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. |
19th October 2011, 00:07 | #10201 | Link | |||
Registered User
Join Date: Feb 2004
Posts: 399
|
Quote:
Quote:
Since I have no issues with madVR going windowed <> FSE, I disabled "Display OSD through madVR's OSD API" completely. Quote:
As I was reporting, it doesn't make any difference when it comes to this bug if madVR does it internally, or if ReClock or I do it manually. From the log I enclosed, do you see any issue on the madVR side?
__________________
XP SP3 / Geforce 8500 / Zoom Player |
|||
19th October 2011, 00:36 | #10202 | Link | |||||
Registered User
Join Date: Apr 2009
Posts: 1,019
|
Quote:
Seems like it could be a pretty good feature—my LCD has 70 zones and while uniformity is pretty good for an LCD as a result, there doesn't seem to be any uniformity correction available via the service menu, so doing it through the HTPC would be interesting to see just how much better things could get. Quote:
WTW visibility is debatable. Reference-class displays are still required to show 16–255, though in reality most content does not have anything above 235 and it can improve the contrast on many current displays. The option to preserve WTW would be nice though. Quote:
Quote:
Quote:
Here are a couple of crops from Tron Legacy, at 200% My current 3DLUT with data from 20–100% in 10% steps: Shows some discolouration/banding but tolerable for the improvements using a 3DLUT brings. 3DLUT with data from 10–100%: As you can hopefully see, there is a noticeable increase in banding and reddish discolouration in the darker areas. (might be easiest to open them in separate tabs in your browser and switch between the two) This is with modified data too—I only used the Y value from my display at 10%, xy are set to D65, as that seemed to reduce the errors. I'm hoping that 10-bit out to the display will help in reducing those errors. Of course, this all depends on what kind of display you're using. With CRT, any kind of reduction in bit-depth is immediately obvious. On LCDs and LCoS/SXRD it's fairly noticeable, on Plasmas it probably isn't noticeable at all, as they add colour errors like this on their own. It's the main reason I couldn't stand to keep any of the Kuros. This is partly why I would like the ability to tweak the output RGB values being sent to the display from a LUT, rather than giving it input values. I have had success in reducing this kind of error by manually tweaking the LUT with external devices. (but ultimately, none of them have been worth the additional expense over madVR & yCMS) |
|||||
19th October 2011, 04:00 | #10203 | Link |
quack quack
Join Date: Apr 2009
Posts: 259
|
madVR is screwing with keyboard shortcuts in other programs. Specifically Ctrl+Shift+R
Is this a shortcut used by madVR? Can I have an option to disable all shortcuts plox? Feature request: Settings >> rendering >> windowed mode settings >> new option: show seek bar Thx. madVR rocks |
19th October 2011, 07:09 | #10205 | Link | |||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Quote:
Has been posted before. On my to do list, under "low priority". Last edited by madshi; 19th October 2011 at 07:14. |
|||
19th October 2011, 07:21 | #10206 | Link | |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Make sure VSync correction is disabled in Reclock. You reported that you used a profile to fix the clocks. Have you verified whether the profile actually works as intended? |
|
19th October 2011, 07:33 | #10207 | Link | |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
|
|
19th October 2011, 08:03 | #10208 | Link | |
Registered User
Join Date: Feb 2004
Posts: 399
|
Quote:
Zoom Player is really frozen, and hangs there. In this state, its own guarddog kills it (after 20 seconds, it's a feature), or if the guarddog is disabled the only option is to kill the zplayer.exe process manually (what I did for that log, in order to keep it small). I'll try to come up with a method to always recreate the issue... A maybe related question: when frequency switching is involved, exemple 720p@60 to 1080p24, sometimes the switch is instantanous (exactly when the media starts) and sometimes [for the same media!] it takes up to 6-7 seconds before the switch occurs, during which a couple seconds of the media are looped (5-6 times). What is this difference due to ? (I'm asking for madVR's internal frequency switcher specifically)
__________________
XP SP3 / Geforce 8500 / Zoom Player |
|
19th October 2011, 08:06 | #10209 | Link | ||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
|
||
19th October 2011, 08:14 | #10210 | Link |
Registered User
Join Date: Nov 2009
Posts: 2,352
|
Is it the dither error diffusion or ordered dither? I say because I encode my rips in ordered dither (to save bitrate), I have tested ordered dithering over ordered dither and results are horrible so by the moment I have dithering disabled just in case.
I would also like to suggest a profile load option. Or some way where you can switch fast exclusive mode with windowed mode with its preconfigured settings. That is because I normally use mpc-hc to play small files, preview my encodes, watch clips downloaded from youtube, etc, and windowed mode is more convenient as I can pause, do a few things aside, resize window, I mean, more manageable. Other times I would just plug my laptop to the TV use exlusive mode, and Im good to go. I also envy how you managed to port the resizers to be processed by the GPU, it would be great to see spline36 running on GPU for avisynth as well. Welcome back!
__________________
i7-4790K@Stock::GTX 1070] AviSynth+ filters and mods on GitHub + Discussion thread |
19th October 2011, 08:27 | #10211 | Link |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Neither nor. I'm currently using TPDF (random) dithering. I would love to implement some kind of error diffusion, but the basic principle of error diffusion is to spread the error to neighboring pixels. And that contradicts the basic principle of how GPUs work. Maybe I'll find a way to make error diffusion work by using CUDA/OpenCL some time in the future, but it would probably be quite slow, due to the algorithm contradiction I mentioned.
I don't think madVR's dither should collide with ordered dither. Of course it's always better to not dither at all, but to use a higher bitdepth instead. But if you can't use a higher bitdepth, then it's usually better to use dither than to not use it. Maybe applying ordered dither twice is bad, that's quite possible. I have no personal experience with ordered dither. Anyway, that doesn't apply to TPDF dithering (used by madVR). TPDF dithering can be applied multiple times without any negative side effects, other than that applying it multiple times of course raises the noise floor. In audio processing TPDF dithering is used all the time. Maybe in some future version. Not anytime soon, though. Too many crucial things missing to spend my time on things like that right now. Last edited by madshi; 19th October 2011 at 08:31. |
19th October 2011, 08:43 | #10212 | Link |
Registered User
Join Date: Feb 2006
Posts: 1,076
|
Madshi, welcome back after a well-earned break.
First: I've been using madVR for a short period now, coming from JanWillems test-builds. Must say i'm satisfied with madVR's stable output and quality. JanWillems builds however had a possibility to convert 8-bpc video to 10-bpc, and output that. It uses EVR-CP, i know, but somehow it was possible to force 10-bit output. I have the luck that i have equipment that accepts this 10-bit video. So 'bit-wise' that seems to be the better option. But JanWillems builds aren't always stable (thats logical, they aren't labeled experimental for nothing), and also AVR-CP isn't always stable. Thats why i choose to work with madVR. Some posts in this thread before there was some talk about 10-bpc maybe coming to madVR. Can you comment about this, do you think it is possible, and/or are you working on this subject ? Also some posts before there was some talk about MPC-HT with madVR not exiting full-screen correctly. I want to report that i have also noticed this. I run full-screen on my secondary display. When exiting full screen, and coming back to my primary display the MPC-HT's window doesn't have the window-controls in the top bar anymore (dont know what they are called, but i mean the 3 small buttons in the right top for minimizing/maximizing/closing the window). These controls are 'simply not there anymore' Last edited by G_M_C; 19th October 2011 at 09:20. |
19th October 2011, 08:44 | #10213 | Link | ||
Registered User
Join Date: Feb 2004
Posts: 399
|
Quote:
I can reproduce the bug at will here.. only with madVR. Maybe these logs can help. I do _exactly_ the same for both tries, and close ZP in between of course. I launch ZP, play a SD PAL video and then close the player. (madVR's internal frequency switcher is used: 720p60 to 720p50) http://videoff7.free.fr/madVR_ZP_OK.zip (fast ZP control bar, all OK) http://videoff7.free.fr/madVR_ZP_buggy.zip (slow ZP control bar, freeze if seek) Quote:
__________________
XP SP3 / Geforce 8500 / Zoom Player Last edited by TheShadowRunner; 19th October 2011 at 09:04. |
||
19th October 2011, 09:07 | #10214 | Link | |
Registered User
Join Date: Aug 2009
Posts: 136
|
Quote:
|
|
19th October 2011, 10:44 | #10215 | Link | ||
Registered User
Join Date: Apr 2009
Posts: 1,019
|
Quote:
Quote:
It is a combination of both what is being done with the LUT, and how those values end up being displayed on the screen. Using higher precision does have benefits in this area beyond reducing dithering. This is why high end displays use much higher precision LUTs for internal calibration. High-end Eizo monitors use 16-bit LUTs to drive the display panel for example: (of course the panels themselves are not 16-bit native, but they aren't 8-bit either) When you are only sending 8-bit to a display, changing the RGB values being sent to it by even a single digit can have considerable effects on greyscale at the lower end of the scale. With 10-bit, you have four times the precision and it really does make a difference in reducing visible errors. (another reason why I want to manually tweak the LUT output values) I have spent a considerable amount of time hand-tuning LUTs with dedicated devices, and this is a definite limitation caused by a lack of precision when only outputting 8-bit. |
||
19th October 2011, 10:45 | #10216 | Link | |||||||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Quote:
Quote:
Fact is, the problem only occurs with the combination of ZP + madVR. Use a different renderer and/or media player and the problem is gone. So we have no proof. It could be either ZP or madVR or it could be how those two interact. The log you uploaded earlier does not indicate a madVR problem. Quote:
Quote:
Quote:
|
|||||||
19th October 2011, 10:57 | #10217 | Link | ||||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Quote:
Quote:
Those hand-tuned LUTs with dedicated devices, do they do trilinear interpolation? Which input/output bitdepth does the LUT have? Which bitdepth does the display accept and is the LUT output dithered down to the native display bitdepth? |
||||
19th October 2011, 11:17 | #10218 | Link | |
Registered User
Join Date: Feb 2006
Posts: 1,076
|
Cant say for sure, cause looking better is kind of an subjective argument/feeling. But i have the feeling it did. Simply because i could turn of dithering in JanWillems build (or set it the lowest random setting), and outputting the 10-bit video to my receiver, wich in turn does 'its thing' using a Reon-VX video processor. Image seemed to be clearer, having more 'color-fidelity'. But as i said: This is subjective to be sure, i have to do some comparison between the two.
But then again: Another important argument for madVR is that using madVR with a trunk build of mpc-ht is more stable then Jan's experimental builds. Quote:
I'm using 0.74, which is latest afaik. And your right, not using D3D11 mode fixes the problem. I'll keep using the D3D9 path. I set it to D3D11, simply without thinking (like: I have D3D11, so might as well use it). Last edited by G_M_C; 19th October 2011 at 11:22. |
|
19th October 2011, 13:35 | #10219 | Link | |||
Registered User
Join Date: Feb 2004
Posts: 399
|
Quote:
Quote:
Quote:
__________________
XP SP3 / Geforce 8500 / Zoom Player |
|||
19th October 2011, 14:23 | #10220 | Link |
Registered User
Join Date: Mar 2011
Posts: 90
|
sorry to repeat myself but i am looking to pull the trigger today on a new GPU
contenters are the nvidia gt 430 or AMD 5xxx initila research looks like the 430 was blighted on launch by bad drivers but i cant find if this is sorted as i use MadVr and i know you lot are critical about picture quality like me can anyone suggest which is best? and if amd which model within the 5xxx series ar is a basic 6xxx better thanks in advance |
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
Thread Tools | Search this Thread |
Display Modes | |
|
|