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. |
23rd August 2013, 15:09 | #19922 | Link | |
Registered User
Join Date: May 2011
Posts: 94
|
Quote:
|
|
23rd August 2013, 15:59 | #19923 | Link | |
Registered Developer
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
|
Quote:
__________________
LAV Filters - open source ffmpeg based media splitter and decoders |
|
23rd August 2013, 16:41 | #19924 | Link |
Banned
Join Date: May 2012
Location: _Lies|Greed|Misery_
Posts: 114
|
While trying to reproduce a madVR bug I switched between LAV decoders and found the following behaviour I don't understand:
With DXVA (both native and copy-back) the madVR OSD shows much higher rendering stats then with software or CUVID decoding. I thought DXVA is lighter on the GPU than CUVID. Or do I misinterpret the meaning of rendering stats: I know it doesn't show the GPU load but I thought less decoding work for the GPU should decrease the timings? Can someone please explain this phenomena to me? Thanks. |
23rd August 2013, 17:39 | #19925 | Link | |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Quote:
And I guess the ability to change the dithering "pattern" would also be welcome because we are basically adding noise on purpose and it's tempering with/amplified by the successive lossy RGB32>YCbCr 4:2:2>RGB conversions. But yeah, I do realize that PS cannot do more than random dithering. Floyd–Steinberg and/or dithering "size" settings would be amazing someday Last edited by leeperry; 23rd August 2013 at 18:15. |
|
23rd August 2013, 17:41 | #19926 | Link |
Registered User
Join Date: Mar 2002
Posts: 2,323
|
Yes, even MPC-HC/VLC Remote (android app) supports this functionality
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config |
23rd August 2013, 19:17 | #19929 | Link |
Registered User
Join Date: Dec 2002
Posts: 5,565
|
Well, how about you test it with monitoring software? On ATI it seems to work similar: it goes into power saving mode when using DXVA decoding, apparently ATI seems to assume that video decoding = don't need much power. It does not detect that if you use software decoding, because it does not really know the difference between madVR and a video game. It's just another software making use of the GPU.
|
25th August 2013, 13:44 | #19931 | Link |
Banned
Join Date: May 2012
Location: _Lies|Greed|Misery_
Posts: 114
|
OK understood, guys.
Whenever the present queue is set lower than the render queue so the former cannot hold all rendered frames (or rendered + blended ones with FRC used) the rendering time increases. Does this mean the renderer has to wait for the present queue to free up and this waiting time is added to the rendering time? I thought having the queues filled completely was good but it's faster if the present queue is not by setting the present queue higher than the render queue. |
25th August 2013, 17:32 | #19932 | Link |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
Have any of you guys heard about a bug that makes madVR produce artifacts or something? Especially in few occasions where you're using either 59 or 60 Hz.
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
25th August 2013, 23:47 | #19934 | Link |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
Hahaha, my bad. I was kinda in a hurry there. To be more precise, this is what I'm talking about.
At first I thought it could be either a bad configuration setup or a corrupted file, but the guy who reported the issues told me that when he switches his refresh rate to 59 hz (instead of 60 hz) the issue fixes itself, but appears in other videos. He has yet to tell me which types of videos this happens (like their fps and whatnot) but I want to hear everyone's opinions, as this might be something that has already happened around here.
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
26th August 2013, 12:34 | #19935 | Link |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Adding a bit about BFI as it's really exciting to play around with all those BFI/mVR FRC goodies and now that that the BFI new toy syndrom has worn out, here are my real-world conclusions:
-24p@24Hz: interframe jitter is excellent but the 24p judder is really hiccupy, especially on slow pans. Also, flickering is fairly annoying on flat bright objects(a long shot from 100Hz CRT, though ^^) -24p@24Hz + BFI: hardly anymore backlight flickering and the synced BFI hides motion blur amazingly well. The flip side of the coin is that it acts as a motion-based EE and the picture looks really "colored" for lack of a better word. The subjective pop effect is highly increased but it doesn't look natural, it really has a "digital" flat-screen look.....my brain is not tricked into believing that it's watching through a window(the original intent of HD video). I can literally "feel" the syncopated shutter of the BFI and that really gives a "digital" unnatural look. 24p judder is still there and it's painful. -24p@60Hz + mVR FRC: Ah, goodness! This time the motion blur doesn't hide/cheat anymore, it strives to make things look a hell lot smoother. Pop effect is astounding, 24p judder is history and it literally makes me feel like I'm watching through a window, too good to be true OTOH, once BFI is disabled the darn backlight flicker becomes very annoying -24p@60Hz + mVR FRC + BFI: the BFI synced shutter kills mVR's FRC: it's nowhere near as smooth looking anymore and motion blur still looks unnatural. They do fight each other to my eyes. I should state that I'm colorblind and left-handed(we supposedly see faster but righties would hear faster) so YMMV I think 120Hz synced strobing à la nvidia is an excellent idea(too bad it only works on TN panels) but BFI on its own doesn't look natural at all once you get the hang of it IME....it's great on 50p HD DVB-T though. All this to say that I think this monitor would be full of win: http://www.youtube.com/watch?v=1CNOIbmUws0 You can compare its inexistent flickering to Sammy TV's: http://www.youtube.com/watch?v=Z-Y9bH-3Qu8 The latter flicker very fast but I can still pick it up and so do cameras....the BenQ supposedly doesn't flicker whatsoever, is said to support up to 72Hz(using minimal blanking custom timings) and IME the higher the refresh rate the smoother mVR's FRC looks. Also, BenQ is owned by AUO so no panel lottery this time(A-MVA all the way), but of course the backlight bleeding lottery is still there. I've finally scored a 145€ Asus 660 so I'll order that GW2760HS and will report back Last edited by leeperry; 26th August 2013 at 12:47. |
26th August 2013, 14:24 | #19936 | Link | |
Registered User
Join Date: May 2008
Posts: 84
|
Quote:
|
|
26th August 2013, 16:41 | #19938 | Link |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
Haha, leeperry is known for having odd/contradicting opinions on things. Don't take it for the heart.
However this is weird indeed, madVR's FRC is supposed to imitate the feeling you get with 24p@24hz. If you find madVR's method to be better then this points out to a problem with your 24hz display.
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
26th August 2013, 17:08 | #19939 | Link |
Registered User
Join Date: Jan 2009
Posts: 1,210
|
In my experience, if your display only supports <= 60Hz then using FRC exclusively for all content (except 29.97/30/60fps) is simpler and visually indistinguishable from using 24/48Hz. I've tested 24fps on 60Hz with FRC and on 120Hz without FRC at the same time on two identical displays and you simply cannot tell the difference. Even if your nose is touching the monitor while running test patterns or other videos.
It's my observation that FRC does not give a visually indistinguishable result is when the video's framerate is within ~25% of the display's refresh rate. Maybe madshi could provide better insight on this. Last edited by dansrfe; 26th August 2013 at 17:15. |
26th August 2013, 17:39 | #19940 | Link |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Because FRC blends the frames, they're not 1/2/3/4 anymore, they're 1/1.25/1.5/1.75/2 and so on. madshi made very clear that FRC was simulating an infinite refresh rate and that the higher the better.
All I'm seeing is frame blending I guess and it would appear that SVP also provides a blending mode. FI does create unacceptable artifacts on fast moving objects but I can entirely live with frame blending slight "blurring"(can't really find the right word ^^) as 24p doesn't hiccup anymore Anyway, as I said I'm colorblind and left-handed and I'm just sharing the results of my real-world experiments....If anything, I kinda have a hard time understanding how nobody else sees what I see tbh This said, 99% of the world appears to be using PWM flickering LCD monitors that give me instant headaches and eye burns, lol: LED Monitors can cause headaches due to flicker Compare the very fast flickering videos from my previous post to this nasty sluggishly flickering PWM LCD: https://www.youtube.com/watch?v=bKhDS7CHDPk I also cannot bear plasmas that just flicker like hell to my brain, but again it doesn't just happen to me. 99% of the world would be OK with that DELL monitor but you can read that the guy who posted that youtube video sent it back for a refund due to this major problem. Funnily enough, DELL sent non-PWM samples of that S2440L monitor for review purposes and BenQ is now putting its money on a new line of non-flickering A-MVA monitors, too good to be true² What I've learned in the audio world is that we all hear differently, the same way we all see differently. The idea is to share experiments results and find ways to improve our own experience, nothing more really and madshi has really saved my day as far as my HTPC is concerned It would appear that some digital displays use CRT-like tricks when fed 24p so YMMV again, but unprocessed 24p judder is really hard to bear IME on a LCD screen: New Sony HDTV’s with Motionflow Impulse Last edited by leeperry; 26th August 2013 at 22:05. |
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
Thread Tools | Search this Thread |
Display Modes | |
|
|