KoD; 27th July 2015 at 20:14.
KoD is offline   Reply With Quote
Old 27th July 2015, 20:21   #32176  |  Link
TheShadowRunner
Registered User
 
TheShadowRunner's Avatar
 
Join Date: Feb 2004
Posts: 399
Quote:
Originally Posted by madshi View Post
This is only a cosmetical issue, right? Why do seeks seemingly take so long in your case?
Yes it seems only cosmetic, I play rather high bitrate videos from a NAS, guess that's why.
Quote:
Originally Posted by Telion View Post
Would you please increase this to 2 sec at least? I have a rather slow rig where some seeks get into this interval without a black screen, whereas some don't and I see a black screen flashing. So it's a tad irritating to have such fickle seek experience.
Agreed, 2 or even 3 sec would be better.
__________________
XP SP3 / Geforce 8500 / Zoom Player
TheShadowRunner is offline   Reply With Quote
Old 27th July 2015, 20:26   #32177  |  Link
XMonarchY
Guest
 
Posts: n/a
Quote:
Originally Posted by huhn View Post
DXVA deinterlacing is the same deinterlacer as CUVID deinterlacer. HQ is default for DXVA deinterlacing.

and to be more precise CUVID deinterlacing doesn't allow madVR to use IVTC on the interlaced video stream. so there is nothing to win only things to loose.

and different decoder doesn't have better picture quality.

and about openCL looks like nvidia has removed cl_nv_d3d9_sharing in newer drivers. i'm not 100% sure if madVR uses that extensions. if it does everything make sense.
Any word on this from madshi regarding this? Does NNEDI3 use CL_NV_D3D9_SHARING?

With 353.49 Hotfix drivers, my OSD shows "Chroma > NNEDI64" and below "Image > Catmull-Rom AR" . It doesn't say NNEDI3 64, but just NNEDI64. Does that mean NNEDI3 is being used or not?
  Reply With Quote
Old 27th July 2015, 20:31   #32178  |  Link
aufkrawall
Registered User
 
Join Date: Dec 2011
Posts: 1,812
Yes, it must be a new typo. Why not judging by image quality?
aufkrawall is offline   Reply With Quote
Old 27th July 2015, 22:08   #32179  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
You should all really stop speculating on things that don't have a shred of proof that it matters. NNEDI clearly still works. If you want to test something, check if its slower or something.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders
nevcairiel is offline   Reply With Quote
Old 28th July 2015, 00:42   #32180  |  Link
BetA13
cosmic entity
 
BetA13's Avatar
 
Join Date: May 2011
Location: outside the Box
Posts: 258
maybe im too stupid, but how can i change lq / hq super ress?
i know i read it somewhere but cant find it anymore..
i cant see teh setting HQ/LQ . or did i miss something?

greetz

also, i cant see any difference with super ress on or off.. what do i have to look for?
its one of the only settings i dont understand yet..
BetA13 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 00:09.


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