madVR - high quality video renderer (GPU assisted)

[Alexkral]

Any solution?

There is an issue in the c8 with some firmware not detecting the wide color gamut in PC Mode, it is supposedly corrected with firmware update (04.10.06).

[thighhighs]

Anybody else tested 1903 and encountered some issues?
After the update, problems started with my 720p sources. Playback is not smooth, but the OSD shows no problems :scared: For example, 1080p sources work fine, which is a bit strange for me.
Since it seems that the new madVR build will not be soon and MS focused on more important features like blur, I used the magic button and rolled back to 1803. And everything works fine again :)
My current setup:
R7 2700, GTX 1660 Ti, 430.64, PotPlayer x64.

[Daicon]

What version of AMD drivers and Windows 10?
Lastets both.


There is an issue in the c8 with some firmware not detecting the wide color gamut in PC Mode, it is supposedly corrected with firmware update (04.10.06).

I have firmware .15
And i try mode hdmi and pc and blueray, same result

[DMU]

Anybody else tested 1903 and encountered some issues?
There is an assumption that in the 1903 version FSE does not work any more. It seems to work, but like FSW.

[ryrynz]

FSE isn't working for me (1903 + latest Nvidia) but IDC 'cos I don't use it anyway.

[suanm]

Does anyone know why the two optional pulldown lists named '①color tweaks for fire and explosions; ②highlight recovery strength' go grey and they don't work when i click the button 'tone map HDR using pixel shaders'?
I don't know whats going on?What should i do to activate the two optional pulldown lists and make them work normally?

[suanm]

MadVR 0.92.17 is running on windows10 1803 OS and nvidia 385.28 display driver

[el Filou]

@suanm: uncheck "compromise on HDR tone & gamut mapping accuracy" in rendering > trade quality for performance

[Alexkral]

@Daicon

Save this as *.hlsl and load it in MPC-HC. It just convert the gamut to Rec.709 or DCI-P3. You can edit it as a text file, change the gamut option from the second line to DCI-P3 if you get oversaturation.

sampler s0 : register(s0);

static int GAMUT = 0; //Rec.709 = 0, DCI-P3 = 1

static float m1 = 2610.0 / 4096.0 / 4;
static float m2 = 2523.0 / 4096.0 * 128;
static float c1 = 3424.0 / 4096.0;
static float c2 = 2413.0 / 4096.0 * 32;
static float c3 = 2392.0 / 4096.0 * 32;

static float3x3 Rec709 =
{
1.660491, -0.5876411, -0.0728499,
-0.1245505, 1.1328999, -0.0083494,
-0.0181508, -0.1005789, 1.1187297
};

static float3x3 DCIP3 =
{
1.3435783, -0.2821797, -0.0613986,
-0.0652975, 1.0757879, -0.0104905,
0.0028218, -0.0195985, 1.0167767
};

float3 PQ_EOTF(float3 Ep)
{
float3 p = pow(Ep, 1 / m2);
return pow(max(p - c1, 0) / (c2 - c3 * p), 1 / m1);
}

float3 PQ_OETF(float3 Fd)
{
float3 p = pow(Fd, m1);
return pow((c1 + c2 * p) / (1 + c3 * p), m2);
}

float4 main(float2 tex : TEXCOORD0) : COLOR
{
float3x3 mat;
float3 color = tex2D(s0, tex).rgb;
color = PQ_EOTF(color);
mat = GAMUT == 0 ? Rec709 : DCIP3;
color = saturate(mul(mat, color));
color = PQ_OETF(color);
return float4(color, 1);
}

[anipeg]

Does madvr can separate the image scaling and rendering?

When use madvr with BlueskyFRC, it consume too much resource because madvr need upscale image to 4K in 60fps.
LAV decode(2K 24fps) -> BFRC(2K 24fps to 2K 60fps) -> madvr(2K 60fps to 4K 60fps) -> display

If separate the scaling and rendering, it should save 50% GPU resource because madvr only do image upscale in 24fps, not 60fps.
LAV decode(2K 24fps) -> madvr scaler(2K 24fps to 4K 24fps) -> BFRC(4K 24fps to 4K 60fps) -> madvr render(4K 60fps) -> display

[Alexkral]

Does madvr can separate the image scaling and rendering?

No that's not possible, it happens the same to me with SVP.

[Daicon]

@Daicon

Save this as *.hlsl and load it in MPC-HC. It just convert the gamut to Rec.709 or DCI-P3. You can edit it as a text file, change the gamut option from the second line to DCI-P3 if you get oversaturation.

sampler s0 : register(s0);

static int GAMUT = 0; //Rec.709 = 0, DCI-P3 = 1

static float m1 = 2610.0 / 4096.0 / 4;
static float m2 = 2523.0 / 4096.0 * 128;
static float c1 = 3424.0 / 4096.0;
static float c2 = 2413.0 / 4096.0 * 32;
static float c3 = 2392.0 / 4096.0 * 32;

static float3x3 Rec709 =
{
1.660491, -0.5876411, -0.0728499,
-0.1245505, 1.1328999, -0.0083494,
-0.0181508, -0.1005789, 1.1187297
};

static float3x3 DCIP3 =
{
1.3435783, -0.2821797, -0.0613986,
-0.0652975, 1.0757879, -0.0104905,
0.0028218, -0.0195985, 1.0167767
};

float3 PQ_EOTF(float3 Ep)
{
float3 p = pow(Ep, 1 / m2);
return pow(max(p - c1, 0) / (c2 - c3 * p), 1 / m1);
}

float3 PQ_OETF(float3 Fd)
{
float3 p = pow(Fd, m1);
return pow((c1 + c2 * p) / (1 + c3 * p), m2);
}

float4 main(float2 tex : TEXCOORD0) : COLOR
{
float3x3 mat;
float3 color = tex2D(s0, tex).rgb;
color = PQ_EOTF(color);
mat = GAMUT == 0 ? Rec709 : DCIP3;
color = saturate(mul(mat, color));
color = PQ_OETF(color);
return float4(color, 1);
}

WOW, now yes!
I needed to install DirectX

What was happening?
I do not know what this does!
Is it necessary then?
Why have not I seen it in several manuals that I found?

MANY THANKS
I would like to know a little more what this is, and why it is necessary.

a detail:
The image is very very similar now yes. BUT, the color is still somewhat different, I only noticed it in X colors, not all of them. For example I noticed little little little difference in the face (in the image before upload)
But practically perfect

Any other suggestions / recommendations?

thankssssssss

[Alexkral]

I think maybe your TV has some problem to select the right color gamut. With this shader you have the option to select it by yourself, but the result still depends on the gamut that the TV is using.

In MPC-HC you can see the changes you make to the shader in real time.

[Asmodian]

The image is very very similar now yes. BUT, the color is still somewhat different, I only noticed it in X colors, not all of them. For example I noticed little little little difference in the face (in the image before upload)
But practically perfect

HDR tone mapping is not an exact science. Do not assume your results with PLEX are exactly correct, you will not match it perfectly using another tone mapping technique and neither is necessarily more correct, just different.

[Warner306]

I think maybe your TV has some problem to select the right color gamut. With this shader you have the option to select it by yourself, but the result still depends on the gamut that the TV is using.

In MPC-HC you can see the changes you make to the shader in real time.

The AMD video drivers can sometimes change the color space. It has happened several times in the past. It must get confused with the input flag placed on the gamut by madVR...It is unlikely to be the TV because many others are using the same model with different GPUs.

[el Filou]

If separate the scaling and rendering, it should save 50% GPU resource because madvr only do image upscale in 24fps, not 60fps.
LAV decode(2K 24fps) -> madvr scaler(2K 24fps to 4K 24fps) -> BFRC(4K 24fps to 4K 60fps) -> madvr render(4K 60fps) -> displayBut then FRC would need to process a 4K image instead of a 2K one, so madVR would have 2,5 x less work but FRC 4 x more work. Which takes more GPU time?

[Alexkral]

But then FRC would need to process a 4K image instead of a 2K one, so madVR would have 2,5 x less work but FRC 4 x more work. Which takes more GPU time?

Hey, that's a good point, I had not noticed until now. I guess it could still do the calculations to the 2K image and interpolate the results to the upscaled 4K image, but obviously this is not available now.

[suanm]

@el Filou : Thank you very much.You are right.I do as you said above,the two pulldown listd work perfectly.thank you again.

[bigboyman]

Hello everyone. Recently I've been having some trouble getting decent performance out of madVR.
Even with all recommended "Trade quality for performance" options selected and fairly tame filtering I still get low performance (~15 fps when I should be getting at the very least 24, i.e, 10 frame drops per second).

Are my settings too high for my hardware?
Is there some hidden config I forgot to check?
Is it the drivers or the display settings?

My bulid: ~2-3 year old ASUS gaming laptop with

MPC-HC with LAV, madVR (https://www.videohelp.com/software/madVR), CCCP (https://www.videohelp.com/software/Combined-Community-Codec-Pack), XYsubfilter (https://www.videohelp.com/software/XySubFilter) and ASSmod (because they don't always both work)
i7-6700HQ @ 2.60 GHz
GTX 960 (4 GB)
16 GB DDR4 RAM
1TB HDD and 128GB SSD
Windows 10

Probably useful info:

The player uses up nearly all of my GPU's time (yes, the dedicated GPU) but not so much the CPU
Most of the GPU usage is for 3D (and a little bit for decoding)
I have Vsync enabled but it doesn't seem to improve performance at all
Integrated Graphics don't do anything (can't use both GPU's at once for the same thing)
I use madVR mostly for files with resolutions lower than 720p, so a decent chunk of the render time goes to upscaling
Render time is fairly low at the start of the video (20ms render time for like 3 minutes) but then quickly jumps to 70ms where it stays until I reboot my PC. This makes me believe it's not really an issue of my hardware not being good enough to use these settings with good performance.
The render queue is always nearly empty (1-2/24) and the decoder, subtitle and upload queues are always pretty much full (23-24/24)
The player should take at most 1/24 ~ 42ms for rendering each frame (else it drops frames like crazy [~10 per second])


Here's a link to my madVR and LAV settings (https://imgur.com/a/5HbY6iO).

Also if someone knows how to display more text in the OSD that would be very appreciated (so I can see all render steps and post what takes more time).

[huhn]

you are suing always super sampling don't do that. use 1.5 or 2.0 with your card i recommend 2.0.

can you make a screen of the OSD?

[ashlar42]

Here's a link to my madVR and LAV settings (https://imgur.com/a/5HbY6iO).
Is there a reason for having "Reduce compression artifacts" at 9? Such a high value could make sense just for really poor sources. I'd never use it on high resolution material, personally.

[bigboyman]

Tried using 2.0 instead of supersampling and it actually made it slightly worse (~70ms to ~80ms).

Here's the OSD with ShowRenderSteps (https://imgur.com/a/dQL6EwA) (as many as I can see)

[bigboyman]

Is there a reason for having "Reduce compression artifacts" at 9? Such a high value could make sense just for really poor sources. I'd never use it on high resolution material, personally.

I use this bulid mostly for files with resolution below 720p (often 480p) which is why it's so high.

[huhn]

your settings posted doesn't match the OSD and always supersampling or 2.0 have no effect in this case.

error diffusion is used which will melt the GPU. jinc scaling is used which will melt the GPU. most of the post processing filters are used which again will melt the GPU.

[bigboyman]

your settings posted doesn't match the OSD and always supersampling or 2.0 have no effect in this case.

error diffusion is used which will melt the GPU. jinc scaling is used which will melt the GPU. most of the post processing filters are used which again will melt the GPU.

What settings are different in the OSD?
And what alternatives do you suggest I use?

[huhn]

i don't even know where to start. deblocking should be used with NGU sharp for chroma to spare a bit of processing power but it alone is already alot of work for 960.
start from scratch.

[Asmodian]

Yes, for a 960 you will need to turn down or disable some settings, mostly you can only use a few pre/post processing filters. Ordered dithering is also great quality and much faster than error defusion.

If you really think it is something odd happening after three minutes do not look for issues in madVR, nothing changes in madVR after three minutes. That would be due to overheating or power saving options or similar. Is Optimus behaving badly again? That tech does not work very well in my experience. :mad:

Monitor GPU clocks while watching madVR, hwinfo is a good tool.

Edit: try running without any pre/post filters with show render steps so you can see all of them. Is anything obviously very slow?

[bigboyman]

Yes, for a 960 you will need to turn down or disable some settings, mostly you can only use a few pre/post processing filters. Ordered dithering is also great quality and much faster than error defusion.

If you really think it is something odd happening after three minutes do not look for issues in madVR, nothing changes in madVR after three minutes. That would be due to overheating or power saving options or similar. Is Optimus behaving badly again? That tech does not work very well in my experience. :mad:

Monitor GPU clocks while watching madVR, hwinfo is a good tool.

Edit: try running without any pre/post filters with show render steps so you can see all of them. Is anything obviously very slow?

Using fewer pre/post processing filters (image corrections, not upscaling/downscaling) does improve performance by about 10ms but it's also very important for good video quality at low-res meaning disabling kinda defeats the purpose of this whole setup.

Tried ordered dithering and it DOES improve performance without any visible quality loss but by only about 5ms so not a huge gain.

Clock is stable so not an issue but the GPU is hot (~85ºC - hot but not overheat hot) when madVR is working (with 70ms render time). But then again, that's probably due to the hardware limits and not so much Optimus drivers. I should also probably mention I don't have energy saving mode enabled.

Now that I can see the rest of ShowRenderSteps the composition of render time (~65ms because ordered dithering) is as follows (approximately):

Image corrections - 10ms
Deblock - 8ms
Chorma Upscaling - <1ms
Debanding - 2ms
Deringing - 7ms
NGU Doubling - 9ms
Jinc Upscaling - 10ms
Image scaling X+Y - 5.5ms
Subtitles - 7.5ms
Final Step - 3ms

What do you think I can afford to remove / compromise?

[bigboyman]

i don't even know where to start. deblocking should be used with NGU sharp for chroma to spare a bit of processing power but it alone is already alot of work for 960.
start from scratch.

Thanks for the advice. I don't notice any graphical difference (is there supposed to be one?) and can shave 5ms thanks to this.

So far, this thread has already made me lose about 10ms with next to no graphical compromise. Keep it up guys.

[brazen1]

bigboyman,

I've been using a GTX 960 4GB for years. I still do because I have no reason to upgrade until there is a reason. Every one of your settings is different from mine and I've been seeking the most performance I can obtain out of the 960. I'll go as far as telling you they are all completely wrong tbh. Even your LAV Filter settings. I use settings and profiles for every format there is because I have a very diverse collection. 3D, 2D, SDR, HDR, 2160p through 580p, etc. Imo, your best starting point, and perhaps ending point, is provided for you in the guide in my signature. You can d/l the madVR .bin file and replace yours in your madVR folder to make things very simple. It's a copy of what I use. There are also links to pictures with settings if you'd rather do it manually. Granted I am using a 4k display and I think you're using 1080p max. The settings still apply.

[huhn]

just show advanced rendertimes gives you higher rendertimes i never measured it but it's more then zero.

next problem is that you are currently scaling 480p the rendertimes with a 720p source for deblocking will sky rocket.

is this a 960m? what is the clock? 85 C is pretty high for a maxwell GPU.

subtitles should only take 1-3 MS on a 960 even at UHD.

when deblocking is used with a strong setting deringing doesn't do much anymore try to remove deringing if you still get ringing issues try a stronger deblocking/RRN setting.

upscale refinement is recommended over the use of image enhancement. try to replace the 4 top "image enhancer" with the way cheaper AdaptiveSharpen beware replacing image enhancement with upscaling cost more processing power and usually needs higher settings but Adaptive sharpening is so much cheaper and you are not creating as much upscaling artifacts. i can't remember a single use case where enhance details is useful for sources that are deblocked because a deblocking creates huge parts with no details.
the 4 top "enhancer" work best with clean/good source anyway.

using jinc after doubling has a high diminishing return lanczos 3 AR should be similar.

if you can't framerate match using smoothmotion is mandetory in my opinion.

NGU sharp should be massively better for chroma quality but in general it is not a chroma scaler i would use but it's free with deblocking unlike super XBR or NGU AA.

super XBR 100 is worth a shoot it should be faster then NGU AA mid but it looks clearly different.
you could also try always super sample for quadrupling (not for doubling) beware this can backfire spectacularly.

Using fewer pre/post processing filters (image corrections, not upscaling/downscaling) does improve performance by about 10ms but it's also very important for good video quality at low-res meaning disabling kinda defeats the purpose of this whole setup.
this is misguided using more doesn't mean it will result in "more".
see deringing as an example.

[tp4tissue]

where do i enable show render steps.

[huhn]

https://forum.doom9.org/showthread.php?p=1709584#post1709584
hidden options and there is a reason it is a hidden option.

[bigboyman]

bigboyman,

I've been using a GTX 960 4GB for years. I still do because I have no reason to upgrade until there is a reason. Every one of your settings is different from mine and I've been seeking the most performance I can obtain out of the 960. I'll go as far as telling you they are all completely wrong tbh. Even your LAV Filter settings. I use settings and profiles for every format there is because I have a very diverse collection. 3D, 2D, SDR, HDR, 2160p through 580p, etc. Imo, your best starting point, and perhaps ending point, is provided for you in the guide in my signature. You can d/l the madVR .bin file and replace yours in your madVR folder to make things very simple. It's a copy of what I use. There are also links to pictures with settings if you'd rather do it manually. Granted I am using a 4k display and I think you're using 1080p max. The settings still apply.

That seems pretty interesting and I'd definitely want to try it but I can't find any mention of the madVR .bin in the post in your signature.
Maybe because it's so long. If you could provide a direct link that would really help me out.

[bigboyman]

...

is this a 960m? what is the clock? 85 C is pretty high for a maxwell GPU.


As far as I've seen, 80ºC is the point at which the transition from 20ms to 60ms occurs. It doesn't really ever go above 85ºC.
Not sure what the significance of that is but it could be useful info.


subtitles should only take 1-3 MS on a 960 even at UHD.


Using the internal decoder instead of XYsubs or ASSmod doesn't make a difference.
Not sure what I could use here. What settings on madVR would reduce subtitle render time?


when deblocking is used with a strong setting deringing doesn't do much anymore try to remove deringing if you still get ringing issues try a stronger deblocking/RRN setting.


Do you mean to disable "Reduce ringing artifacts" or the anti-ringing filter?
Anyway, I disabled both. The difference is not insingificant - the ringing becomes a lot more obvious.
I wouldn't call it distracting, but definitely more noticeable.
Anyway, disabling "Reduce ringing artifacts" shaves about 8ms and the AR filter only about 1ms.
I chose to disable both since it's a huge performace gain (9-10ms) for only slightly annoying ringing and the AR filter barely improves ringing anyway.


upscale refinement is recommended over the use of image enhancement. try to replace the 4 top "image enhancer" with the way cheaper AdaptiveSharpen beware replacing image enhancement with upscaling cost more processing power and usually needs higher settings but Adaptive sharpening is so much cheaper and you are not creating as much upscaling artifacts. i can't remember a single use case where enhance details is useful for sources that are deblocked because a deblocking creates huge parts with no details.
the 4 top "enhancer" work best with clean/good source anyway.


Only saves like 2ms and looks significantly worse, so I'm not doing this one.


using jinc after doubling has a high diminishing return lanczos 3 AR should be similar.


Actually god-tier advice. Goes down by 9-10ms and looks the same.


if you can't framerate match using smoothmotion is mandetory in my opinion.


Don't think this is the problem not to mention my render time shoots up like crazy (30ms up).


NGU sharp should be massively better for chroma quality but in general it is not a chroma scaler i would use but it's free with deblocking unlike super XBR or NGU AA.
...

Already using it, thanks :)

------------------------------------------------------------------------------------------

So, just to recap: Right now, despite some slightly annoying ringing, I've managed to get render time ~42ms. This is pretty good, but I'd like it to be consistently lower since it's still to close to 1/24, which means there are still frequent (but massively less noticeable) frame drops.

EDIT: Also disabled "Disable scaling if image size ..." and bumped compression artifact reduction to 14 (still medium quality) since it improves image quality with about 0 added cost

[bigboyman]

One more thing I thought I should mention is that for some reason, using a 720p source bumps frame render time to around double what I have in 480p. This relationship seems to be linear with increasing resolutions. Should there really be such a pronounced increase in frame render time?

[Asmodian]

Yes, 960x720 has exactly twice the number of pixels that 720x480 does. This does indicate that most of your performance cost is due to chroma upscaling, artifact removal, and/or image enhancements (the things before upscaling). Upscaling is also slower with larger input sizes but it scales with output resolution as well, so it wouldn't be 2x slower.

As far as I've seen, 80ºC is the point at which the transition from 20ms to 60ms occurs. It doesn't really ever go above 85ºC.
Not sure what the significance of that is but it could be useful info.

This is the source of your performance issues.

Your laptop's cooling is not good enough so your GPU downclocks a lot to keep temps down. Is it too dusty or is a fan broken, or something? Have you checked clock speeds when this happens? You will need to keep temps at 80°C or below.

[bigboyman]

...
This is the source of your performance issues.

Your laptop's cooling is not good enough so your GPU downclocks a lot to keep temps down. Is it too dusty or is a fan broken, or something? Have you checked clock speeds when this happens? You will need to keep temps at 80°C or below.

I'm trying a laptop cooler but it doesn't do much.
It could be dusty but I'd have to open it and check.

Thing is, nVIDIA website says the max operating speed is around 100ºC.
Not to mention, even relatively inactive, my GPU still hovers around 80ºC, so it's not like madVR makes the temp shoot up like crazy. Is the throtelling threshold lower than advertised?

[huhn]

just give us GPU clock speed and we will know if it is throttling.

if you get 89 C at idle your system isn't cooled properly.

laptop usual are not using normal maxwell cards but tweaked far slower version with a similar name and it's unlike they will allow 100 C the GPU may survive that but that doesn't the rest around it.

as far as i know the throttling temp for maxwell was around 80 where it will at least not boost anymore.

[bigboyman]

just give us GPU clock speed and we will know if it is throttling.

if you get 89 C at idle your system isn't cooled properly.

laptop usual are not using normal maxwell cards but tweaked far slower version with a similar name and it's unlike they will allow 100 C the GPU may survive that but that doesn't the rest around it.

as far as i know the throttling temp for maxwell was around 80 where it will at least not boost anymore.

It goes from about 1200 - 900MHz to 405MHz the instant I open a file on the player (unless it's the first time it got to high temps since boot). Temperature doesn't even seem related to the clock change. It could be 83ºC before and 78ºC after and it still does this. It will sit for as long as it feels like it at that clock speed (even after closing the player) before going back.

[huhn]

that's how throttling works. as soon as the throttling temp is reach in your case ~80 it will lower the clock to avoid going over that temp.

the normal boost clock of a maxwell GPU is over 1200 mhz.
your system is "killing" itself right now. your system is throttling so hard i doubt the cooler is still working.

[Asmodian]

It is possible for the thermal paste in an old laptop to harden and "crack" off the GPU or cooler. This hurts cooling performance a lot and also shows that really fast throttling behavior because there is no thermal mass for the GPU to heat up before it starts throttling.

Not related to madVR, any 3D load like a game will have the same problem. :(

[ttnuagmada]

So i have a weird one that started happening maybe a month ago. Basically when I open a video with a player that is using MadVR, there is a chance that my computer will hard-lock. I cannot for the life of me figure it out. I have tried with both Potplayer and MPC-HC.

I'm running a 3770K and a 980ti. I had the CPU overclocked, but returning to stock seemed to have no effect.


There have been 2 situations where I've noticed it.


I'm opening a video - I never see a picture before it locks, it's always before there is a picture. This is without going fullscreen.
Changing out a 3Dlut.


I'm pretty solidly convinced that it is not a hardware/stability issue. I can game all day long without issues. I have stress tested the CPU with IBT and the GPU with 3dmark stress test and have had absolutely no issues with that or doing anything else. The PC is watercooled and my temps are fine. There are no WHEA errors or any other kind of hints in the event log either.

I have tried all of the different decoders in LAV. I have tried it without LAV installed at all. I have tried several different Nvidia driver versions, using DDU with each install.

I'm convinced it has something to do with MadVR, because I have tried to replicate it using the default renderer and it doesn't happen. It also seems to be completely random. Sometimes I'll have opened several videos before it happens, and then go several more before it happens again. Sometimes it will happen 5 times in a row.

The thing that gets me though is that I had made no changes to anything from before it started to when it started. I was running the same GPU driver, hadn't changed any MadVR settings or any other filter settings. Maybe there was a Windows 10 update that did it? I've been on 1809 for several months.

any help would be appreciated.

edit: also, most of my video is played over a network, so to rule out anything network/harddisk related I tried it with local videos as well and it will still happen.



Anyone have any ideas on this? I still cannot figure it out, and this is the only situation that anything like this happens. My computer is otherwise perfectly stable.

I've tried everything I can think of short of re-installing windows.

[huhn]

try debug mode and look into the log after you reboot. don't except anything from this.

[jkauff]

Anyone have any ideas on this? I still cannot figure it out, and this is the only situation that anything like this happens. My computer is otherwise perfectly stable.

I've tried everything I can think of short of re-installing windows.
Make a note of your madVR settings, and copy any scripts into a text file. Then run the "restore default settings.bat" file found in the madVR folder. Chances are good that the problem will be gone. Add your settings back one at a time until you get the problem again--if you do.

[ttnuagmada]

Make a note of your madVR settings, and copy any scripts into a text file. Then run the "restore default settings.bat" file found in the madVR folder. Chances are good that the problem will be gone. Add your settings back one at a time until you get the problem again--if you do.

I've uninstalled MadVR and reinstalled it, would that not accomplish the same thing?

[ryrynz]

No, there are registry settings also. Restore default settings.

[ttnuagmada]

No dice.

[clsid]

Use the debug version. Then you can see what madVR does right before the hang occurs.

Scan your harddrive for bad sectors.
Scan for damaged Windows files (https://www.ghacks.net/2018/03/16/use-dism-to-fix-issues-sfc-cant/).

[mclingo]

Anyone have any ideas on this? I still cannot figure it out, and this is the only situation that anything like this happens. My computer is otherwise perfectly stable.

I've tried everything I can think of short of re-installing windows.

i had a similar issue a couple of times before, maybe not exactly the same though.

before resorting to a full windows rebuild, have you chucked the full monty at it in one go, rather than doing everything bit by bit?, before you do a full rebuild try this all in one go.


Remove NVIDIA driver using DDU and reboot.

Remove MADVR completely

Remove Lav completely.

Reboot again

reinstall Nvidia driver.

Install MPC-BE and run up a movie just to check all is working ok at this point.

install LAV

Install MADVR in a !!NEW!! location

Switch MPC-BE render to MADVR and try it, leave MADVR at stock settings for now.