madVR - high quality video renderer (GPU assisted)

[leeperry]

when you're in a lazy mood "let mVR decide" seems actually quite convenient, even though it seems too soft to my taste.

[Oguignant]

Hi, today activate the "crop black bars" option. Save 6 ms. Does anyone know of any disadvantages for using it?

[Mistery 73]

a greeting to all, activate 200% super sampling you think it is helpful when upscaling to 1080p to 4K? what do you advise me to use as an algorithm after doubling when upscaling to 4K 1080p and use as a quality algorithm NGU very high to me like a very defined image sharp

[kalston]

http://imgur.com/a/JG6db

Heh, I got it on my hands now. Looks good to me.

There is actually a more recent master that I should try and get my hands on (and that you own). Regardless even with the old master NGU makes it look markedly better than NNEDI IMO. So far I prefer NGU with everything I have, basically. Even though a bit of softness is tempting sometimes.

[ttnuagmada]

Can someone give me an idea of where to start as far as what the best settings would be for a bluray remux on a 1080p screen (ST60)? System has a 980ti and 3770K@4.6ghz

[ryrynz]

Can someone give me an idea of where to start as far as what the best settings would be

There are no "best settings" what you're asking for is a recommendation and these will vary from person to person. You're not going to learn anything if you just copy and paste someone else's preferences.
There's no right or wrong settings with the exception of stupidly excessive chroma settings in comparison to the luma.

[Sunset1982]

Want to buy new videocard for madvr.
Perfect 23.976 is needed (both ati and nvidia cards i own have 23.978 and 23.971).
Dont like custom resoluion = prefer autochange refresh rate in madvr display modes.
Advise variants pls

Amd cards do 23,976 close to perfect. With my rx480 i get no framedrops for 4-5 hours. Performance is ok wirh ngu medium or high and i expect it to get even better with newer drivers. I got a xfx480 gtr because it is relative small and fits in my htpc. If you got a bigger case i would go for a red devil becaul its more silent.


I put my nvidia gtx 1060 into my gaming pc because of the custom resolution thing and the d3d11/ed crash bug.

[SpinB]

Fucking Hell, madshi you really nailed it this time, the combination of NGU + super-xbr is an outstanding improvement in image quality, +/- 23% of GPU use and the laptop temps and fans are not all over the place, 0.91.4 should be renamed version 1.0.

:thanks:

[Damien147]

Why the hell would I want to image double(using ngu-med) and downscale a 3840x1600 movie to 4096x2160!?:scared:I get dropped frames with my rx470 like that.It's the only ''bad'' occasion though between 7-8 4k movies I've tried.

[pose]

Why the hell would I want to image double(using ngu-med) and downscale a 3840x1600 movie to 4096x2160!?:scared:I get dropped frames with my rx470 like that.It's the only ''bad'' occasion though between 7-8 4k movies I've tried.
What the heck youre talking about? :confused: If NGU is too slow for your specific task then just dont use it.

[Asmodian]

Can someone give me an idea of where to start as far as what the best settings would be for a bluray remux on a 1080p screen (ST60)? System has a 980ti and 3770K@4.6ghz

I suggest NGU high for image upscaling with let madVR decide for everything else.

[Uoppi]

A while back I remember there was some discussion about Error Diffusion 1 being OK for 8-bit displays but NOT Error Diffusion 2. Was that info legit? Or did I misunderstand it?

I'm using ED2 currently and wondering if I should switch to ED1 or ordered dithering.

[har3inger]

Hi, today activate the "crop black bars" option. Save 6 ms. Does anyone know of any disadvantages for using it?

It's generally useless unless your source has hard-coded letterboxing. If it does, it removes those portions of the image to simulate a lower resolution source. This costs less GPU to upscale or downscale but costs more CPU to detect if there are black bars.

There's also a theoretical possibility madvr will crop out actual image data by erroneously thinking that black stuff at the edges is black bars, but madshi's done a great job of putting in rules that essentially prevent this from ever occurring.

A while back I remember there was some discussion about Error Diffusion 1 being OK for 8-bit displays but NOT Error Diffusion 2. Was that info legit? Or did I misunderstand it?

I'm using ED2 currently and wondering if I should switch to ED1 or ordered dithering.

I don't think this is the case. ED1 and ED2 are different styles of dithering which trade off either noise for dither patterns and vice versa. It depends a lot on what screen/tv you use to figure out which is better for you. Most of the time the differences between ED1 and ED2 (and even ordered dithering) are going to be invisible. You can set your display bitdepth to 2 or 3 in display properties to get a very obvious representation of how each algorithm looks.

[Stereodude]

A while back I remember there was some discussion about Error Diffusion 1 being OK for 8-bit displays but NOT Error Diffusion 2. Was that info legit? Or did I misunderstand it?

I'm using ED2 currently and wondering if I should switch to ED1 or ordered dithering.
ED is a waste on a 10-bit display if you're using 10-bit output from madVR. That's my recollection of the topic. Even on an 8-bit display using 8-bit output the differences are very subtle. Generally your GPU power is better used elsewhere.

[Asmodian]

I agree, ED 1 or 2 for 8-bit or less but only when you have extra GPU power, ordered dithering offers great quality and is quite fast.

[Damien147]

What the heck youre talking about? :confused: If NGU is too slow for your specific task then just dont use it.

I just like to set and forget.For everything else(sd/hd and few 4k I've tried) current settings work fine,it just happened.

[ryrynz]

I just like to set and forget.For everything else(sd/hd and few 4k I've tried) current settings work fine,it just happened.

You can do that generally once you have profiles set up.

[Sunset1982]

A new radeon crimson driver is announced for tomorrow, named "relive".

It will have a lot of new features and bug fixes. Let's hope it brings a performance boost for NGU with it. I will report back as soon as I got it and tested it.

https://youtu.be/-U9KnDGUtc0

[omarank]

Gotta know where to look and on what scenes, it's quite subtle most of the time until a whole lot of red comes into the frame.
Yes, I know. Do you have a good sample which can show differences between NNEDI3 16 and Jinc chroma doubling?

[clsid]

A new radeon crimson driver is announced for tomorrow, named "relive".

It will have a lot of new features and bug fixes. Let's hope it brings a performance boost for NGU with it. I will report back as soon as I got it and tested it.

https://youtu.be/-U9KnDGUtc0
It also has a new feature called "Chill", which reduces framerates (read: lowers GPU clocks) during static scenes. Similar to what NVIDIA does with their "Optimal" power mode setting. It would not surprise me if it causes performance issues with MadVR if it is turned on.

[burfadel]

Chill is only enabled, at least for now, in certain games. It won't be activated with madVR, but even if it were it wouldn't affect frame rate since it's constant for most videos.

[Asmodian]

Chill is only enabled, at least for now, in certain games. It won't be activated with madVR, but even if it were it wouldn't affect frame rate since it's constant for most videos.

It sounds like it would affect madVR if it was enabled for it. Some scenes in video are static and dropping the clock speed during them could cause madVR to start dropping frames. We might not notice though, since the frames should all look the same anyway.

[huhn]

nearly every video is CFR and it doesn't really matter what happens in the scene the needed processing doesn't change a lot so the clock shouldn't change even with a static scene.

[ryrynz]

Yes, I know. Do you have a good sample which can show differences between NNEDI3 16 and Jinc chroma doubling?

Sure.

Jinc AR vs NNEDI3 16 (http://screenshotcomparison.com/comparison/193040)
Super-xbr 100 AR vs NNEDI3 16 (http://screenshotcomparison.com/comparison/193047)

I've actually gone back to super-xbr 100 AR. 125 is a bit bright and not as natural for my liking.
If you'd like the video I can cut a sample. The above are upscaled 800%, differences are still easily discernible at 200% though.

Look at these two, easily the worst options in this scenario. Look at the bleed through on Recon and Bilateral is well... it speaks for itself really.
Recon Soft vs Bilateral (http://screenshotcomparison.com/comparison/193045)

Going back to version 0.88.8 I'm actually quite liking Jinc 4 AR as well. Perhaps that should be reinstated for chroma, GPU usage is similar.
Jinc AR vs Jinc 4 AR (http://screenshotcomparison.com/comparison/193046)

*Edit* Based on what I've seen with the other scalers though it's likely that Jinc 3 is more accurate, maybe this isn't a big loss after all.

IMO Super-xbr with AR really is the king of chroma right now in terms of accuracy and performance based on what I've seen so far.

*Edit* NGU also seems to be offering quite some bang for buck also, at least in this example.

[Asmodian]

Jinc 4 seems to have worse ringing with no benefit?

I do like Recon Soft with SR too, does SR help with the bleed through in this example?

edit: and where is NGU? I would love to see this example with NGU med as well. :)

[ryrynz]

Jinc 4 seems to have worse ringing with no benefit?

Thought it looked beneficial as a softer chroma option.


I do like Recon Soft with SR too, does SR help with the bleed through in this example?


Nah, that's just how reconstruction works.


edit: and where is NGU? I would love to see this example with NGU med as well. :)


Recon soft SR1 VS NGU med. (http://screenshotcomparison.com/comparison/193054)

Interesting comparison, Jinc 4 AR is a lot cleaner and provides the same colors as Bicubic 75 AR which was considered to be one of the better chroma choices some years ago.
Jinc 4 AR vs Bicubic 75 AR (http://screenshotcomparison.com/comparison/193055)

For anyone interested in doing their own testing with this, download here (http://s000.tinyupload.com/download.php?file_id=05007139025542992175&t=0500713902554299217567771).

I'd probably consider Jinc 4 AR for chroma if it made it's way back in, would want something better than my GTX 960 for it though.
*Edit* Have changed my mind, will probably go with NGU once it's been optimized for chroma.

[khanmein]

@madshi got chance for D3D12? https://github.com/baldurk/renderdoc/releases/tag/v0.32

[Asmodian]

Recon soft SR1 VS NGU med. (http://screenshotcomparison.com/comparison/193054)

Interesting comparison, Jinc 4 AR is a lot cleaner and provides the same colors as Bicubic 75 AR which was considered to be one of the better chroma choices some years ago.
Jinc 4 AR vs Bicubic 75 AR (http://screenshotcomparison.com/comparison/193055)

For anyone interested in doing their own testing with this, download here (http://s000.tinyupload.com/download.php?file_id=05007139025542992175&t=0500713902554299217567771).

Thanks a lot! :)

NGU med does very well at keeping the red out of the white.

Edit:
I decided to make some comparisons where I could reference the original RGB, it is so hard to decide how it should look. NGU veryHigh looks the best to me. :)
comparisons at 400%
Original RGB vs NGU medium (http://screenshotcomparison.com/comparison/193071)
superXBR 100 AR vs NGU medium (http://screenshotcomparison.com/comparison/193070)
NGU medium vs NGU veryHigh (http://screenshotcomparison.com/comparison/193073)
bicubic 60 AR vs jincAR (http://screenshotcomparison.com/comparison/193069)
NNEDI3 32 vs NGU medium (http://screenshotcomparison.com/comparison/193074)

original RGB image (http://www.mediafire.com/view/n33o1t0nj9imf38/BrownFox.png)
4:2:0 lossless H.264 mkv (http://www.mediafire.com/file/gtrhq95j4pqdbhp/BrownFox.mkv)

[omarank]

Sure.

Jinc AR vs NNEDI3 16 (http://screenshotcomparison.com/comparison/193040)
Super-xbr 100 AR vs NNEDI3 16 (http://screenshotcomparison.com/comparison/193047)

I've actually gone back to super-xbr 100 AR. 125 is a bit bright and not as natural for my liking.
If you'd like the video I can cut a sample. The above are upscaled 800%, differences are still easily discernible at 200% though.
Interesting. Thanks!

Going back to version 0.88.8 I'm actually quite liking Jinc 4 AR as well. Perhaps that should be reinstated for chroma, GPU usage is similar.
Jinc AR vs Jinc 4 AR (http://screenshotcomparison.com/comparison/193046)

Jinc 4 AR is a lot cleaner and provides the same colors as Bicubic 75 AR which was considered to be one of the better chroma choices some years ago.
Jinc 4 AR vs Bicubic 75 AR (http://screenshotcomparison.com/comparison/193055)
Agreed.

For anyone interested in doing their own testing with this, download here (http://s000.tinyupload.com/download.php?file_id=05007139025542992175&t=0500713902554299217567771).
Thanks for the sample!

[TheElix]

Thanks a lot! :)

NGU med does very well at keeping the red out of the white.

Edit:
I decided to make some comparisons where I could reference the original RGB, it is so hard to decide how it should look. NGU veryHigh looks the best to me. :)
comparisons at 400%
Original RGB vs NGU medium (http://screenshotcomparison.com/comparison/193071)
superXBR 100 AR vs NGU medium (http://screenshotcomparison.com/comparison/193070)
NGU medium vs NGU veryHigh (http://screenshotcomparison.com/comparison/193073)
bicubic 60 AR vs jincAR (http://screenshotcomparison.com/comparison/193069)
NNEDI3 32 vs NGU medium (http://screenshotcomparison.com/comparison/193074)

original RGB image (http://www.mediafire.com/view/n33o1t0nj9imf38/BrownFox.png)
4:2:0 lossless H.264 mkv (http://www.mediafire.com/file/gtrhq95j4pqdbhp/BrownFox.mkv)
Thanks for these comparisons! The difference is very clear. Seems like for quality chroma upscaling the priority should be like this: NGU Very High > NGU High > NGU Med > NGU Low (?) > Super-XBR 100 AR > followed by everything else.

[ryrynz]

I decided to make some comparisons where I could reference the original RGB, it is so hard to decide how it should look.

Really need some animated or real world content shot in 4:4:4..

It's easier to pick a winner when there's just one color like in above scenario but it may just be that more times than not NGU very high is the most accurate.
Some extra tests between the top contenders.

Super-xbr 100 AR vs NGU low (http://screenshotcomparison.com/comparison/193084)
NGU Low vs NGU med (http://screenshotcomparison.com/comparison/193085)
NGU med vs NGU high (http://screenshotcomparison.com/comparison/193086)
NGU high vs NGU very high (http://screenshotcomparison.com/comparison/193087)
NGU low vs NGU very high
(http://screenshotcomparison.com/comparison/193089)
Seems like for quality chroma upscaling the priority should be like this: NGU Very High > NGU High > NGU Med > NGU Low (?) > Super-XBR 100 AR > followed by everything else.

NNEDI3 32 is quite nice and IMO looks better than NGU low but it's considerably more demanding at least on this HD4000, it's more demanding than NGU very high which makes it hard to recommend.
One wonders if perhaps it's time for NNEDI3 to retire from chroma duties.. That asks for one more I guess..

NNEDI3 32 neurons vs NGU very high (http://screenshotcomparison.com/comparison/193091)

[Damien147]

You can do that generally once you have profiles set up.

I might check that if it happens again.

[StinDaWg]

The real world impact to any of those chroma comparisons is so close to nothing/placebo it's barely even worth discussing.

As madshi has said a million times, luma is way more noticeable/important.

[Werewolfy]

@madshi
I just wanted to say that NGU for 1080p --> 2160p is awesome! I used Jinc for upscaling because I found it more natural than NNEDI3 and I was limited to 32 neurons. But now I use NGU, it's very natural, sharp and I can use the high preset. The picture is so pristine with this upscaler, it's actually the first time that I can easily see the difference on almost every scenes. I don't know how you did it but you really did a great job with this upscaler, it totally beats NNEDI3 + SR with high quality sources.

The only downside is with 480p anime. To be fair, NGU is more faithful to the source than NNEDI3 but it REALLY doesn't like mosquito noise. I use NNEDI3 256 neurons + SR1 on this type of content and for the moment I will stick with it. NNEDI3 hides these kind of artefacts very well, the reason why I use only SR1 is because anything higher will show again these artefacs like NGU. This week-end I will try 720p anime to see what NGU does.

But I will say it again, NGU is unbeatable with high quality sources!

I will also test the new UI this week-end.

[LoKi128]

Why can't the chroma upscaling algorithm be used beyond the initial conversion to 4:4:4? For example with my settings, it'll upscale chroma initially with ReconSharp, then upscale luma with NGU-veryHigh and the chroma again but now with NGU-med. Couldnt ReconSharp (or any of the initial-chroma-only scalers) be used for the second scaling step as well?

Is it just not optimized for scenarios beyond 4:2:0 -> 4:4:4? However, wouldn't it just be the same scenario as the initial conversion, having a chroma image that is exactly half the luma image if a doubling algo is used, like NGU, NNEDI or SuperXBR?

[Q-the-STORM]

Why can't the chroma upscaling algorithm be used beyond the initial conversion to 4:4:4?

Reconstruction uses luma information to be more accurate than a regular upscaler...
so using it past original luma resolution, doesn't make much sense, since upscaled luma is just as accurate as upscaled chroma...

[HillieSan]

The new AMD Crimon Relive drivers do not boost NGU on my RX 480. Performance is the same as with the previous drivers. The drivers are buggy. I have to restart my receiver otherwise it does not find a sound device, At boot time I get collored noise for a few second before it goes to desktop.

[ShiftyFella]

I'm late to the party, jumped from 90.24 to 91.4, didn't read much of the previous discussions. My biggest gripe is why I can't have nice things? Why doubling\quadrupling options are so limited all of sudden when it comes to NNEDI3? http://i.imgur.com/tK8AUVs.png before I could easily use 256->128 for luma and 32->32 or 64->32 for chroma with no problem. I would like my options back, as 16 neurons for chroma sometimes produces glitches or color artifacts that 32 never has, so I would like to avoid 16 neurons and impact on performance of 32 vs 16 is negligible. Another thing is setting very high for chroma doubling becomes unavailable when you drop to 32 neurons for luma and you have to stay with 16 neurons or drop down to bicubic60, old options and behavior was better IMO

Played a bit with NGU high, I like how sharp it is but for low res\bitrate I still prefer NNEDI3 but need to do more testing to decide. I'm not so thrilled with NGU low, NNEDI3 at 32 or superxbr with SR looks better for all content. Performance vs NNEDI3 is amazing tho. On my 290x difference in performance of Superxbr vs NGU low is there, on your typical resolutions gap is not really noticeable but if you're like me, who likes to use SS with 4k content on 1440p screen, difference becomes more apparent that superxbr is faster and actually can be usable. I'm happy that superxbr is still around as low power\impact option, I miss XBR AB:'(

720p doubled to 1440p:
NGU high for chroma\luma is 11ms
NNEDI3 64 for luma and 32 for chroma is 20ms
Superxbr 5ms with SR@3 9ms

1080p SS doubled and scaled down with SSIM1D to 1440p:
NGU high\high is 25ms
NNEDI3 64\32 is 43ms and 27ms for NNEDI3 32\16
Superxbr 11ms with SR@3 20ms

SuperXBR vs NGU low:
1080p doubled -> 1440p
surepxbr: http://i.imgur.com/JKXyjQc.png
ngu low: http://i.imgur.com/0o3yBjc.png

2160p doubled -> 1440p
superxbr: http://i.imgur.com/E86Ow7D.png
ngu low: http://i.imgur.com/7Q8LOly.png

[Backflash]

The new AMD Crimson Relive drivers do not boost NGU on my RX 480. Performance is the same as with the previous drivers. The drivers are buggy. I have to restart my receiver otherwise it does not find a sound device, At boot time I get colored noise for a few second before it goes to desktop.

Are you running audio through HDMI? I'd get separate device for that. Have no noise on boot, it seems it doesn't like your receiver. Do you have same issue if you try to connect it directly to TV?

No changes to MadVR whatsoever on my end, at least with current things I have, I think NGU has slightly(1-3ms) lower render time for 10bit 1080p but it's been a while since I tried it extensively before new drivers, just launched to look at timings. Might be wrong there.
Primary change is purely for dx11/12 multi/single threaded performance(around +20% in synthetic).

[Damien147]

....yep,it happened again.I am using NGU med for everything but for the specific resolution (3840x1600) I have to use jinc to upscale to 4096x2160.I checked the guide (http://forum.doom9.org/showpost.php?p=1271417&postcount=3) but I failed.Can you help me what to type/do with the profiles?

[HillieSan]

Are you running audio through HDMI? I'd get separate device for that. Have no noise on boot, it seems it doesn't like your receiver. Do you have same issue if you try to connect it directly to TV?

No changes to MadVR whatsoever on my end, at least with current things I have, I think NGU has slightly(1-3ms) lower render time for 10bit 1080p but it's been a while since I tried it extensively before new drivers, just launched to look at timings. Might be wrong there.
Primary change is purely for dx11/12 multi/single threaded performance(around +20% in synthetic).

I had no issues with the previous drivers. I use a long cable (12 meters) to the receiver. If I turn on the receiver after turning on the HTPC all is fine, but not the other way around. I think the new HDMI cable analyser is causing this issue. Can I turn it off?

[huhn]

Why can't the chroma upscaling algorithm be used beyond the initial conversion to 4:4:4? For example with my settings, it'll upscale chroma initially with ReconSharp, then upscale luma with NGU-veryHigh and the chroma again but now with NGU-med. Couldnt ReconSharp (or any of the initial-chroma-only scalers) be used for the second scaling step as well?

Is it just not optimized for scenarios beyond 4:2:0 -> 4:4:4? However, wouldn't it just be the same scenario as the initial conversion, having a chroma image that is exactly half the luma image if a doubling algo is used, like NGU, NNEDI or SuperXBR?

doubling chroma is pretty much placebo in it self after reaching 4:4:4. so this is a huge waste of processing power.

and that's what i don't get with these new doubling settings.
the change was made to avoid bad settings like chroma doubling but with default settings using NGU med for chroma doubling when NGU very high is used and don't let me start about quadruple chroma super sampling... it is doing this with default settings.

activate 200% super sampling is disabled.

always super sample shouldn't be used. at the moment NGU will super sample when upscaling by a couple of pixels.

NGU very high is using NGU low for chroma quadrupling when super sampling 720p to 2880p while scaling to 2160. changing chroma from "let madVR decide" to "normal" will HALVE the rendertimes. NGU chroma doubling shouldn't be used with default settings and that never ever.

[Damien147]

As said earlier I want to use jinc for a specific resolution and for everything else NGU med.I created 2 profiles,first one is jinc and second is NGU med.I typed this in profiles auto select rules:

if (srcWidth <= 3840) and (srcHeight <= 1600) "Profile 1"
else "Profile 2"

Nothing happens,can you guys correct me?

[har3inger]

I do like Recon Soft with SR too, does SR help with the bleed through in this example?


I find that SR2 is the minimum strength you need to fix those "bleed" problems the reconstruction/bilateral scalers face. However, sometimes the "bleed" is what cleans up edges so nicely in other cases, where you want to remove very thick, dark, bloaty lines in the chroma.

As said earlier I want to use jinc for a specific resolution and for everything else NGU med.I created 2 profiles,first one is jinc and second is NGU med.I typed this in profiles auto select rules:

if (srcWidth <= 3840) and (srcHeight <= 1600) "Profile 1"
else "Profile 2"

Nothing happens,can you guys correct me?


Sometimes madvr doesn't like <= or >= very much in my experience. Try using < 3841 and <1601. Else, if you have a green checkmark, it should select the profiles correctly.

[Asmodian]

As said earlier I want to use jinc for a specific resolution and for everything else NGU med.I created 2 profiles,first one is jinc and second is NGU med.I typed this in profiles auto select rules:

if (srcWidth <= 3840) and (srcHeight <= 1600) "Profile 1"
else "Profile 2"

Nothing happens,can you guys correct me?

Anytime the resolution is above 3840x1600 in either dimension madVR should use Profile 2. This probably isn't what you want so this might be better:
if (srcWidth <= 1920) and (srcHeight <= 1088) "Profile 1"
else "Profile 2"

Sometimes madvr doesn't like <= or >= very much in my experience. Try using < 3841 and <1601. Else, if you have a green checkmark, it should select the profiles correctly.

This is true for refresh rates and frame rates but not for resolutions, in my experiance. With resolution <= has always worked perfectly for me.

[ryrynz]

The real world impact to any of those chroma comparisons is so close to nothing/placebo it's barely even worth discussing.

As madshi has said a million times, luma is way more noticeable/important.

Jesus. Might as well say that the real world impact of ANY of madVR's settings are close to nothing/placebo if that's the case.. you can quite clearly see the changes here even without enlargement. But sure, let's just remove all the chroma options.. it's not important, why have choices for this.. it's barely worth discussing. Madshi I vote to remove everything but reconstruction soft! :rolleyes:

Also I don't even know why you bothered to post :confused:
Maybe your memory isn't too good, most of us have been here since 09/10, I think we know what we're doing.

[Damien147]

Tried what you said but same result,I have a green checkmark both ways but nothing happens.I don't know if I have to add something.

[pirlouy]

The real world impact to any of those chroma comparisons is so close to nothing/placebo it's barely even worth discussing.
At least, they took time to make decent screenshots comparison, which is highly respectable.

[Damien147]

if (srcWidth <= 1920) and (srcHeight <= 1088) "Profile 1"
else "Profile 2"


That worked but backwards,had to put ngu on profile 1 and jinc on profile 2.:confused:I will leave it like that.Thank you both guys.:)

[kalston]

Thanks for those chroma upscaling examples and the sample! Using this sample I think I like NGU low a lot on my current setup. Fast and clean.