madVR - high quality video renderer (GPU assisted)

[Clammerz]

Originally Posted by brazen1

Isn't 'smooth motion' a last resort setting and a huge memory hog?

To summarise some of the responses you may/have receive(d):

• Some users don't like the delay when mode switching between different refresh rates
• Some users cannot achieve (or want to put in the effort to achieve) perfect refresh rate matching
• A lot of displays have "quirks" when using 23/24p refreshrates (downsampled chroma, cannot disable TV processing, internal judder in TV processing pipeline)

Regarding the latter, you'll also find a lot of users are willing to accept the quirks of the TV for temporal resolution/display BFI/etc.

Originally Posted by mclingo

Hi, are there any tricks to searching on this forum

Try Google:

inurl:"https://forum.doom9.org/showthread.php?t=146228" "3D"

Originally Posted by maiden

Trying to get my movies to stop being jerky.

Originally Posted by maiden

new pic playing through remote desktop

You'll need to get a screenshot without RDP, and it may be useful to get a general idea if there's a trend in what the numbers are showing (e.g. presentation glitch number rises, or XYZ queue drops every Y often over a period of time where you see the issue occurring multiple times).

Originally Posted by mclingo

does anything one 90% is too high for NGU sharp medium on an RX 550?

There have been reports of people getting frame drops when pushing their GPUs to 90+% usage. One thing to keep in mind that since madVR uses multiple components of a GPU, the core GPU utilisation may not be a true indicator of what components are at their max utilisation. Testing and checking frame rendering times and frame drops is the only way to know for sure if you are over the max capabilities for your card.

Originally Posted by psyside

Well i tried everything, and there is basically 0, not 5% just 0 difference in IQ

At first glance there's a couple of differences that stand out to me, but I'm not really one to pixel peep. If you can't see a difference then there's nothing stopping you from using VMR9. No one is forcing you to use madVR, but at the same time, you likely won't find someone to hold your hand and tell you exactly what you should set.
You've seen a few suggestions of what may improve and what people prefer but it's up to you to try to find out what looks best to you, since only you have your eyes.

[austinminton]

Originally Posted by Manni

You can use profiles with MadVR and only have FSE on with 3D. That's what I do.

Thanks for the suggestion. That worked for me. Guess its still a bug in fse, but we are all moving away from it in any case.

On a side note, I just realized "fullscreen" as a boolean value in profile rules does not work even though its listed as such on the post on profile rules.

[psyside]

Originally Posted by Clammerz

At first glance there's a couple of differences that stand out to me, but I'm not really one to pixel peep. If you can't see a difference then there's nothing stopping you from using VMR9. No one is forcing you to use madVR, but at the same time, you likely won't find someone to hold your hand and tell you exactly what you should set.
You've seen a few suggestions of what may improve and what people prefer but it's up to you to try to find out what looks best to you, since only you have your eyes.

Hi, i must first say that im sorry, since members here where very helpful, i'm not here to bash madVR or anything, i'm just trying, to figure out whats wrong in my case. I personally think the program i use to COMPARE the settings, is somehow overriding, and using madVR same as MPHC, since the pictures look identical to me, and few of my friends i asked, i never told them anything, i just ask them what they think about the pics.

With the help of Asmodian i managed to get madVR run much much better, but the issue is i still cannot see a difference, i zoomed 500-600% and used frame by frame method to get to exact same point of the video, no difference.

Usually you CAN very easy spot madVR difference, in videos i saw on youtube, and on tons of screenshots, the difference is huge, but im my case i see nothing even i use screenshots vs videos method, and we both know screeshots method is quite easier to spot.

What could be the cause? i have no idea really.

[nsnhd]

Originally Posted by psyside

What could be the cause? i have no idea really.

Have you tried other players ?
Personally, I prefer PotPlayer since it gives me no dropped frames while MPC-HC/BE does with the same madVR settings. Sometime you can't find explanations, just keep trying

[psyside]

Originally Posted by nsnhd

Have you tried other players ?
Personally, I prefer PotPlayer since it gives me no dropped frames while MPC-HC/BE does with the same madVR settings. Sometime you can't find explanations, just keep trying

MPHC works fine, i just cant find a good player with different renderer to use it in order to take screenshot from to compare it with madVR enabled MPHC.

Edit, does this looks normal?



deinterlacing and limited range?

[Asmodian]

No, your render queue is empty, it should be 7-8 / 8 or maybe 6-8 / 8. In a similar fashion the present queue should be at least 2-4 / 4. The rendering time looks OK and even if it was paused the render queue should be full (present queue would be 1-1 / 4).

Are you using DXVA2 native decoding in LAV filters? This is known to provide poor quality on Nvidia GPUs, use DXVA2 copy-back instead. Still you might not notice, the issue is only a slight blurring of the chroma planes.

Your source may be soft enough that there aren't any visible differences, even with SSIM compared to Bicubic. SSIM is good at preserving single pixel scale detail and fine texture but if you don't have any in the source you cannot preserve anything.

Don't use DXVA2 native! Especially on Nvidia GPUs, but you don't want to use native decoding at all. Use DXVA2 (copy-back) instead.

The OSD doesn't show a decoding speed issue, the decoder queue would be empty if that was the issue, and the (DXVA2) indicates hardware decoding is active.

When using native you cannot pick the hardware device to use (it has to be the same device the display is plugged into) and you are not playing anything right now, which is why the Active Decoder is inactive. Look at that same screen while playing something.

[varekai]

Originally Posted by Asmodian

Don't use DXVA2 native! Especially on Nvidia GPUs, but you don't want to use native decoding at all. Use DXVA2 (copy-back) instead.

Short question:

Nvidia GTX 960 GPU
My settings
LAV Filters
Hardware Acceleration
Hardware Decoder to use --> NVIDIA CUVID <--

Should I switch to --> DXVA2 (copy-back) <--

[Asmodian]

Yes, absolutely. That or D3D11 with the GPU selected:

[Siso]

Originally Posted by Asmodian

Yes, absolutely. That or D3D11 with the GPU selected:

Is the device selection important for dxva2 copyback, or is it only for d3d11?

[varekai]

Originally Posted by Asmodian

Yes, absolutely. That or D3D11 with the GPU selected:

Thanks for your input, appreciate it!

Hardware Acceleration
Hardware Decoder to use --> D3D11 <-- (it's not in there)

DXVA2 (copy-back) (OK!)
Hardware Device to use: Nvidia Geforce GTX 960 (OK!)

Also noticed you are using LAV Video Decoder 0.70.2.80-git
Should I upgrade LAV Filters?

[nevcairiel]

Originally Posted by Siso

Is the device selection important for dxva2 copyback, or is it only for d3d11?

You can use it to use a secondary GPU with DXVA2, but its not absolutely required.

With D3D11 using "Auto" mode uses D3D11 Native mode if available, selecting a specific GPU always forces Copy-Back.

[Siso]

Originally Posted by nevcairiel

You can use it to use a secondary GPU with DXVA2, but its not absolutely required.

With D3D11 using "Auto" mode uses D3D11 Native mode if available, selecting a specific GPU always forces Copy-Back.

So "automatic" works for win 7?

[Asmodian]

Originally Posted by varekai

Also noticed you are using LAV Video Decoder 0.70.2.80-git
Should I upgrade LAV Filters?

If you do the D3D11 option will appear.

[varekai]

Originally Posted by Asmodian

If you do the D3D11 option will appear.

I looked in there and can only find the 0.70.2 which is what I have installed.

Also noticed this:
Active Decoder: <inactive>
Active Hardware Accelerator to use: <none>

I got 3 displays 2 monitors 1 TV so
Hardware Device to use: Automatic
would be fine?

[Jesco]

You might want to look here for the nightly builds:

https://files.1f0.de/lavf/nightly/

[Asmodian]

Ah, maybe you need a nightly build? I upgraded to 0.70.2.83-git.

I don't think madVR cares if LAV uses DXVA2 copy-back or D3D11 cb, they both work very well. The big improvement with D3D11 is during native decoding, but you should only use that if you are on a battery.

Originally Posted by varekai

Also noticed this:
Active Decoder: <inactive>
Active Hardware Accelerator to use: <none>

This is because you aren't currently playing something, those show was is being actively used. LAV will fall back to software if the video isn't compatible with hardware decoding.

Originally Posted by varekai

I got 3 displays 2 monitors 1 TV so
Hardware Device to use: Automatic
would be fine?

If you use D3D11 than, no, you need to set it to the GPU to get copy-back.

[varekai]

Originally Posted by Jesco

You might want to look here for the nightly builds:

Thanks for the link!
So installed LAVFilters-0.70.2-83 and found out that D3D11 is not supported on my Win7... bummer...:\

[Jesco]

Sorry for this (probably noob) question. I have an AMD HD7970, running W10 OS.

Current LAV settings:

Hardware Decoder to use: DXVA2 (copy-back)
Hardware Device to use: AMD Radeon R9 200 series

Will I benefit any by switching to a nightly LAV version and use D3D11 as hardware decoder instead of DXVA2 (copy-back)? Will it take some load of my cpu?

[cork_OS]

I'm using CUVID instead of DXVA2cb due to Adaptive HW Deinterlacing.

[nevcairiel]

Originally Posted by Siso

So "automatic" works for win 7?

D3D11 decoding is not supported on Windows 7. It requires Windows 8+.