View Full Version : madVR - high quality video renderer (GPU assisted)
huhn
4th April 2016, 00:23
i guess most are using window mode.
there is no simply answer. if you don't see any issues as it is now you can leave it as it is. if you get dropped frames or high present times try a different mode.
the number of presentation modes is pretty high if you count backbuffer and present in advance in.
on a dual display setup window mode as a chance to fail compared to FSE or overlay mode.
and the options names you are using here aren't the names in the madVR options. so i can only guess what the program you are using is changing in the madVR settings.
70MM
4th April 2016, 01:11
i guess most are using window mode.
there is no simply answer. if you don't see any issues as it is now you can leave it as it is. if you get dropped frames or high present times try a different mode.
the number of presentation modes is pretty high if you count backbuffer and present in advance in.
on a dual display setup window mode as a chance to fail compared to FSE or overlay mode.
and the options names you are using here aren't the names in the madVR options. so i can only guess what the program you are using is changing in the madVR settings.
Im using JRiver with madvr.... JRiver couldn't change the option names in madvr could it?
Asmodian
4th April 2016, 01:45
Im using JRiver with madvr.... JRiver couldn't change the option names in madvr could it?
Are you using madVR's option window or JRiver's?
If you are using MadVR's using the full option names is helpful to avoid confusion. :)
To answer your questions: use whichever of these options offer you no dropped frames and/or presentation glitches. All these options exist because there doesn't seem to be a set of them that works best on every system so it takes some experimentation to find what works best for you.
I run only 'enable windowed overlay', with 'enable automatic fullscreen exclusive mode' disabled. I also have "use Direct3D 11 for presentation" disabled (it overwrites 'enable windowed overlay' too).
This isn't best for everyone but it works very well for me. There is nothing wrong with using fullscreen exclusive mode too, so madVR uses exclusive instead of overlay when fullscreen, but I do not need fullscreen exclusive for stable playback.
ryrynz
4th April 2016, 03:18
I am curious why when my 980ti was only at 80-90% usage it was dropping a number of frames each second. Surely that only happens when the GPU is hitting 100%?
That is basically 100%. You can't max it out.
QBhd
4th April 2016, 03:55
That is basically 100%. You can't max it out.
Indeed... My R270X can go as high as 91% before frames drop, but not all the time. 88-90% is the normal max usage.
QB
70MM
4th April 2016, 05:03
Are you using madVR's option window or JRiver's?
If you are using MadVR's using the full option names is helpful to avoid confusion. :)
To answer your questions: use whichever of these options offer you no dropped frames and/or presentation glitches. All these options exist because there doesn't seem to be a set of them that works best on every system so it takes some experimentation to find what works best for you.
I run only 'enable windowed overlay', with 'enable automatic fullscreen exclusive mode' disabled. I also have "use Direct3D 11 for presentation" disabled (it overwrites 'enable windowed overlay' too).
This isn't best for everyone but it works very well for me. There is nothing wrong with using fullscreen exclusive mode too, so madVR uses exclusive instead of overlay when fullscreen, but I do not need fullscreen exclusive for stable playback.
I am using madvrs option window.
Im going to try and turn some of these settings off like you and check for dropped frames. Maybe the more turned off without dropped frames is better...:rolleyes:
70MM
4th April 2016, 06:46
Im using the GTX 970 card and have my dithering set to Erro Diffusion - option 2, is my card DX 11 compatible?
Am I using the right dither setting?
Error Diffusion - option 2
Use Direct-Compute to perform very high-quality error diffusion dithering. Low-noise, mid dither pattern. Requires a DX 11-compatible graphics card.
Betroz
4th April 2016, 07:15
Im using the GTX 970 card and have my dithering set to Error Diffusion - option 2, is my card DX 11 compatible?
Yes...your card is DX12 compatible too. I use the same setting.
khanmein
4th April 2016, 07:22
Im using the GTX 970 card and have my dithering set to Erro Diffusion - option 2, is my card DX 11 compatible?
Am I using the right dither setting?
Error Diffusion - option 2
Use Direct-Compute to perform very high-quality error diffusion dithering. Low-noise, mid dither pattern. Requires a DX 11-compatible graphics card.
i'm also using option 2
Betroz
4th April 2016, 07:26
You guys were right. A 980Ti for MadVr in a HTPC case is too much, with temps going up to 83C...so I swapped it out for my 970 card (and that gets hot enough).
When I use Super-Xbr for image doubling/quad and there is a scene when peoples faces are up close, their faces are nice and sharp, but when in the background their faces get much more grainy. With NNEDI3 64 and up, faces in the background are more sharp than with super-xbr. So this is normal? Kind of bugs me out cause I want things to be perfect :D
ryrynz
4th April 2016, 10:47
When I use Super-Xbr for image doubling/quad and there is a scene when peoples faces are up close, their faces are nice and sharp, but when in the background their faces get much more grainy. With NNEDI3 64 and up, faces in the background are more sharp than with super-xbr. So this is normal? Kind of bugs me out cause I want things to be perfect :D
Sounds normal to me, you can supplement super-xbr with some additional sharpening though and find the right mix of options to give you the effect you're after.
Queiroz
4th April 2016, 11:17
Hello everyone.
My 60" 1080p plasma died on me a few weeks ago then i bought a new UHD 4K 60" screen a few days ago and now i have trouble to watch almost everything, framedrops, stutter, i will explain the situation, my receiver Onkyo TX-NR525 can do only 4K@30hz and htpc video card is gtx 670 so i just want to play my videos smoothly.. is it still possible?
I'm using shark007 codecs + mpc-be madvr x86 + reclock
I'll uninstall everything and use the mpc-be madvr 64bits edition now without reclock.
I hope it works.
Any advice would be really appreciated.
Thanks in advance.
Kind Regards,
Queiroz
XMonarchY
4th April 2016, 12:23
I enlighten a lot of people about madVR and many of them ask about what HTPC they should build JUST for film playback. I usually advice Intel i5 + GeForce GTX 960 due to its overall good performance and H.265 Hardware Rendering capabilities.
Assume the system is going to use the latest versions of MPC-HC x86 (interval LAV filters disabled), ReClock, LAV Filters (external x86 + x64 install package), and madVR. Would i5 + GTX 960 run with settings listed below without issues?
LAV Video Settings:
Hardware Decoder - DXVA2 (Copy-Back) - All Resolutions (SD, HD, UHD 4K, H.264, HEVC, VC-1, MPEG-2, DVD, VP9), except for MPEG-4
Output Formats - All enabled, except for AYUV.
RGB Output Levels - Untouched (as input).
Hardware/GPU Deinterlacing - Disabled (No interlacing).
Software Deinterlacing Algorithm - No software deinterlacing (disabled).
madVR Settings for 1080p:
Devices
Properties - 0-255, 10bit, no 3D, HDR - 120nit
Calibration - Rec. 709 3DLUT + Rec. 601 3DLUT
Display Modes - Treat 25p movies as 24p, 1080p23, 1080p24, 1080p59, 1080p60
Processing
De-Interlacing - Default
Artifact Removal - Enabled, Low (Default strength) + Medium (Fade in/out strength) for High Quality content, Medium (Default strength) & High (Fade in/out strength) for LQ content
Image Enhancements - None enabled
Zoom Control - None enabled
Scaling Algorithms
Chroma Upscaling - NNEDI3 32n + SuperRes w/ Strength 3
Image Downscaling - SSIM 2D w/ Strength 100% + AR Relaxed + Scale in LL
Image Doubling - None enabled
Image Upscalnig - Jinc + AR
Image Refinement - SuperRes w/ Strength 3, Refine the image every ~2x upscaling step
Rendering
General Settings - Exclusive FullScreen, Use D3D11 for presentation, Present a new frame for every V-sync, CPU Queue - 16, GPU Queue - 8
Windows Mode - Present several frames in advance, video frames presented in advance - 3
Exclusive Mode - Present several frames in advance, video frames presented in advance - 3
Stereo 3D - None enabled
Smooth Motion - None enabled
Dithering - ED Option 2, Use Colored Noise, Change dither for every frame
Trade Quality for Performance - None enabled
Questions:
- I know madVR is not very CPU-dependent, but would Intel i3 CPU instead of Intel i5 CPU handle such high quality settings?
- If the settings I provided would not provide a problem-free performance, then:
A. What minimum specs would you recommend for HTPC that would use the settings listed above without any dropped/delayed frames or presentation issues?
B. What settings would you change/advice to change to make Intel i5 (or Intel i3) + GTX 960 run problem-free and yet maintain the highest quality possible for such a rig?
XMonarchY
4th April 2016, 14:24
You guys were right. A 980Ti for MadVr in a HTPC case is too much, with temps going up to 83C...so I swapped it out for my 970 card (and that gets hot enough).
When I use Super-Xbr for image doubling/quad and there is a scene when peoples faces are up close, their faces are nice and sharp, but when in the background their faces get much more grainy. With NNEDI3 64 and up, faces in the background are more sharp than with super-xbr. So this is normal? Kind of bugs me out cause I want things to be perfect :D
Yeah, for me even faces get too grainy and just look low-quality when I use XBR for Doubling, even though I do not use any sharpening features. NNEDI3 Luma Doubling is absolutely stunning and IMHO the strongest and the most image-improving aspect of madVR, aside from maybe ED dithering. With all other very high quality settings + SuperRes make it a hell of a lot better.
Queiroz
4th April 2016, 14:48
I've uninstalled everything here...
Now i'm using MPC-BE with MadVR x64 and LAV codecs x64 and after using the settings posted by the user XMonarchY, it is finally working for me.. i mean my 1080p videos are working just fine w/ no stutter.. i just have one question about upscaling, now when i press "info" on the TV remote it says "1920x1080 / 24p" the tv still does the upscaling from hd and lower content ? sorry if i'm asking in the wrong section but any help would be appreciated... also i'm using DXVA Cuda which seems to be working, do i have to change to Copyback? Windows 10 x64 pro here with latest nvidia drivers
Asmodian
4th April 2016, 17:39
DXVA copyback is better than CUDA now but you do not have to change it.
It sounds like you have Windows set to output 1080p to your TV. If you want madVR to upscale to 4K (recommended) then you need to set the resolution to 4K in Windows.
Uoppi
4th April 2016, 17:51
Questions:
- I know madVR is not very CPU-dependent, but would Intel i3 CPU instead of Intel i5 CPU handle such high quality settings?
I have a dual core Ivy Bridge Pentium from 2013 coupled with GTX 960. In my use at least (1080p, 2D), the GPU is maxing out and not the CPU.
Queiroz
4th April 2016, 17:55
DXVA copyback is better than CUDA now but you do not have to change it.
It sounds like you have Windows set to output 1080p to your TV. If you want madVR to upscale to 4K (recommended) then you need to set the resolution to 4K in Windows.
I see.
My windows is set to 3840 x 2160 / 30p (due to my receiver limitation) because if i plug my video card direct to the TV it display 4K@60hz anyways i can't upgrade my receiver so i have to find a way to play my videos,shows,movies..
but i think MadVr changes the resolution because of the option that XMonarchY posted.. i've allowed Madvr to autochange the resolution to 1080p23, 1080p24, 1080p59 and 1080p60 and it changes to 1080p24 almost everytime...
it is working just fine i just want to know if the TV itself is doing any upscale or if i have to do it using madvr options.
:thanks:
BluesFanUK
4th April 2016, 18:13
DXVA copyback is better than CUDA now but you do not have to change it.
It sounds like you have Windows set to output 1080p to your TV. If you want madVR to upscale to 4K (recommended) then you need to set the resolution to 4K in Windows.
I thought using the LAV video decoders like CUDA and DXVA weren't worth the hassle?
sneaker_ger
4th April 2016, 18:25
If your CPU is fast enough to play your source material you don't need to bother with hardware decoders. That hasn't changed.
huhn
4th April 2016, 18:30
@XMonarchY
don't use 10 bit output.
the number of 10 bit panels out there so unbelievable low.
and samsung and sony are not revealing there panel bit.
for the simple reason that good processing with an 8 bit panel is better than a poor processing with an 10 bit panel.
and you forget something very very important and that is resolution.
XMonarchY
4th April 2016, 21:24
@XMonarchY
don't use 10 bit output.
the number of 10 bit panels out there so unbelievable low.
and samsung and sony are not revealing there panel bit.
for the simple reason that good processing with an 8 bit panel is better than a poor processing with an 10 bit panel.
and you forget something very very important and that is resolution.
I didn't forget resolution. I mentioned it 4 times - 1080p23, 1080p24, 1080p59, 1080p60 , but I added another 1080p mention to the madVR Settings sub-title.
Many plasma sets can't do proper 10bit or can do it only at 16-235, but I've tested my CCFL LCD (12bit input in NVidia CP) with 10bit tests and it passed all of them. There's no gradation/banding or at least a strong reduction compared to 8bit. On MY TV, 12bit input in NVidia CP (10bit colors + 2bit from internal processing) and 10bit in madVR is the best option.
If I were to advice someone something like Panasonic ST/VT/ZT series, then of course I would suggest to select 8bit in NVidia CP and 8bit in madVR :) .
bcec
5th April 2016, 03:20
I see.
My windows is set to 3840 x 2160 / 30p (due to my receiver limitation) because if i plug my video card direct to the TV it display 4K@60hz anyways i can't upgrade my receiver so i have to find a way to play my videos,shows,movies..
I recommend you run two cables. One from your gfx card to your TV (1), the other from your gfx card to your receiver (2). From the perspective of your HTPC, it will think it has two monitors connected to it, your TV and Receiver. You will then use (2) for audio. This way, you can enjoy sending 4k @ 60Hz on your TV while still using your existing receiver for audio. This is my current setup.
(1) can be done via HDMI 2.0 to HDMI 2.0 if your gfx card supports HDMI 2.0, or via DP-to-HDMI 2.0 adapter.
(2) can be done even with a DVI to HDMI cable.
huhn
5th April 2016, 03:24
I see.
My windows is set to 3840 x 2160 / 30p (due to my receiver limitation) because if i plug my video card direct to the TV it display 4K@60hz anyways i can't upgrade my receiver so i have to find a way to play my videos,shows,movies..
but i think MadVr changes the resolution because of the option that XMonarchY posted.. i've allowed Madvr to autochange the resolution to 1080p23, 1080p24, 1080p59 and 1080p60 and it changes to 1080p24 almost everytime...
it is working just fine i just want to know if the TV itself is doing any upscale or if i have to do it using madvr options.
:thanks:
try 2160p23, 2160p24, 2160p25, 2160p29, 2160p30, 1080p50, 1080p59, 1080p60.
or if you have an iGPU try to connect the iGPU to the receiver and direct connect the TV.
6233638
5th April 2016, 10:12
Sorry that I haven't had a chance to look at any of the new anti-ringing options at all yet.
Thanks for the chroma scaling option to restore the old behavior, chroma quality is noticeably improved now - though it obviously comes at a performance cost.
There seems to be a bug with linear light downscaling and mixed x/y scaling.
I'll try to upload a sample this evening which demonstrates the problem.
What's happening is that, with luma x downscaling and luma y upscaling, activating linear light downscaling results in a very dark image.
If NNEDI3 luma doubling is activated under these conditions (no chroma) the image turns green.
This is unaffected by the new chroma quality option.
EDIT: It's the combination of sigmoidal light upscaling, and linear light downscaling that seems to cause this.
Disabling either one fixes it.
XTrojan
5th April 2016, 11:03
I recommend you run two cables. One from your gfx card to your TV (1), the other from your gfx card to your receiver (2). From the perspective of your HTPC, it will think it has two monitors connected to it, your TV and Receiver. You will then use (2) for audio. This way, you can enjoy sending 4k @ 60Hz on your TV while still using your existing receiver for audio. This is my current setup.
(1) can be done via HDMI 2.0 to HDMI 2.0 if your gfx card supports HDMI 2.0, or via DP-to-HDMI 2.0 adapter.
(2) can be done even with a DVI to HDMI cable.
The only issue with this is that 2 monitors are registered, won't the other unused monitor consume resources? And cause compability issues?
bcec
5th April 2016, 17:05
The only issue with this is that 2 monitors are registered, won't the other unused monitor consume resources? And cause compability issues?
Nope. At least I have not observed any drawbacks at all.
mysterix
5th April 2016, 20:47
I use LAVFilter to decode video and it has it's own dithering options:http://forum.doom9.org/attachment.php?attachmentid=15342&stc=1&d=1459885097So, as it seen, it has not "no dithering" mode. On the other side, MadVR has not "no dithering" mode too. Does it mean that dithering implements twice in this case? Should I use some other videofilter with MadVR instead of LAVFilter to avoid increased dithering artifacts and resource wasting?
As for dithering mode "none" in MadVR, it is said in it's description, however, that rounding implements in this case. So, that is not totally nothing very likely and is undesirable by the same reason as well. Is it right?
markanini
5th April 2016, 20:57
Sorry that I haven't had a chance to look at any of the new anti-ringing options at all yet.
Thanks for the chroma scaling option to restore the old behavior, chroma quality is noticeably improved now - though it obviously comes at a performance cost.
There seems to be a bug with linear light downscaling and mixed x/y scaling.
I'll try to upload a sample this evening which demonstrates the problem.
What's happening is that, with luma x downscaling and luma y upscaling, activating linear light downscaling results in a very dark image.
If NNEDI3 luma doubling is activated under these conditions (no chroma) the image turns green.
This is unaffected by the new chroma quality option.
EDIT: It's the combination of sigmoidal light upscaling, and linear light downscaling that seems to cause this.
Disabling either one fixes it.
Well done on finding the cause of the dark image bug. It was driving me crazy!
Q-the-STORM
5th April 2016, 20:58
I use LAVFilter to decode video and it has it's own dithering options:http://forum.doom9.org/attachment.php?attachmentid=15342&stc=1&d=1459885097So, as it seen, it has not "no dithering" mode. On the other side, MadVR has not "no dithering" mode too. Does it mean that dithering implements twice in this case? Should I use some other videofilter with MadVR instead of LAVFilter to avoid increased dithering artifacts and resource wasting?
LAV doesn't dither if you use madVR...
unless you do YUV -> RGB conversion... so leave that on untouched, then only madVR is dithering..
mysterix
5th April 2016, 21:32
LAV doesn't dither if you use madVR...
unless you do YUV -> RGB conversion... so leave that on untouched, then only madVR is dithering..
Thank You!
Is it right that MadVR does by it's dithering, in general, same correction of color space as videofilter does by YUV -> RGB conversion and movies will be look correctly on simple PC monitor if leave that option on "untouched"? Or, maybe, should I do something else to watch videos correctly in this option mode?
CarlosCaco
6th April 2016, 00:46
Someone could tell me about the option present a frame every v sync? I made some tests but i really don t knoe if help or not... this option low a little me tv refresh rate, sometimes seens to smooth sometimes seens to crap... when should i use this? Why d3d9 don t have the same? If i try reclock should i use that or not?
Asmodian
6th April 2016, 01:43
Someone could tell me about the option present a frame every v sync? I made some tests but i really don t knoe if help or not... this option low a little me tv refresh rate, sometimes seens to smooth sometimes seens to crap... when should i use this? Why d3d9 don t have the same? If i try reclock should i use that or not?
This option influences how madVR interfaces to D3D11. With it off madVR will only give D3D11 a new frame if it has a new frame, relying on D3D11 to duplicate frames for each v-sync. With it on madVR will give D3D11 each duplicate as a new frame for every v-sync, this takes more performance but also might prevent issues if the driver has trouble duplicating frames or something.
Ideally it would be better if madVR worked perfectly with it disabled but the only downside is a small performance hit. It is a good option if it prevents presentation glitches but if you don't have presentation glitches I wouldn't worry about it at all. Reclock is not relevant.
CarlosCaco
6th April 2016, 03:51
This option influences how madVR interfaces to D3D11. With it off madVR will only give D3D11 a new frame if it has a new frame, relying on D3D11 to duplicate frames for each v-sync. With it on madVR will give D3D11 each duplicate as a new frame for every v-sync, this takes more performance but also might prevent issues if the driver has trouble duplicating frames or something.
Ideally it would be better if madVR worked perfectly with it disabled but the only downside is a small performance hit. It is a good option if it prevents presentation glitches but if you don't have presentation glitches I wouldn't worry about it at all. Reclock is not relevant.
thanks for the explanation, this option alters my tv behavior, i donīt know why the refresh rates decrease a little, also alters the dropped/repeated frames every...
and with reclock alters the the clock adaptations compared to d3d9 and d3d11 without this...
do you know how d3d9 handle the frames and v sync?
Asmodian
6th April 2016, 07:58
thanks for the explanation, this option alters my tv behavior, i donīt know why the refresh rates decrease a little, also alters the dropped/repeated frames every...
and with reclock alters the the clock adaptations compared to d3d9 and d3d11 without this...
do you know how d3d9 handle the frames and v sync?
How odd, I would leave it checked then.
I am pretty sure D3D9 always presents a new frame for every v-sync.
Asmodian
6th April 2016, 08:08
It's a real shame 'move subtitles' breaks with italic text - one line with just a "<i>" is enough to break all lines for the remainder of the movie. You can't reset the setting in madVR and have it back to normal, the only way is to restart the movie and skip to previous position every time italic text shows up.
Moving the subtitles into the active video area is one of my favorite features in madVR. Is there an another way of doing this, without having to manually enter new positions for every file you want to play?
It would be great with simple profiles, i.e. something like "if croppedAR = 1.35 -> move subtitles to x.y" but I don't think that's possible.
Breaks? Do you have the newest xySubFilter? What format subtitles?
I usually disable that option but I don't think this is expected to be broken. :confused:
Neo-XP
6th April 2016, 09:42
Breaks? Do you have the newest xySubFilter? What format subtitles?
I usually disable that option but I don't think this is expected to be broken. :confused:
Yes it is unfortunately : http://forum.doom9.org/showthread.php?p=1761566#post1761566
Telion
6th April 2016, 18:05
There is another way to workaround this issue - you can just mux an external .srt subtitle file into a single .mkv with the video, as the bug with <i> only affects external .srt but not internal ones.
Sunspark
7th April 2016, 04:37
Hi, I just wanted to leave feedback on a system that isn't high-end but still fairly decent. I'm using an Intel Broadwell HD Graphics 6000 GPU with an older version of MadVR .89.18 using the D3D 9 path, LAV 0.68 on an Intel NUC5i5RYK running Windows 7.
Since I cannot use the high-end scaling algorithms, I focused on the ones that I could. I'm sensitive to things like PWM, dithering and see momentary distortion--if I am watching something in motion then look away at a still object. So I am picky to say the least.
There was too much talk in the past about people liking Jinc AR/Jinc AR for both image and chroma upscaling and I think they are wrong for just selecting it as automatically the best.
There's something about Jinc that I find has this sort of blurry oily look to it. It's subtle, but it is there and I can't help noticing it.
After I tested awhile with chroma upscaling on video, most easily done with video that you are not image upscaling or downscaling, I think Mitchell-Netravali with the anti-ringing filter checked is a good one to use for chroma upscaling. It seems to have that right mix of clarity with just a hint of smoothing which I think also comes in handy with lower resolution sources or bitrates.
People's faces are a good thing to focus on, not as a still, but while in motion. Skin pores too visible? Then the algo is too sharp. Skin pores or other stuff too smeared pushing them slightly into the uncanny valley direction? Then the algo is too smooth. Same thing with backgrounds.. dancing pixels vs smears.
Certainly it has been eclipsed by the new high-end stuff, but if you're not on a high end system like mine I urge you to consider Mitchell-Netravali for chroma upscaling media, including anime, not solely video.
For image upscaling, I tend to be lazy and just use DXVA for both chroma and image. If you don't want to do that, then Lanczos 3 Anti-Ringing is a fine choice.
You don't want it too sharp or too smooth. Lanczos 4 is going to be smoother than 3, which you don't need if you're already using Mitchell-Netravali for chroma which is already leaning on the smooth side. Adding both will be too smooth I find. 3 is a better choice than 4.
I never used Spline much at all. I found it was too similar to Lanczos but different in a Coke vs Pepsi sense.
Everyone's setup is different.
A link I will throw in here that I just stumbled across,
http://faculty.cs.tamu.edu/schaefer/research/supplemental/scanline/index.html there is a paper too if you shorten the URL. The link shows the effect of filters on SVG rasterization. Maybe it has been seen here before, but new to me and interesting to play with, the patterns are designed to expose aliasing problems.
Sean
7th April 2016, 05:11
If I were to advice someone something like Panasonic ST/VT/ZT series, then of course I would suggest to select 8bit in NVidia CP and 8bit in madVR :) .
I have a 60" ST60 and I'm still trying to figure out the best balance of options for it (I watch about a 50/50 split of 720p regular TV and Anime). Does 10bit+ not help even though there's an option in the Nvidia CP? I haven't really messed around with that yet.
ryrynz
7th April 2016, 08:41
I have a 60" ST60 and I'm still trying to figure out the best balance of options for it (I watch about a 50/50 split of 720p regular TV and Anime). Does 10bit+ not help even though there's an option in the Nvidia CP? I haven't really messed around with that yet.
Even feet away from my VT50 I can't see much of a difference, so you won't notice any difference between 8 and 10 bit on these panels at the usual viewing distance at all.
CarlosCaco
7th April 2016, 13:42
You don't want it too sharp or too smooth. Lanczos 4 is going to be smoother than 3, which you don't need if you're already using Mitchell-Netravali for chroma which is already leaning on the smooth side. Adding both will be too smooth I find. 3 is a better choice
Lanczos 4 seens sharper than 3 to me... BTW someone know if the same algorithm produces diferents results depending on the gpu that used (excluding dxva)
Like lanczos 4 rendered by nvidia 930m appears sharper than rendered by Intel hd 5500?
videonerd
7th April 2016, 15:52
Hi madshi is DontRenderAfterStop still applicable in the latest build?
huhn
7th April 2016, 16:39
Lanczos 4 seens sharper than 3 to me... BTW someone know if the same algorithm produces diferents results depending on the gpu that used (excluding dxva)
Like lanczos 4 rendered by nvidia 930m appears sharper than rendered by Intel hd 5500?
of cause lanczos 4 is sharper. more tabs sinc based algorithm = sharper image.
that doesn't mean lanczos4 is a better choice.
the math used on the image should have the same results. if it is different something else is the cause.
XMonarchY
7th April 2016, 17:59
I have a 60" ST60 and I'm still trying to figure out the best balance of options for it (I watch about a 50/50 split of 720p regular TV and Anime). Does 10bit+ not help even though there's an option in the Nvidia CP? I haven't really messed around with that yet.
AFAIK, ST60 specifically shows banding on grayscale gradient when 10bit/12bit option is used. I would stick with 8bit. On VA/SPVA/MVA panels, 12bit option helps, but Plasma uses dithering by default + you get madVR dithering, so its best to stick to 8bit.
XMonarchY
7th April 2016, 18:05
Does anyone know if using Anti-Ringing for SuperRes (Strength 3, 2x Upscaling Step) in Image Refinement improves visual quality?
Warner306
7th April 2016, 18:41
Does anyone know if using Anti-Ringing for SuperRes (Strength 3, 2x Upscaling Step) in Image Refinement improves visual quality?
Not likely. At least not in test images.
perlita.pajares
7th April 2016, 20:25
Sounds like a driver issue, but it's hard to say. Have you tried different GPU drivers for both Intel and NVidia GPUs? You could also try different media players, just to be safe. In any case, I think this is unlikely to madVR's fault, although I can't really say for sure.
I have tested with differents drivers and players and same response. If it is not madvr faults, it must be lav filters'.
Anyone with same problem?
Thanks
Sunspark
7th April 2016, 23:47
Found a performance bug in my setup with a fix. I was wondering why the rendering time on chroma was higher than it used to be in the past. It might be related to LAV 0.68.
Some numbers using a high resolution video playing in chroma upscaling mode only.
In LAV Video settings-
DXVA2 Copyback 1.6 ms
DXVA2 Native 2.6 ms
Yes, native was 1 ms slower than copyback.. which makes no sense to me, but that's what the HUD was saying.
This is with the following checked ON in madvr options:
use a separate device for presentation
use a separate device for dxva processing
Now, if you uncheck these options the rendering time in copyback jumps to 12+ ms.
I have no idea what interaction is at play here but the rendering time difference is huge with those two checked and unchecked.
Magik Mark
8th April 2016, 00:03
Is it possible for madvr to do the ff:
1. If video frame rate is 24 fps then switch to display refresh rate to 48hz
2. If video frame rate is 25 fps then switch to display refresh rate to 50hz
Thanks
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.