madVR - high quality video renderer (GPU assisted)

[omarank]

It's still a 4K image. On a quick check I didn't know where to look. I did a quick check but didn't see any obvious differences between Jinc or super-xbr chroma upscaling.
The differences between the two chroma upscaling algorithms on a still image like this may not be obvious. You may look at the trees or the mountains in the background. When Jinc chroma is used, there is a natural depth to the image, while when super-xbr chroma is used, the image loses that depth. Well this is how I perceive it. And these differences are more observable when mono dithering is used.

Thanks! You were using strength=1.0 and softness=0.0, right? I'm asking just to be sure, because madVR did store and reuse the settings you used from the older SuperRes.
Right.

[chros]

super-xbr chroma upscaling now supports higher quality AR algo -> 6% slower
When Jinc chroma is used, there is a natural depth to the image, while when super-xbr chroma is used, the image loses that depth.
The new chroma upscaling super-xbr + HQ AR is slower now than Jinc + AR , so it's a no go for me (tested on my 650m in p5 powerstate to clearly see the difference).

And 1 more interesting thing: D3D9 exclusive Old path is way-way faster (~40%) then the New path with the same settings (when you switch on several frames in advance)! Is there any advantage (quality wise) to use the New path if everything else seems to work fine with the old one?

Edit: and D3D11 exclusive is the slowest from all them.

[xabregas]

I don't understand you reasoning behind this.

Biggest questions: what would madVR do if you selected s-xbr upscaling and nnedi3 doubling? What image scaling algo would madVR use after doubling with s-xbr?
It's under doubling because you can only double, not scale by any (arbitrary) ratio, performance has nothing to do with this. The up-/down-scaling algo used after doubling is important for image quality.


So super-xbr is like NNEDI3?? Cant be used to upscale 1280x720 videos to 1920x1080 screens?? So it will double and then downscale. Anyway, what would be the best downscale method after doubling? Im using spline and i get better results doubling it then disable doubling and use luma upscaling alone with jinc :X

[fairchild]

And 1 more interesting thing: D3D9 exclusive Old path is way-way faster (~40%) then the New path with the same settings (when you switch on several frames in advance)! Is there any advantage (quality wise) to use the New path if everything else seems to work fine with the old one?

Edit: and D3D11 exclusive is the slowest from all them.

Using the same clip, it's the opposite for me, old-path D3D9 is like 19ms rendering times vs 7ms for new-path D3D9. Probably depends on your GPU/drivers and maybe even MadVR settings. The only quality wise advantage I believe is if you plan to use the 10-bit option which requires D3D11 exclusive.

Also D3D11 8-bit is about the same as new-path D3D9 with. (around 7ms)

[aufkrawall]

When comparing render times, I think it's important to force the graphics card to run with full clocks.

Maybe this is not possible with GCN Radeons, since Powertune is aggressively lowering GPU clock when there is no full load (which is a good thing, apart from this kind of benchmarks).

I can force my GTX 980 to constantly run with full boost clock when doubling 720p60 with NNEDI3 64.
Under these conditions, the render times of D3D9 and D3D11 are almost identical.
However, I can have huge differences when the GPU isn't running with full clock due to insufficient load.

So I think it's pointless to compare rendering times if the GPU load isn't high enough.

[huhn]

a 24/23p source on a 60 hz display should be faster with old path and that always.

old path is doing more work when the frame rate is not the same as the display rate so it is slower but has more control of what is really happening.

fairchild please check your powerstates for these tests.


using a r9 270
i have 38 ms using spline 3 ar for 720p to 1080p scaling and the reason is lowest powerstate possible even with nnedi3 32 i get lower ms.

[aufkrawall]

How big can the impacts of different presentation handling be, compared to Jinc or even Nnedi3?

[fairchild]

a 24/23p source on a 60 hz display should be faster with old path and that always.

fairchild please check your powerstates for these tests.

I use lav copyback and it forces high powerstates when the clip allows. But I just checked an HEVC clip which doesn't use copyback and it still had same results. 12ms with new path d3d9 and 23ms with old path d3d9.

I am using 14.12 official drivers if that makes a difference, also my settings are pretty pedestrian:

ordered dithering with neither option checked
processing-image enhancements: FineSharp - Strength 1.0 (pre-resize manual active for HD)
chroma upscaling: lanczos3 (AR)
image doubling: super-xbr 100 (2x) (active on 960x540 or less)
image upscaling: lanczos3 (AR)
image downscaling: Catmull-Rom (AR+LL)

[chros]

I use lav copyback and it forces high powerstates when the clip allows. But I just checked an HEVC clip which doesn't use copyback and it still had same results. 12ms with new path d3d9 and 23ms with old path d3d9.

I am using 14.12 official drivers if that makes a difference, also my settings are pretty pedestrian:

a 24/23p source on a 60 hz display should be faster with old path and that always.
I'm using 1080p/23.97 content on 1080p/24.01 display (so there's no luma resize) and the framerate is pretty identical. More then that the frame rate switcher is set to these values, and it's working great, and I got the same result with all the different contents:
1080p24, 1080p25, 1080p30, 1080p60

And I wasn't talking about rendering times, but GPU load (though these 2 should depend on each other), using nvidia inspector monitor.
So maybe ATI drivers are behaving completely differently.

Edit: my nvidia driver version is: 347.88 (but I had the same result with the older drivers as well)

[aufkrawall]

I use lav copyback and it forces high powerstates when the clip allows. But I just checked an HEVC clip which doesn't use copyback and it still had same results. 12ms with new path d3d9 and 23ms with old path d3d9.

I can hardly believe these figures. Almost twice as fast, just because of different API?
I'd really rather assume that is somehow a counting/calculation error.
You could check this with madVR settings which almost cost twice as much, e.g. more Needi3 neurons if interop performance is sufficient.

[huhn]

I use lav copyback and it forces high powerstates when the clip allows. But I just checked an HEVC clip which doesn't use copyback and it still had same results. 12ms with new path d3d9 and 23ms with old path d3d9.

I am using 14.12 official drivers if that makes a difference, also my settings are pretty pedestrian:

you should check the power states. dxva has rarely a clear effect on it. on the other side cuvid forces high power state.

just check it with GPU-z

[fairchild]

I can hardly believe these figures. Almost twice as fast, just because of different API?
I'd really rather assume that is somehow a counting/calculation error.
You could check this with madVR settings which almost cost twice as much, e.g. more Needi3 neurons if interop performance is sufficient.

I'm just reporting my findings for my rig. Not trying to convince anybody or anything. Different hardware using different drivers using different settings will produce different results.

you should check the power states. dxva has rarely a clear effect on it. on the other side cuvid forces high power state.

just check it with GPU-z

Well what I've noticed is that when I use D3D9 old-path it maintains 450mhz/1250mhz throughout the video, where as D3D9 new-path or D3D11 with present a frame for every vsync, it fluctuates from 450mhz/1250mz to 1050mhz/1250mhz. Not sure why this causes the reported rendering times by MadVR to go down so drastically.

I'm going to try not using FSE while using D3D11 windowed mode to see if that bug that always forces me to go back to FSE is squashed. (the composition rate changes randomly and this mismatch causes the videos to be choppy. eg when the composition rate is 23.976 and I am watching a 60p video clip)

You were right that the dxva copy-back doesn't seem to affect the clock rates much. I think it's best for the card to stay at a single clock rate, say the 450mhz/1250mhz mid state instead of constantly fluctuating. low power state for my card is 300mhz/150mhz

[aufkrawall]

I'm just reporting my findings for my rig. Not trying to convince anybody or anything. Different hardware using different drivers using different settings will produce different results.

Yeah, but it's kind of esoterism to conclude D3D11 would be magically 100% faster than D3D9 just because of the rendering times (which are not a secure method to measure performance, even according to madshi) without really benchmarking this under clean conditions.

[fairchild]

Yeah, but it's kind of esoterism to conclude D3D11 would be magically 100% faster than D3D9 just because of the rendering times (which are not a secure method to measure performance, even according to madshi) without really benchmarking this under clean conditions.

I actually said in my original post that both D3D9 new-path and D3D11 were about equal. D3D9 old-path is the one that shows the huge increase in rendering times. Anyways... Just use whatever works best for you.

[aufkrawall]

Oops, sorry. I misread that.

[SithUK]

I am using MPC-HC with madVR and a 3D LUT. I am using a laptop as the source. The laptop has an i5 quad core and ATI 4500 equivalent GPU. I am struggling with video playback FPS. Is there a guide that anyone can suggest that suggest how to increase performance madVR performance with a 3D LUT (or even without it)? The big win for me is running the 3D LUT to achieve an accurate Rec709 gamut.

[Warner306]

I am using MPC-HC with madVR and a 3D LUT. I am using a laptop as the source. The laptop has an i5 quad core and ATI 4500 equivalent GPU. I am struggling with video playback FPS. Is there a guide that anyone can suggest that suggest how to increase performance madVR performance with a 3D LUT (or even without it)? The big win for me is running the 3D LUT to achieve an accurate Rec709 gamut.

You need to lower your image scaling settings. I'd start by choosing a lower-quality chroma upscaler such as Bicubic because the difference can be neglible with most sources. Then I'd use image upscaling such as Lanczos3 or lower. Lastly, turn off debanding and any other extra processing. If still over the average rendering queue, start experimenting with the " Trade quality for performance" checkboxes.

[har3inger]

I would actually start first with the trade quality for performance stuff. A better image upscaler or smooth motion on is more noticeable than madvr doing math in 10 bit instead of 16 bit. The setting to toggle 10 bit buffers gives quite a bit of performance for its negligible tradeoff.

[seiyafan]

I think ATI 4500 has a performance similar to HD 2000, if that's the case it would have a pretty hard time doing anything other than the bare minimum.

[SithUK]

I think ATI 4500 has a performance similar to HD 2000, if that's the case it would have a pretty hard time doing anything other than the bare minimum.

I've had a look and the Ati 540v that I have is basically a mobility hd 4550. Userbenchmark.com has it as comparable to an Intel hd2000.

Should I try selecting all the tick boxes in the trade performance for quality section? The default install has the first half of the options selected already.

The default chroma, up scaling, and down scaling options are already as suggested above. (Bicubic and lanczos)

I tried adjusting the 3d lut bit rate option on the trade performance for quality page to the lowest value of 6. I still had stuttering playback.

Is there anyway to divert the load from the gpu to the cpu?my processor is showing max 6%, and that is likely for only one of the 4 cores.

Is my hope of using a 3d lut with this laptop realistic? (Cpu is i5 m450 @ 2.4Ghz, 4Gb ram, Ati 540v gfx)

[Ver Greeneyes]

Is there anyway to divert the load from the gpu to the cpu?my processor is showing max 6%, and that is likely for only one of the 4 cores.Is the decoding being done on the GPU? If you're using the LAV filters (through MPC-HC or otherwise), check the Hardware Acceleration section in the video decoder settings.

[SithUK]

Is the decoding being done on the GPU? If you're using the LAV filters (through MPC-HC or otherwise), check the Hardware Acceleration section in the video decoder settings.

Mpc-hc
Lav filter video settings
Hardware acceleration is using Dxva2 (native)

Even if I set the option to none, would it transfer the madvr 3d lut related processing off my gpu?

[nevcairiel]

CPUs are incredibly bad at video processing (or maybe more accurate, GPUs are extremely good at it), so any video processing on the CPU would be extremely inefficient.

[chros]

Well what I've noticed is that when I use D3D9 old-path it maintains 450mhz/1250mhz throughout the video, where as D3D9 new-path or D3D11 with present a frame for every vsync, it fluctuates from 450mhz/1250mz to 1050mhz/1250mhz. Not sure why this causes the reported rendering times by MadVR to go down so drastically.
I actually said in my original post that both D3D9 new-path and D3D11 were about equal. D3D9 old-path is the one that shows the huge increase in rendering times. Anyways... Just use whatever works best for you.

The 1st sentence in your 1st quote is the answer to your question: if you use Fullscreen Exclusive Oldpath your gpu can do the same processing with slightly worse rendering times on much lower GPU frequency!
That means Oldpath is way faster than the rest.
For me it's important with my laptop, that can be easily as loud as a hover, lower power states/frequencies means less heat which means less noise.

I think ATI 4500 has a performance similar to HD 2000, if that's the case it would have a pretty hard time doing anything other than the bare minimum.

As we talk about this, try out the Fullscreen D3D9 (Oldpath) presentation, it's faster then the rest. (Untick 'present frames in advance' in Fullscreen Exclusive setting. You can check what madvr is using with CTRL+J)

And if you use subtitles (I do) and MPC/MPC-HC , than there's 1-2 subtitle related settings (like animation related) which can affect speed as well. Try them out and check your gpu usage in the meantime with GPU-Z.

CPUs are incredibly bad at video processing (or maybe more accurate, GPUs are extremely good at it), so any video processing on the CPU would be extremely inefficient.
More than that, they use separate part of the gpu (vpu) which nothing to do with other madvr settings, at least with nvidia GPUs (I don't know anything about ATI cards).

[nevcairiel]

More than that, they use separate part of the gpu (vpu) which nothing to do with other madvr settings, at least with nvidia GPUs (I don't know anything about ATI cards).

Wat, processing the video just runs on the normal shaders.
Processing is not decoding.

[madshi]

It s been a while since I asked this question but are there ANY plans at all in the future to get madVR work with PowerDVD? I would LOVE to use a 3Dlut file with powerDVD!!!
You'd have to ask the PowerDVD devs to add support for madVR.

Have the values for SuperRes for chroma changed? I can't remember it worsening ringing with default values, but "now" it does.
Changed compared to what?

i prefer super-xbr strict AR. super-xbr high-quality AR is something ringing.:scared:
Yes, it has more ringing in some situations, but it's also better in some situations. It's not a good idea to judge an algorithm by only using one single image.

I must say super-xbr is amazing, but why not put it in luma upscaling below jinc?
Because it can only double the resolution. It cannot do anything else.

Woah, that's messed up! Look at all those green lines.. Madshi, any room for improvement there?
Maybe. But there's a limit at some point. Either I supress all ringing, or I allow some. If I do allow some, there will always be some situations where the allowed ringing is harmful. But if I surpress all ringing, that can sometimes be harmful, too.

Thanks for the new version of madVR, especially this.

After some testing, Bilateral is now my chroma upscaling algorithm for anime.
It's pure magic, especially with high quality sources :D
Really !?

I didn't even try this algorithm because many users here said it was horrible.
With lowres sources, heavy color bleeding can be observed.
Bilateral can be horrible, or it can leave all other algos in the dust. It all depends on the source.

And 1 more interesting thing: D3D9 exclusive Old path is way-way faster (~40%) then the New path with the same settings (when you switch on several frames in advance)! Is there any advantage (quality wise) to use the New path if everything else seems to work fine with the old one?
No quality advantage, but should be produce more reliable smoothness. If the old mode works fine you, by all means use it. The new path shouldn't really be much slower, though - except maybe if you have a rather slow GPU and if your refresh rate is much higher than the movie frame rate.

[madshi]

madVR v0.88.16 released

http://madshi.net/madVR.zip

* modified SuperRes algorithm slightly (faster, less artifacts, less sharp)
* added SuperRes "use alternative color space" option
* replaced SuperRes "HQ downscaling" option with new "algo" multi option
* removed SuperRes "softness" option (now always 0.00)
* fixed: decoder sending BT.2020 decoding matrix wasn't properly detected
FEEDBACK

All SuperRes users, please play with this new build. There are 2 new options: "algo" and "use alternative color space". I've intentionally named them in such a way that you can't know which does which exactly. I would like to hear your feedback about which settings you like.

This time the upscaling factor is less important (higher factor might make differences more obvious, though). Also this time don't try to find optimal passes/sharpness values. Instead it might make sense to dial passes/sharpness up as much as your GPU can handle, so you can better judge the differences between the various new options. Higher passes/strength values will make the differences more obvious.

Please make sure you have the other sharpening algos (FineSharp, LumaSharpen, AdaptiveSharpen) disabled. And please make sure you have "refine the image only once after upscaling is complete" activated.

[leandronb]

Hi guys. today i started watching a movie and noticed that was not smooth, i saw the stats on madvr and was showing a lot of dropped frames. tried another file and playback was smooth, so i thought it was file specifc.
i noticed that when in windowed mode playback is smooth even in 60hz mode, but when i maximize window it have the same problem than in fullscreen exclusive.
there is some pictures for explanation.
i noticed that rendering time increased drastically when in fullscreen or maximized window but only for this file. i fixed the problem by disabling double luma resolution, i was using always and nnedi3 64 neurons, but i always used this option and no problem happened before. also all my used settings i think is in the screenshots, i am using madvr v0.88.8, nvidia gtx660ti, i5 3570k, 8gb ddr3.
even i lower resolution movies that i think is the heaviest for doubling resolution played fine with no dropped frames and rendering time under 14ms. so what coulb be the problem?
http://i.imgur.com/3c0R3Tr.png
http://i.imgur.com/Sbp6s8Q.png

another question i have is an older one in fact. when i started tweaking a little more with madvr settings read that nnedi3 improves even more the image quality. i noticed that when using this option the image gets a little glitchy with some random dots appearing, i noticed that this was discussed here. what i did to fix was using the mpc x64 with madvr, i don't know if i did it right but the glitches are gone. but my question is, in current versions there is still this problem? do you guys think that nnedi3 worth the performance hit from 4ms to 16ms rendering time?
http://i.imgur.com/HTkFT7W.jpg
http://i.imgur.com/7wAC530.jpg

thanks in advance.

[madshi]

Here you go.

Went from windowed (1s) -> fullscreen (1s) -> windowed (1s) -> fullscreen (~10s) -> alt + f4 to close player in fullscreen. Queues not filling occurred, as well as playback appearing to stutter but not showing dropped frames.

https://www.dropbox.com/s/hi8yiom7xcwayc3/madVR%20-%20log.txt?dl=0

Not sure if it would make much of a difference, but here is another debug where I waited for queues to completely fill before switching modes. The bug occurred in both attempts and is easily repeatable by just toggling between windowed and fullscreen exclusive mode twice.

Windowed (3s) -> fullscreen (3s) -> windowed (3s) -> fullscreen (~10s) -> alt + f4.

https://www.dropbox.com/s/cqo5yif5vykomad/madVR%20-%20log2.txt?dl=0

Again as a reminder, this only seems to occur with D3D11 10bit. D3D9 10bit and D3D11 8bit both seem to be fine.
Your log says that copying the frames to the backbuffer takes ages (about 15ms per frame). Which seems to be a GPU driver or D3D11 issue. Probably nothing I can do about it, unfortunately... :(

Hi guys. today i started watching a movie and noticed that was not smooth, i saw the stats on madvr and was showing a lot of dropped frames. tried another file and playback was smooth, so i thought it was file specifc.
i noticed that when in windowed mode playback is smooth even in 60hz mode, but when i maximize window it have the same problem than in fullscreen exclusive.
there is some pictures for explanation.
i noticed that rendering time increased drastically when in fullscreen or maximized window but only for this file. i fixed the problem by disabling double luma resolution, i was using always and nnedi3 64 neurons, but i always used this option and no problem happened before. also all my used settings i think is in the screenshots, i am using madvr v0.88.8, nvidia gtx660ti, i5 3570k, 8gb ddr3.
even i lower resolution movies that i think is the heaviest for doubling resolution played fine with no dropped frames and rendering time under 14ms. so what coulb be the problem?
Attaching stuff to this forum is not a good idea. It usually takes ages for the attachments to get approved.

In any case, what does the debug OSD (Ctrl+J) say about the movie framerate? And is deinterlacing on or off for the problematic file? My best guess is that it's a 60i file, which madVR has to deinterlace to 60p. Which consumes twice as much processing power as a 24p file does.

another question i have is an older one in fact. when i started tweaking a little more with madvr settings read that nnedi3 improves even more the image quality. i noticed that when using this option the image gets a little glitchy with some random dots appearing, i noticed that this was discussed here. what i did to fix was using the mpc x64 with madvr, i don't know if i did it right but the glitches are gone. but my question is, in current versions there is still this problem?
This is a GPU driver issue. I don't know in which driver versions this occurs and in which not.

[MS-DOS]

Tried new SR with SXBR, 10 passes 1.00 str. Alternative color space makes the edges darker, a bit blurrier and maybe a little more artificial on pretty much everything I test. Subjectively it looks worse. Algo options don't seem to do anything at all, as I can't notice even a slightest difference.

[madshi]

Thanks.

The difference between each neighbor "algo" option is relatively small. But if you switch from e.g. 0 to 3 and back, there should be a visible difference - at least when using a very sharp source and a scaling factor of at least 2.0x. The higher the scaling factor and the sharper the source, the bigger the difference will be between the "algo" and the "use alternative color space" options.

[aufkrawall]

I really like the new alternative color space option.
It gives an anti aliasing effect, but doesn't seem to introduce AA artifacts. However, strength should be highered in return to make contoures not look too thin.
I prefer algo 3: 1 changes brightness for some colors too much and 2 is too soft.
I still don't see a real gain in more passes than 1.
So my new favorite settings for SuperRes are:
passes 1, strength 0.7, algo 3, alternative color space used.
At least it looks great with the low res cartoon example.

Edit: madshi, would it be possible to use lower values than 0.1 with AdaptiveSharpen with the next version?

[tFWo]

My experience with new SR: I turned it off :( as i struggle to find the purpose and actual refinement in this image refinement.

720p movie i tried actually looked softer and fuzzier with any of the SR algos. Alt. color space made it even worse. Turning SR (with alt color space on) off was like removing a thin layer of dirt :scared:

On SD movie (the one i used for those images in my last post) i can see the difference clearly between different algos when zoomed. 0 is sharpest, 2 is softest. 3 is different (sharper than 2). Combination of algo 2 + alt color space is nightmare fuel - Rings of Saturn :D .

Turning SR on with low number of passes with strength 1.00 doesn't do much but more importantly doesn't do anything positive to the picture. Increasing number of passes just brings more artifacts with negligible increase in sharpness.

What is the purpose of this new SR?

[SecurityBunny]

Your log says that copying the frames to the backbuffer takes ages (about 15ms per frame). Which seems to be a GPU driver or D3D11 issue. Probably nothing I can do about it, unfortunately... :(


:( Oh well. Thanks for looking into it at least. I'll just stick with D3D9 10-bit since it can switch window modes problem free.

[aufkrawall]

I can understand that the softness doesn't look good always, and with NNEDI3 it's also an option to just enable AdaptiveSharpen to get a different, but very nice and sharp result.

[Ver Greeneyes]

Even if I set the option to none, would it transfer the madvr 3d lut related processing off my gpu?No, but it would at least do the decoding on the CPU, giving madVR more to work with. My laptop GPU is pretty similar to yours - a Mobility Radeon HD 4650 - and it can handle up to 720p with a 3DLUT just fine. 1080p is too much for it however (at least when downscaling to the laptop monitor - but it struggles to remain smooth on a 1080p external monitor as well). I have it set to do the decoding on my CPU ("None").

[Halfpant.exe]

Liking the new SR changes
I watch a lot of low quality 480p videos
& passes 2, strength. 75 with algo 3 & alternative color space turned on is giving a fairly nice results
& for hd content i am using
algo 0 with alternative color space turned off & adaptive sharpen to 1.5

Sent from my MotoE2(4G-LTE) using Tapatalk

[leandronb]

Your log says that copying the frames to the backbuffer takes ages (about 15ms per frame). Which seems to be a GPU driver or D3D11 issue. Probably nothing I can do about it, unfortunately... :(


Attaching stuff to this forum is not a good idea. It usually takes ages for the attachments to get approved.

In any case, what does the debug OSD (Ctrl+J) say about the movie framerate? And is deinterlacing on or off for the problematic file? My best guess is that it's a 60i file, which madVR has to deinterlace to 60p. Which consumes twice as much processing power as a 24p file does.


This is a GPU driver issue. I don't know in which driver versions this occurs and in which not.
I edited my post with images, i think the file is 23p, but i think the resolution is a bit weird, but don't know. the osd is in the images.

[Anima123]

I edited my post with images, i think the file is 23p, but i think the resolution is a bit weird, but don't know. the osd is in the images.

In your case, since the scale factor is too small, there's no need to activate image doubling. You really should not choose 'always use doubling' option.

[leeperry]

Hah, it's true that being a mVR tester is a full-time job as I just left for a week, missed two subversions and one of my fav settings(SR softness) is already gone :o

Sounds like one more week and NEDI was out too :scared:

So I can remove super-xbr-25 again?
25 is useless as far as I can tell, I'd love to try 37 if any possible please.

I don't mean to sound ironic, pedantic or negative or anything like that so please do not take what I'm gonna ask in any bad way, mVR is an amazing VR and I'm more than thankful for the time and prime rate expertise you invest in it but: Is there any way to run xsbr and SR on my own via a PS script in my media player please?

I realize that you are the game creator and that you make the rules but post-processing settings are highly subjective, many apps come with a "simple/advanced settings" checkbox that has the former enabled by default. Could you for instance imagine Ahead Nero with 5 suboptions? Both newbies and experts want to be able to use it but they have different needs, one part of them likes as many options as possible and the other not. If at some point the latter wanna learn more about what's happening in the back, they can always go expert, learn and finetune it all optimally. Sometimes you don't know you need something till you try it.

At this point, you wanna dumb down mVR as much as possible in order not to scare new users away and all testers have to constantly complain, whine and try to prove their point that their favorite setting should stay(if it's not gone already) :o

currently HQ enabled seems to be better in several cases, although maybe worse in some others. But current tendency goes in favor of HQ enabled.

If you have examples of where HQ looks worse, please show screenshots.

TBH I couldn't care less whether other users would prefer having HQ on or off, it's a free world and we all have different rigs, different tastes and different eyes. Since when is a "tendency" supposed to decide what looks best to me? :(

Especially during summer when most of the EU and US will be away on vacation for at least a few weeks. They were not here, they cannot complain....."Ibiza or mVR, gotta make choices in life dude :D"

If I understood correctly, SR constantly compares and tries to find the best upscaled frame so they really only make sense in motion IMHO. Having HQ on or off might not make a difference on a single screenshot out of its context for all I know, I can always try to post screenshots that would prove my (subjective?) point but I'm talking about differences on motion pictures(as I do not watch slideshows of tiny pixel art and nasty test patterns).

Maybe HQ enabled would look better to me on a plasma, 4K, TN or something....what matters to me is that on my 1080p Sammy LED LCD TV with .13 via DVI(pure untouched 4:4:4 RGB TMDS) I definitely far prefered having HQ off, the same way I really hate the obvious EE created by NNEDI3 and sxbr on 720p@1080p. I like them a lot on tiny videos because NEDI isn't quite extracting as much edges information as them.

Please post screenshots that show where/why you like NEDI better than super-xbr.

If this sample doesn't prove my point to your eyes then nothing will, really: http://forum.doom9.org/showpost.php?p=1722469&postcount=30138
Here's an extreme scenario 40" 720p sample (https://mega.co.nz/#!C5BzAbqJ!od8S-ee7ZxM0v-Y0j3Eg8Xtw1O9QrK_oQXwwH6Vj6lQ) from this movie (http://www.imdb.com/title/tt1043852/), I do own the original BD so I could compare them but with NNEDI3@CR AR LL with the NNEDI3 SR defaults@ED2-8bit, EE is seriously too strong to my eyes as if all background elements in the room were right on the front, it literally kills perspective when OTOH NEDI with the SR non-double defaults looks fantastic to me

NEDI is a like a sharper J3AR to me, the other two come with very strong EE as clearly visibly to me on this sample. I do prefer sxbr over NNEDI3 because it's far less GPU intensive, looks equally über-sharp and works stunningly well on tiny videos. Different needs, different solutions and I'm not stuck with the 13.12 drivers anymore :)

Sorry to say, but the default settings are sharpness 1.0 and softness 0.0. Probably you still had the old defaults from the old SuperRes algo stored.
I did update from the previous build and my settings were nothing like this, sharpness was 28 IIRC.

Anyway, 1.0 sharpness was way too sharp to me in .13 and if SR is bound to become a silly "SR" checkbox with zero subsetting then it sounds like I would be stuck with .13 forever or would have to run SR on my own in my media player(which is totally fine as I like to set all settings to my own personal subjective taste, I'm crazee like that sometimes).

Not sure who you're agreeing with here, maybe yourself? But sorry, no knob planned for now.
Very nice, now you're calling me a nutcase agreeing with himself :)

My post was right under this one and if you look closely I did put a "^" character at the beginning in order to avoid a useless quote and hopefully making clear that I was referring to it:
To me, xbr100 is looking a little too sharp. I tested NEDI+SR (defaults) vs xbr75 doubling alone and they looked pretty much alike except xbr alone will run much faster
It does sound like he'd try 87 should he be able to, the same way I'd love to try 37.

Which raises the same question: would you be so kind as to please explain us how to run sxbr and SR in our media player so we could go bonkers and try any value we like? No more whining afterwards I promise, you'll save time reading it and me writing it :p

Or maybe you could please call those PS scripts from an external text file and allow those would like to get their hands dirty to specify their own favorite settings?

:thanks:

[Nevilne]

That's kinda rude. Look into MPDN.

[leeperry]

^ Already thought about it but I like PotPlayer too much, seamless playback of up to 1000 movies being its strong point.

FWIU these scripts are available as PS scripts so what I would like to do should be totally doable and madshi wrote mVR so he can chose any default setting he likes. I fully respect that and again didn't mean to sound rude in any way, mVR is very obviously light years ahead of all other DS video renderers and we're all extremely thankful for it. I've been testing it since day 1 and it's been a hell of a ride PQ-wise :devil:

[Warner306]

I've had a look and the Ati 540v that I have is basically a mobility hd 4550. Userbenchmark.com has it as comparable to an Intel hd2000.

Should I try selecting all the tick boxes in the trade performance for quality section? The default install has the first half of the options selected already.

The default chroma, up scaling, and down scaling options are already as suggested above. (Bicubic and lanczos)

I tried adjusting the 3d lut bit rate option on the trade performance for quality page to the lowest value of 6. I still had stuttering playback.

Is there anyway to divert the load from the gpu to the cpu?my processor is showing max 6%, and that is likely for only one of the 4 cores.

Is my hope of using a 3d lut with this laptop realistic? (Cpu is i5 m450 @ 2.4Ghz, 4Gb ram, Ati 540v gfx)

DXVA image upscaling could be the way to go for a graphics card that slow. Is is possible to use a 3D LUT with DXVA chroma and image enabled?

[SithUK]

DXVA image upscaling could be the way to go for a graphics card that slow. Is is possible to use a 3D LUT with DXVA chroma and image enabled?

I mostly watch 1080p content at 1080p resolution. Madvr operates using near neighbour by default I think when there is no scaling required. What other options could I try to improve performance with a 3d lut? There are no performance issues when I run without a 3d lut, but the main attraction to me is the calibrated rec709 gamut.

[Eyldebrandt]

My experience with new SR: I turned it off :( as i struggle to find the purpose and actual refinement in this image refinement.

720p movie i tried actually looked softer and fuzzier with any of the SR algos. Alt. color space made it even worse. Turning SR (with alt color space on) off was like removing a thin layer of dirt :scared:

On SD movie (the one i used for those images in my last post) i can see the difference clearly between different algos when zoomed. 0 is sharpest, 2 is softest. 3 is different (sharper than 2). Combination of algo 2 + alt color space is nightmare fuel - Rings of Saturn :D .

Turning SR on with low number of passes with strength 1.00 doesn't do much but more importantly doesn't do anything positive to the picture. Increasing number of passes just brings more artifacts with negligible increase in sharpness.

What is the purpose of this new SR?

+1.
This is more or less my exact thought.
On HQ content, with high resolution screens, NNEDI3 doubling + AS gives better results compared anything with SuperRes inside the plot.
Few passes = does nothing worth turning it on
high passes = does everything you definitely don't want.

[Ver Greeneyes]

I mostly watch 1080p content at 1080p resolution. Madvr operates using near neighbour by default I think when there is no scaling required. What other options could I try to improve performance with a 3d lut? There are no performance issues when I run without a 3d lut, but the main attraction to me is the calibrated rec709 gamut.Did you try CPU decoding? I think that will probably help a lot, assuming your CPU can handle it.

[madshi]

I really like the new alternative color space option.
It gives an anti aliasing effect, but doesn't seem to introduce AA artifacts. However, strength should be highered in return to make contoures not look too thin.
I prefer algo 3: 1 changes brightness for some colors too much and 2 is too soft.
I still don't see a real gain in more passes than 1.
So my new favorite settings for SuperRes are:
passes 1, strength 0.7, algo 3, alternative color space used.
At least it looks great with the low res cartoon example.
Could you please try some other content, too? Preferable also some sharp and some soft real life video instead of Anime, maybe also with ringing and without ringing. These new SuperRes options may look different with different kinds of material. Of course I need to find values which work best as an allround solution.

Edit: madshi, would it be possible to use lower values than 0.1 with AdaptiveSharpen with the next version?
Ok, no big problem.

My experience with new SR: I turned it off :( as i struggle to find the purpose and actual refinement in this image refinement.

720p movie i tried actually looked softer and fuzzier with any of the SR algos. Alt. color space made it even worse. Turning SR (with alt color space on) off was like removing a thin layer of dirt :scared:

On SD movie (the one i used for those images in my last post) i can see the difference clearly between different algos when zoomed. 0 is sharpest, 2 is softest. 3 is different (sharper than 2). Combination of algo 2 + alt color space is nightmare fuel - Rings of Saturn :D .

Turning SR on with low number of passes with strength 1.00 doesn't do much but more importantly doesn't do anything positive to the picture. Increasing number of passes just brings more artifacts with negligible increase in sharpness.

What is the purpose of this new SR?
Rings of Saturn, hehe... :D

Liking the new SR changes
I watch a lot of low quality 480p videos
& passes 2, strength. 75 with algo 3 & alternative color space turned on is giving a fairly nice results
& for hd content i am using
algo 0 with alternative color space turned off & adaptive sharpen to 1.5
If you had to choose one setting for all sources, which would it be?

I edited my post with images, i think the file is 23p, but i think the resolution is a bit weird, but don't know. the osd is in the images.
Yes, the resolution is at fault. Scaling near 1080p to 1080p has quite high demands. As Anima123 already mentioned, in that situation it doesn't really make much sense to force image doubling. Basically you're doubling near 1080p to near 4K and then scale it back down to 1080p.

Is there any way to run xsbr and SR on my own via a PS script in my media player please?
Not with madVR, no. madVR does currently not support custom shaders which change video resolutions. This is planned for a future version, but it will take some time to get there.

I realize that you are the game creator and that you make the rules but post-processing settings are highly subjective, many apps come with a "simple/advanced settings" checkbox that has the former enabled by default. Could you for instance imagine Ahead Nero with 5 suboptions? Both newbies and experts want to be able to use it but they have different needs, one part of them likes as many options as possible and the other not. If at some point the latter wanna learn more about what's happening in the back, they can always go expert, learn and finetune it all optimally. Sometimes you don't know you need something till you try it.

At this point, you wanna dumb down mVR as much as possible in order not to scare new users away and all testers have to constantly complain, whine and try to prove their point that their favorite setting should stay(if it's not gone already) :o
I consider myself an expert, too, but I still want to get rid of as many options as possible. Even most experts do not want to adjust settings to different values for each movie. I know that some madVR users would prefer me to keep all options, but sorry, not going to happen. I do try to remove only options which I consider not useful enough to be worth the effort. And I know that some users may disagree about some of those decisions. But I make my decisions after listening to your feedback, and of course I have my own opinion, too, and I try to fuse them to get to the best possible end result. Anyway, you'll have to live with whatever decisions I make, sorry.

TBH I couldn't care less whether other users would prefer having HQ on or off, it's a free world and we all have different rigs, different tastes and different eyes. Since when is a "tendency" supposed to decide what looks best to me? :(
What looks best to you isn't always fixed, though. One day you love a noisy picture and ask for ways to add even more grain. The next day you want to get rid of as much noise as possible. And when testing the dithering algos, you changed your mind a few times, too. Just saying this so that you can understand better that I can't make decisions based on the feedback of one single user. Which options look good or bad can be a matter of taste, sometimes, but sometimes it also simply depends on with which content you tested. Some options look better with some content but worse with other content. I need to find options which look better for the majority of content and the majority of users. And that's what I'm trying to do. Furthermore, HQ downscaling is actually the scientifically and technically more correct solution. Actually, the non-HQ-downscaling implementation was broken. It did not work as it was originally intended. Still, you liked it better, which is your right. But several other users had a different opinion, *I* had a different opinion, and then add to that the scientific background and the fact that the non-HQ option was actually doing stuff wrong. All this combined led me to the decision to use HQ downscaling.

If I understood correctly, SR constantly compares and tries to find the best upscaled frame so they really only make sense in motion IMHO.
No. What you're describing is one method/implementation of SuperRes. But that's totally different to what Shiandow is doing. Shiandow's SuperRes strictly looks at each frame separately.

If this sample doesn't prove my point to your eyes then nothing will, really: http://forum.doom9.org/showpost.php?p=1722469&postcount=30138
This is a custom reencoded Anime sample with ringing artifacts in the source. I've not compared this sample with HQ downscaling on/off. But in any case, do you really want to base your decisions on an Anime reencoding which has EE artifacts in the source already? Of course testing such sources is useful, too, but only as one of many tests.

Anyway, 1.0 sharpness was way too sharp to me in .13 and if SR is bound to become a silly "SR" checkbox with zero subsetting then it sounds like I would be stuck with .13 forever or would have to run SR on my own in my media player(which is totally fine as I like to set all settings to my own personal subjective taste, I'm crazee like that sometimes).
There are new options to play with in build v0.88.16. Let me know what you think. I do value your feedback, but your feedback doesn't "overwrite" the feedback of other users.

Very nice, now you're calling me a nutcase agreeing with himself :)

My post was right under this one and if you look closely I did put a "^" character at the beginning in order to avoid a useless quote and hopefully making clear that I was referring to it:

It does sound like he'd try 87 should he be able to, the same way I'd love to try 37.
Neither he nor anyone else in the proximity of your original comment made a wish at all about a super-xbr sharpness knob. So there was nobody you could agree with there, except yourself... :p

I mostly watch 1080p content at 1080p resolution. Madvr operates using near neighbour by default I think when there is no scaling required. What other options could I try to improve performance with a 3d lut? There are no performance issues when I run without a 3d lut, but the main attraction to me is the calibrated rec709 gamut.
Try setting chroma upscaling to Bilinear.

[madshi]

+1.
This is more or less my exact thought.
On HQ content, with high resolution screens, NNEDI3 doubling + AS gives better results compared anything with SuperRes inside the plot.
Few passes = does nothing worth turning it on
high passes = does everything you definitely don't want.
Regardless of settings?

Did you have a different opinion about the old SuperRes algorithm?

[reverepink]

Recommend me some settings for video playback, please.

My setup is:

Win 8.1
PotPlayet x64
MadVR and LAV Video Decoder
Dell U3011 2560*1600 Monitor
NVIDIA GTX 650Ti

I mostly try to watch HD remuxes and some anime webrips.

[dbcooper]

Any performance improvements with AMD Catalyst 15.7 drivers?