madVR - high quality video renderer (GPU assisted)

[leeperry]

most of bad luck that the Polaris driver sabotages even that
You don't say, I'm about to finally upgrade from W7SP1 to W8.1 as I need 10bit CLUT support and I thought I'd go x64 as new AMD drivers won't be released in x86 anymore but if NGU is slow with the drivers for the new boards then I might as well stick to x86 coz Reclock is a pre for me. 13.12 are still top notch on W7, hopefully they'll work on W8.1 and I can get a 290-something later if need be as that should also work with the 13.12's.

[Anima123]

Here's a new test build for those of you who still can't get NGU to activate (e.g. Optimus users):

http://madshi.net/madVR091d.rar

Installation instructions:

1) Make sure you're in release mode.
2) Overwrite the files.
3) Activate debug mode.
4) Create a short log file and upload it somewhere for me to look at (don't attach to this forum).

There's a chance it might work now. If it does, I don't need a log. If it still doesn't work, please upload a log, thanks.

The player just skipped NGU this time, with debug mode. Logs as following url: https://www.dropbox.com/s/sangxsn2m1vev3o/madVR%20-%20log.rar?dl=0

[hannes69]

Coil whine:
Not a new problem. Can affect all graphic cards regardless of manufacturer, series, price. It is dependant of two factors: Quality of the coils (including "antiwhine" solutions) and the frequency applied to the coil (resonation).
I am NOT a fan of throwing away working products and replacing it with new ones (pollution, inconveniences with shipping,... unneeded waste of money,...) A graphic card with coil whine is a working device with a little problem that has to be solved (throwing away is NOT a solution in my opinion, it is a "solution" for the incapable man who can´t help himself and has only the option for spending money;) )
As already mentioned by others:
First approach is to change the frequency. The coil is whining when "fired" with certain frequencies. Depending on the electronic function the whining coil has, there is a chance that changing the frequency of the GPU or the GPU memory lowers or eliminates the coil whining. This is possible for both frequencies (GPU and GPU memory) and in both directions (overclocking or underclocking). You have to try out all the possibilities.
There is also the possibility to change the whining by overvolting or undervolting because a coil could then be fed with more or less current which can influence the whining.
Example from my experience (other device but same story): My TFT monitor had coil whining. I changed the vertical frequency from the default 60Hz to 75Hz by using a custom resolution. The whining has disappeared completely.
Second approach is a "physical" one. A coil is whining because it is vibrating with a certain audible frequency and we hear that because the coil is giving the sound to the air by moving within it a little bit. When eliminating this movement it can´t whine anymore. Take a piece of a hose or a drinking straw as a stethoskope and locate the whining coil. Now fix the coil by using a good amount of nail polish, plastic spray, glue, whatever. Use your "glue" only on the coil, don´t mess around it all over the place. If correctly applied, the whining has gone.
This doesn´t work always, but there is a good chance. If you don´t use much too much of the glue, nothing can go wrong with this method. Worst case the whining stays.
On a GPU there are no "dangerous" voltages, if only applied to the coil there will be no temperature problem (local overheating).
I have done this method several times with success. The devices worked several years after the procedure total normally.
Maybe don´t do that within guarantee time because when the service looks at such changes the guarantee could be denied.
Good luck!
:)

[Knight77]

Guys I need an advice, could you help?

I have a GTX 960 and a 4K TV. Currently I'm using Super-xbr 100 with chroma and luma upscaling to 4K, it's ok but I can see the GPU is like 70% used and the fan is spinning a lot producing noise. I tried NNEDI3 32 and the new NGU but it cannot handle it at all...it's really superslow and unwatchable. Based on this I was planning to buy a Gtx 1060 but I'm having doubts...it is better to spend some more money and get a 1070 or it wouldn't change much? By the money point of view actually there is some difference so I'd like to understand if it is worth it or not.

Any reply is appreciated.

[MariaX9]

Quick Question is "Scale in Linear Light" recommended for SSIM downscaling?

[aufkrawall]

Yes, scaling down in LL is closer to the original than scaling down in gamma light.

[Warner306]

I finally had a chance to try NGU.

GTX 750 Ti and 1080p display

Artifact Removal - Debanding: medium/high
Dithering: Error Diffusion 2
Image enhancements: off
upscaling refinement: off

1080p -> 1080p
Chroma upscaling: NGU medium

28 ms

720p -> 1080p
Chroma upscaling: Bicubic 75 + AR
Luma Doubling: NGU medium
Image Downscaling: SSIM 1D 100%

42 ms

480p -> 1080p
Chroma upscaling: superxbr100 + AR
Luma Doubling: NGU high
Image upscaling: Jinc + AR

26 ms

1) I am in favour of removing redundant settings. NNEDI3 is fairly redundant now that a superior scaler is available at the same performance level. Why use the lesser scaler if given a choice?

2) Not sure about SuperRes. NGU is already very sharp. Should definately stay for lowered powered algorithms.

3) Why would anyone use NEDI? Again, at that processing level, NGU is available and superior.

I am eagerly anticipating a lower performance level of NGU to replace super-xbr image doubling at 720p -> 1080p. If you could do that, then I would be impressed.

[aufkrawall]

1) I am in favour of removing redundant settings. NNEDI3 is fairly redundant now that a superior scaler is available at the same performance level. Why use the lesser scaler if given a choice?

NNEDI3 can still be useful if you watch a lot of cartoons/animes which contain a lot of artifacts. Especially since you can combine NNEDI3 with madVR's deringing filter, which is not possible with NG1 without ugly side-effects.


3) Why would anyone use NEDI? Again, at that processing level, NGU is available and superior.

Agreed, I find NEDI redundant for a long time already.
The same goes for the redundant bicubic/softcubic options (M-N, C-R), nearest neighbor, anti-ringing for SuperRes (Edit: I'd also remove the LL checkbox for SR, seems quite theoretical to me)...
Of the chroma scalers, I'd remove bileteral completely and the reconstruction options, except of sharp AR on/off.


If you could do that, then I would be impressed.
If "NGU low" will be better than super-xbr for the same costs, imho also super-xbr could be removed.

[madshi]

madVR v0.90.1 released

http://madshi.net/madVR.zip

* added NGU "low" quality algorithm
* "soften edges" and "add grain" now allow smaller adjustment steps
* fixed: install batch didn't work
* fixed: on some PCs NGU didn't activate
* fixed: on some PCs keyboard shortcuts didn't work
* fixed: NGU grain was static
* "ShowRenderSteps" hack no longer slows down NGU as much as before
* "ShowRenderSteps" hack only slows down stuff as long as the OSD is visible
I hope that OSD and NGU now work for everyone (again)?

Please let me know what you think about the NGU "low" variant. With my AMD 7750 it's almost twice as fast as "medium", and it's now faster for me than Jinc/super-xbr + SuperRes 1. However, it depends on which image upscaling algorithm you've selected, because Jinc/super-xbr double both luma and chroma at the same time, while if you tell NGU to only double luma, chroma needs to be doubled separately. So if you use an expensive algorithm for chroma doubling, NGU-Low may be slower again than Jinc/super-xbr + SuperRes. So for best speed vs quality ratio I'd suggest to combine NGU-Low for luma doubling and to choose e.g. CatromAR for "image upscaling".

[leeperry]

OSD works :thanks:

[XTrojan]

Tested NGU Low on GTX 970 on Luma channel from 1080->2160.

Super-Xbr is still much more efficient(2x?), and it doubles chroma as well, I used Jinc3 for Chroma upscaling with the NGU.

[madshi]

Super-Xbr is still much more efficient(2x?), and it doubles chroma as well, I used Jinc3 for Chroma upscaling with the NGU.
Which image upscaling algorithm did you use during this test? See v0.91.1 announcement post for details about why I'm asking.

[XTrojan]

Which image upscaling algorithm did you use during this test? See v0.91.1 announcement post for details about why I'm asking.
I used v0.91.1

Jinc3, Doubling Luma was NGU Low

Other one was Super-Xbr100 + (Superres3 + anti-ringing)

NGU Low+Jinc3 was still draining way too much, about twice as much as Super-Xbr100.

[aufkrawall]

To define efficiency you also have to take account of the result and not just the cost, and I don't think that super-xbr can compete with NGU low in terms of the result. NGU low is still much sharper and very artifact free.
Haven't taken a real look at the performance yet, but the image quality still seems to be very good. Looks even better to my eyes than NGU medium with the American Dad sample. It's blurrier, but also less aliased.
As a chroma scaler, there still doesn't seem to be a real difference to the high setting, so we most likely can save a lot of GPU power here.

madshi, any chance that the deringing filter will be more compatible with NGU in the future? I suspect then we could even get rid of NNEDI3.

[Shiandow]

Quick Question is "Scale in Linear Light" recommended for SSIM downscaling?

Yes, scaling down in LL is closer to the original than scaling down in gamma light.

I'm not entirely sure which is better. Generally downscaling in linear light is better, but SSIM also enhances the details, for which it is better to use gamma light. I've been trying to come up with a way to get the best of both worlds, but I haven't found a complete solution, yet.

[madshi]

I used v0.91.1

Jinc3, Doubling Luma was NGU Low

Other one was Super-Xbr100 + (Superres3 + anti-ringing)

NGU Low+Jinc3 was still draining way too much, about twice as much as Super-Xbr100.

Did you turn off SuperRes when using NGU? NGU doesn't really need SuperRes, so activating SR for NGU is more or less a waste of GPU power.

Please compare this:

1) super-xbr100 + SuperRes at your preferred level
2) NGU-Low *without* SuperRes, and please use CatromAR in the "image upscaling" settings section

If 1) is faster for you, I'd be very surprised (unless you're using an RX480, maybe).

any chance that the deringing filter will be more compatible with NGU in the future? I suspect then we could even get rid of NNEDI3.
I don't really know how to improve the deringing filter much further. It is what it is. At some point I might try to use the NGU algorithm framework to create a better deringing filter, but I'm not sure how slow that would be. Since NGU itself already derings pretty well I'm not sure how much need there still is for a separate deringing filter, when upscaling with NGU?

[madshi]

I'm not entirely sure which is better. Generally downscaling in linear light is better, but SSIM also enhances the details, for which it is better to use gamma light. I've been trying to come up with a way to get the best of both worlds, but I haven't found a complete solution, yet.
Interesting. Why is SSIM's detail enhancement better in gamma light, if I may ask?

[thighhighs]

I cant test on uhd tv now, but i test 1080p29 supersampling on full-hd screen. Settings: no deringing, low\high debanding, chroma upscaling bicubic 75 AR, image downscaling bicubik 150 AR LL, orderer dithering, gpu gtx690.

super-xbr 100 doubling + SuperRes 3: 28ms
NGU-low doubling: 37ms

Looks super-xbr more fast, dont remove it :)

[madshi]

super-xbr 100 doubling + SuperRes 3: 28ms
NGU-low doubling: 37ms
Same questions I asked XTrojan:

a) Did you disable SuperRes when using NGU?
b) Which algorithm did you have selected in the "image upscaling" section when testing NGU?

copy&paste from my reply to XTrojan:

Please compare this:

1) super-xbr100 + SuperRes at your preferred level
2) NGU-Low *without* SuperRes, and please use CatromAR in the "image upscaling" settings section

If 1) is faster for you, I'd be very surprised (unless you're using an RX480, maybe).

[XTrojan]

Same questions I asked XTrojan:

a) Did you disable SuperRes when using NGU?
b) Which algorithm did you have selected in the "image upscaling" section when testing NGU?

copy&paste from my reply to XTrojan:

Please compare this:

1) super-xbr100 + SuperRes at your preferred level
2) NGU-Low *without* SuperRes, and please use CatromAR in the "image upscaling" settings section

If 1) is faster for you, I'd be very surprised (unless you're using an RX480, maybe).
Used Superres for both, is NGU bad with superres or did i miss something?

[aufkrawall]

I don't really know how to improve the deringing filter much further. It is what it is. At some point I might try to use the NGU algorithm framework to create a better deringing filter, but I'm not sure how slow that would be. Since NGU itself already derings pretty well I'm not sure how much need there still is for a separate deringing filter, when upscaling with NGU?
I was hoping that with your own scaler, deringing filter and upscaler could work more "coordinated" to prevent it turning the "reduced ringing" into "aliased ringing".
I replied to huhn regarding the anime example with MPEG2 artifacts:
http://forum.doom9.org/showthread.php?p=1786198#post1786198

NNEDI3 64 + deringing:
http://abload.de/thumb/nnedi3deringihupk.png (http://abload.de/image.php?img=nnedi3deringihupk.png)

NGU high:
http://abload.de/thumb/ngu4su7y.png (http://abload.de/image.php?img=ngu4su7y.png)

The amount of ringing gives NGU quite some headaches.
That "aliased ringing" effect with NGU + deringing is very well visible with the other American Dad cartoon example.

[thighhighs]

super-xbr 100 doubling + SuperRes 3: 28ms
NGU-low doubling: 37ms
I use NGU without SuperRes. Doubling luma and chroma in both cases here. Image upscaling Jinc.
Another settings

NGU-low double luma only, image upscaling C-R + AR: 18ms. It fast.

[madshi]

Used Superres for both, is NGU bad with superres or did i miss something?
As far as I can see, NGU produces results which are so good that SuperRes doesn't know how to further improve them. So using SuperRes with NGU simply doesn't bring any benefit. It may even harm image quality. It's better (for image quality and speed!) to use NGU on its own, without SuperRes.

[JarrettH]

Is it necessary for madvr to be 50 mb?

[Shiandow]

Interesting. Why is SSIM's detail enhancement better in gamma light, if I may ask?

Linear light might be more physically accurate, but if you want to measure detail then gamma light is closer to how humans perceive detail. In theory something like CIELAB would be even better, but the difference are pretty small so I'm not sure if that will matter that much.

Obviously you run into problem because, paradoxically, humans estimate brightness as if the image was in linear light, which means some parts of SSIM should be done in linear light, while others should be done in gamma light (or Lab or similar), and I haven't entirely figured out which is which.

[madshi]

NGU-low double luma only, image uscaling C-R + AR: 18ms. It fast.
And which has overall better image quality now? This setting? Or your super-xbr + SuperRes setting?

I was hoping that with your own scaler, deringing filter and upscaler could work more "coordinated" to prevent it turning the "reduced ringing" into "aliased ringing".
Sorry, but no. I don't know how to coordinate them.

[aufkrawall]

Is it necessary for madvr to be 50 mb?
Seems like the archives are not much compressed anymore. I'd recommend 7z with ultra compression, archive is only 16MB of size then.

[cork_OS]

As far as I can see, NGU produces results which are so good that SuperRes doesn't know how to further improve them. So using SuperRes with NGU simply doesn't bring any benefit. It may even harm image quality. It's better (for image quality and speed!) to use NGU on its own, without SuperRes.
Thank you, I just had a question about SR effect on NGU.
So maybe NGU should just ignore "SR on" mark?

[madshi]

Is it necessary for madvr to be 50 mb?
I've all my pixel shaders precompiled in a million different versions. That consumes a lot of space. But it helps keeping madVR startup times small.

Linear light might be more physically accurate, but if you want to measure detail then gamma light is closer to how humans perceive detail. In theory something like CIELAB would be even better, but the difference are pretty small so I'm not sure if that will matter that much.

Obviously you run into problem because, paradoxically, humans estimate brightness as if the image was in linear light, which means some parts of SSIM should be done in linear light, while others should be done in gamma light (or Lab or similar), and I haven't entirely figured out which is which.
Ouch. I wonder if getting this exactly right would noticeably improve image quality. I rather guess that the visible difference might be small. But it's a wild uneducated guess.

So maybe NGU should just ignore "SR on" mark?
I've already thought about that. But leeperry would run amok if I did that... :p

[JarrettH]

i've already thought about that. But leeperry would run amok if i did that... :p

;) :d :d

[heiseikiseki]

madVR v0.90.1 released

http://madshi.net/madVR.zip

* added NGU "low" quality algorithm
* "soften edges" and "add grain" now allow smaller adjustment steps
* fixed: install batch didn't work
* fixed: on some PCs NGU didn't activate
* fixed: on some PCs keyboard shortcuts didn't work
* fixed: NGU grain was static
* "ShowRenderSteps" hack no longer slows down NGU as much as before
* "ShowRenderSteps" hack only slows down stuff as long as the OSD is visible
I hope that OSD and NGU now work for everyone (again)?

Please let me know what you think about the NGU "low" variant. With my AMD 7750 it's almost twice as fast as "medium", and it's now faster for me than Jinc/super-xbr + SuperRes 1. However, it depends on which image upscaling algorithm you've selected, because Jinc/super-xbr double both luma and chroma at the same time, while if you tell NGU to only double luma, chroma needs to be doubled separately. So if you use an expensive algorithm for chroma doubling, NGU-Low may be slower again than Jinc/super-xbr + SuperRes. So for best speed vs quality ratio I'd suggest to combine NGU-Low for luma doubling and to choose e.g. CatromAR for "image upscaling".

Everything works here.
Thanks so much!

And NGU looks amazing for Anime!

for gtx 860m (optimus)
720p->1440p in NGU-low (chroma and doubling) for 35ms which is near to jinc ar or super-xbr.

[madshi]

720p->1440p in NGU-low (chroma and doubling) for 35ms which is near to jinc ar or super-xbr.
Is the doubling just for luma or for both luma and chroma? I'd suggest to only use luma doubling. And in image upscaling settings try CatromAR.

[Shiandow]

Ouch. I wonder if getting this exactly right would noticeably improve image quality. I rather guess that the visible difference might be small. But it's a wild uneducated guess.

Yeah I wonder that as well... FWIW I'm pretty sure downscaling in gamma light causes problems, so either running SSIM in linear light magically causes no problems whatsoever, or there's still room for improvement.

Long story short, linear light SSIM requires more testing.

[madshi]

either running SSIM in linear light magically causes no problems whatsoever, or there's still room for improvement.

Long story short, linear light SSIM requires more testing.

For now it seems to work well enough for me. When switching between gamma and linear light, the biggest differences I can see seem to be caused by the downscaling operation itself, not by the SSIM processing. Not 100% sure about that, though.

[aufkrawall]

And NGU looks amazing for Anime!

Not just anime, everything that's not utterly broken. :D

I can use NGU low for 1080p -> 4k -> WQHD btw. even with deband and NGU low chroma scaler (but not doubler). But SSIM is too much, bicubic works.
Edit: Nope, I can't. Queues get empty.

[madshi]

One thing that would be worth investigating:

Does SSIM downscaling produce any visible benefits over e.g. Bicubic150, when downscaling an image that was doubled before?

The reason I'm wondering is that one key benefit of SSIM is reproducing texture detail nicely. But when doubling an image, there's not much texture detail in the doubled image. So I think Bicubic150 might work just as well as SSIM in that situation. It's quite possible that SSIM only really helps when directly downscaling a high-res source. Anybody up for testing this in depth?

[thighhighs]

And which has overall better image quality now? This setting? Or your super-xbr + SuperRes setting?Super-XBR doubling + SuperRes 3 not better, but more slow vs NGU-low luma\C-R AR (or lanczos3 AR, ~18ms too) in my small test.

[jerryleungwh]

I am using a 960M GPU and NGU-low for image doubling is too much for my GPU when I try to upscale 720p->1080p even when I'm just using bicubic75 for both chroma upscaling and image downscaling and is causing a lot of frame drops, what would be a recommendation for me to use the algorithm for image doubling?

And I have been trying to use NGU-med for chroma upscaling as that's the best my GPU can handle, so how does it compare with NNEDI32 SR1(which is what I have been using before the update) for chroma upscaling?

[Q-the-STORM]

win10 x64 + GTX 780 Ti SC
NGU works fine here, didn't have hotkey issues... madVR v0.90.0 seemed to start up very slowly (2-3 seconds instead of starting instantly), but v0.90.1 is starting instantly again, so that's back to normal...

When playing 1080p videos on a 2160p screen, using NGU doubling on chroma and luma, there should only be NGU-med showing up in the OSD, but mine is showing:
chroma > NGU-med
luma > NGU-med
chroma > lanczos3 AR


lanczos3 with AR is set in image upscaling, but since this is just 1920x1080 to 3840x2160 doubling, why is lanczos showing up in the OSD?

[madshi]

I am using a 960M GPU and NGU-low for image doubling is too much for my GPU when I try to upscale 720p->1080p even when I'm just using bicubic75 for both chroma upscaling and image downscaling and is causing a lot of frame drops, what would be a recommendation for me to use the algorithm for image doubling?
Which algo for "image upscaling"?

[madshi]

When playing 1080p videos on a 2160p screen, using NGU doubling on chroma and luma, there should only be NGU-med showing up in the OSD, but mine is showing:
chroma > NGU-med
luma > NGU-med
chroma > lanczos3 AR
Looks like you probably have NGU-med active in "chroma uspcaling" and "luma doubling", but you have "chroma doubling" unchecked? At least that's what your OSD suggests.

[Q-the-STORM]

Looks like you probably have NGU-med active in "chroma uspcaling" and "luma doubling", but you have "chroma doubling" unchecked? At least that's what your OSD suggests.

oh, forgot to check the box in my 1080p profile.... thx

[jerryleungwh]

Which algo for "image upscaling"?

I use Jinc for that, but I thought it isn't used because after doubling a 720p image it is then downscaled back to 1080p?

[madshi]

I use Jinc for that, but I thought it isn't used because after doubling a 720p image it is then downscaled back to 1080p?
Depends on your trade quality settings. Try Catrom AR for image upscaling.

[heiseikiseki]

Is the doubling just for luma or for both luma and chroma? I'd suggest to only use luma doubling. And in image upscaling settings try CatromAR.

Thanks!
In that setting I can use NGU-med for lumba for 29ms !

[huhn]

RX 480
FHD - UHD 23p
NGU low 19 ms
NGU mid 29 ms
NGU high ~85ms
super xbr 100 12 ms
nnedi3 32 22 ms
nnedi3 64 32 ms
nnedi3 128 46 ms


i know the card doesn't like current NGU but but adding NGU low isn't bad i guess

[AndreaMG]

@Madshi, "addgrain" is not working anymore on latest release for me.

[aufkrawall]

It should work (it does here), are you sure NGU is actually active?

[AndreaMG]

It should work (it does here), are you sure NGU is actually active?
Yes I am. Do you find it subtler or it only changes patter each frame now?

[Ver Greeneyes]

Looks like the GTX 1060 in my new laptop can't quite handle 720p60->1440p60 luma doubling with NGU low.. that's a pity. 16ms isn't much to work with though :) I'm going to have to spend some quality time remaking my profiles to take NGU into account!