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 July 2013, 16:36 | #19621 | Link | ||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Have you added a bug entry to the bug tracker about this? If so, there's no need to discuss it here in the forum any further. No need to post a link to your tracker issue, either. I'll get to that when I find the time... |
||
19th July 2013, 16:38 | #19622 | Link | |
Registered User
Join Date: Apr 2012
Posts: 169
|
Quote:
|
|
19th July 2013, 16:41 | #19623 | Link |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
To be honest, I've no clue about which generation is 16bit alpha blend capable and which is not. That's not a spec that's advertized anywhere. I don't think there's even a "CAP" bit for that in Direct3D, so I can't even ask Direct3D or the GPU itself if it can do it. We'll have to collect information from users to see which GPU generations exactly support XySubFilter+madVR and which don't...
|
19th July 2013, 17:02 | #19625 | Link |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Just tried with my old NVidia 9400 mainboard. I can reproduce the problem there. *However*, by enabling the option "use 10bit image buffer instead of 16bit" the problem goes away. So that should be a good workaround for older GPUs.
|
19th July 2013, 17:09 | #19626 | Link | |||||
Chief Blur Buster
Join Date: Sep 2002
Location: Toronto
Posts: 38
|
I'm the operator of the Blur Busters Blog, and creator of TestUFO motion tests, so I've become an expert in motion blur...
First: I agree, low priority issue for now, but this is an early canary warning that this is going to be slowly a more important feature in the future. However, I would like to inform you of a few things: Quote:
Quote:
Quote:
Software based BFI (60fps on 120Hz) would do this on normal 120Hz LCD monitors: --> 8.3ms visible:8.3ms dark == 50% reduction of motion blur However for doing 60fps @ 120Hz software BFI on a LightBoost display, something rather interesting happens: less motion blur on LCD than plasma (LCD colors are still terrible, though) LightBoost flashes the backlight for as short as 1.4ms, so over two frames (one 8.3ms visible, one 8.3ms black, total 16.7ms cycle) you are now are getting: --> 1.4ms visible + 6.9ms backlight off + 1.4ms black frame + 6.9ms backlight off So effectively, you've combined hardware-based BFI (simulted via backlight) and software-based BFI (actual black frame), you are now getting: --> 1.4ms visible + 15.3ms black frame (combined hardware+software) This is a 92% reduction in motion blur over a 60Hz LCD (1.4ms of sample-and-hold blur, instead of 16.7ms of sample-and-hold blur) (web animation of sample-and-hold blur). Which produces CRT-quality motion of 60Hz sources on LightBoost LCD displays; that's less motion blur than plasma displays. This is since plasma displays have about 5ms of motion blur, due to red/green phosphor decay, and plasma motion blur becomes noticeable during very fast pans (e.g. when viewing TestUFO: Moving Photo test at 1440pixels/sec) while this Moving Photo test remains crystal sharp on a LightBoost display, especially at LightBoost=10%. I should point out that at 1440 pixels/sec, 1ms of sample-and-hold motion blur equals 1.4 pixels of motion blur. So 5ms does add up to quite a lot of motion blur during fast high-def pans where there is no source-based motion blur (e.g. soft focus, long exposure, overcompression). For 1080p 60fps high-bitrate video (especially 4K60fps downconverted to 1080p@60fps, to create ultra-sharp high-contrast edges), it is possible to begin to notice 1ms differences in motion blur during fast screen-width-per-second panning (1920 pixels/sec), because that's currently adding 1/1000th of 1920 = 2 pixels of motion blur. Very subtle. But human eyes can see 2 pixel differences at 1080p, if you're not too far from the display. So, it's in favour to expand the ratio of blackness to visible frame, shave off the milliseconds of visible frame, and that's where combining hardware BFI with software BFI can come in, to bypass a hardware BFI limitation (e.g. LightBoost not functioning at 60Hz). The 92% reduction of blur is also confirmed via this graph too, during motion measurement tests (see graph, as well as various testimonials). However, I agree, most people won't be watching movies on a computer monitor, but it bears worth pointing out that LightBoost computer monitors have recently leapfrogged over televisions in terms of achievable motion resolution, even though they have terrible color quality compared to televisions. Quote:
(1) You don't need to rapidly switch the LCD pixel color value from on/off/on/off. This can reduce contrast severely. Hardware BFI assisted via backlight method (rather than LCD panel method) doesn't necessarily need to turn on/off LCD pixels or even completely; it just turns on/off the backlight at precisely synchronized intervals (or segments of backlight at a time, as in scanning backlights) Although terminology should not be used interchangeably, BFI/scanning backlights/strobe backlights all essentially, at the human eye level, roughly equivalent -- these approaches simply adds a dark period between refreshes -- to reduce the sample and hold effect (animation demo of sample and hold effect -- this animation looks very different on CRT than LCD). (2) BFI at the LCD panel level (the only method available to software) also potentially causes some nasty inversion artifacts on some panels, due to the interference with the positive/negative voltages (lagom pixel-walk, techmind inversion explanation). So fine-single-pixel checkerboard texture patterns can appear during BFI. (3) Some LED backlights are designed to pulse brighter when on, so you can compensate for BFI dimness by using brighter backlight flashes to compensate. Doing hardware-based BFI can avoid these problem. The hardware-based BFI avoids that by using the backlight instead, preventing the need for LCD pixels to go back-and-fourth. Quote:
I agree, low priority issue for now, but this message is an early canary warning that this is going to be slowly a more important feature in the future. Last edited by Mark Rejhon; 19th July 2013 at 17:33. |
|||||
19th July 2013, 17:09 | #19627 | Link | |
Registered User
Join Date: Apr 2012
Posts: 169
|
Quote:
I couldn't see any difference in the image between having 10 and 16Bit buffer, what does it really affect? |
|
19th July 2013, 17:28 | #19628 | Link | ||||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Quote:
Quote:
------- So here's the new official xy-VSFilter / XySubFilter thread: http://forum.doom9.org/showthread.php?t=168282 Please post subtitle specific questions there, I mean things like incorrectly formatted subtitles etcs. If you have subtitle questions that are more madVR related than XySubFilter, you can post them here instead, e.g. if you run into GPU performance problems with XySubFilter or things like that... Last edited by madshi; 19th July 2013 at 17:30. |
||||
19th July 2013, 17:33 | #19629 | Link |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Hi Mark, good to see you around
Apparently Sammy TV's don't do full frame BFI, just like a CRT, more like half/quarter vertical height or so....anyway yes, by the time mVR reaches 1.0 we will see whether there are more HDMI2/DP 120Hz+ displays on the market. |
19th July 2013, 18:06 | #19630 | Link |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
I've made an official thread for the xy-VSFilter Project, which I'd encourage everybody to use for any XySubFilter discussions not directly related to madVR:
xy-VSFilter Project (High Performance VSFilter Compatible Subtitle Filters) |
19th July 2013, 18:21 | #19632 | Link | |
Registered User
Join Date: Jun 2011
Posts: 121
|
Quote:
As for the refresh rate, not sure whats the correct refresh rate of my screen but CTRL+ J gives me 60.0014HZ on my sony TV and 59,9475HZ on my DELL desktop monitor Not sure what this means, but i have no ghosting whatsoever on my desktop monitor and on my sony tV i have i`ve just installed xysub but i dont know where is the option to put subtitles in the botton black bar on 2.35 movies TIA |
|
19th July 2013, 18:31 | #19633 | Link |
Registered User
Join Date: Apr 2012
Posts: 169
|
Currently, madvr sets both the decoding and subtitle queue via cpu queue parameter. Could we have those separated in the future (if it makes sense internally)?
edit: About the only reason I'd want to have it is for more responsive jumping, since the decoding queue never falls below 63/64 while subtitle queue does drop to 11/64 on some heavily styled subs (and that's on a 4GHz X4 970BE cpu...) Last edited by Soukyuu; 19th July 2013 at 18:37. |
19th July 2013, 19:57 | #19634 | Link |
Registered User
Join Date: Jan 2009
Posts: 1,210
|
What exactly is the goal in setting queues? Should we bring it as low as possible or as high as possible? Does increasing the queue size add lag to initial render steps and/or skipping around the video? How do we know what the optimal queue sizes are for the machine we use?
|
19th July 2013, 20:00 | #19635 | Link | |
Registered User
Join Date: Dec 2002
Posts: 5,565
|
Quote:
I didn't want to open a new thread myself because I was hoping for an official one. It's better if a team member does it, so he can always update the start post with fresh info straight from the source. |
|
19th July 2013, 20:27 | #19636 | Link | |
Registered User
Join Date: Apr 2012
Posts: 169
|
Quote:
So to find out if your queue needs adjusting, play a video of your choice and look at queue numbers on the OSD. If you get dropped frames and any queue shows as empty (or dangerously low), try increasing it by a few steps. I imagine the default values are sufficient in the majority of scenarios. So far I only saw ONE video with .ass subtitles that dropped frames on default settings (the same I had to increase my cpu queue to 64 for). Even those subs that used to be regarded as "very heavy" with VSFilter (not Xy-VSFilter) decode without problems on my machine on default settings. |
|
19th July 2013, 21:51 | #19638 | Link |
Registered User
Join Date: Mar 2007
Posts: 934
|
madshi, I have a log where I closed MPC-HC 1-2 seconds after I saw a black frame using smooth motion. How much of the 3.25 GiB log do you need?
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7 |
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
|
|