madVR - high quality video renderer (GPU assisted)

[mimi123]

Originally Posted by djfred93

MadVR can't do 10 bit output. A lot of answer is in this thread .

Very thanks for your reply, the tread you tell me has been very useful.

I read these tread and I understand there is minus to none improvement on image quality, at least for the normal BD content, playing at 10 bits, is my understanding correct?

Anyway, can you tell me a good card, or some, to get the full potential of Madvr?

[Qotscha]

Originally Posted by MistahBonzai

The use of openCL error diffusion demands too many GPU cycles that I could put to better advantage elsewhere. Simple playback of native 1080P via MadVR at default settings uses 19% GPU while simply enabling openCL error diffusion bumps it up to a steady 68% - and that's with a PCIE 3 2GB HD 7850 (latest Catalyst 13.2 driver pkg), i73770@3.4GH, Windows 7 64bit on a SSD. Display is 40" Sony Bravia (1080P w/10-bit panel) via HDMI 1.3b.

A bit of background... I mostly use MPC-HC, MadVR, LAV, FFDshow (raw), Reclock, AviSynth and AC3 Filter. Viewing/testing consists of analysis of video sample quality utilizing AviSynth(MT) based utilities - I don't watch movies or TV but I love to tweak :-). These samples often exceed 1080P (1920*1080/29.976 progressive) - going up to 4096*2304 DCI-P3.

For the past few weeks I've focused on AviSynth OpenCL variants - notably nnedi3ocl. In my estimation the current MadVR release (87.3) performs similar (within 15% or so) to what I have seen via AviSynth. My take on this particular usage of OpenCL is that it facilitates load balancing across the GPU/CPU while nnedi3 can certainly improve up-scaling quality - you can work wonders with low-rez porn clips :-).

It isn't a free lunch. And it's very heavily GPU dependent - resolution is everything. Up-scaling 854*480P to 1080P vis Interframe2 and/or SVP is about the practical limit on my rig before encountering occasional dropped frames. Note that I simultaneously employ video smoothing (frame interpolation) utilizing frame doubling, sharping and the like as needed. YMMV :-)

I offer-up a sincere thanks to all involved. And Madshi..you are the bomb! :-)

Thanks for clarification. Now I believe that error diffusion really is that heavy. It just felt first a little hard to believe that my card, which was formerly "fully madVR capable", i.e. able to do 1440*1080i60 at 1080p60 with Jinc 3 + AR, can't utilize error diffusion at 1080p50 with any settings.

I fully agree about the quality of NNEDI on low resolution content.

And thank you madshi for your continuous work to improve viewing experience!

[leeperry]

Originally Posted by madshi

You can tell madVR to double your image size, or to quadruple it with NNEDI3. If you tell madVR to only double it, you get 720*576. If you tell madVR to quadruple it, you get 1440*1152. But whether madVR actually doubles or quadruples depends on which exact settings you're using in the "image doubling" section. There are several radio boxes in there you can check which controls all this. In any case, if the output of "image doubling" is smaller than the target resolution, madVR will use the algorithm from the "image upscaling" settings page to upscale the "image doubling" output to the needed target resolution. If the output of "image doubling" is larger than the target res, madVR will usw the image downscaling algorithm to achieve the target res.

OK sounds good, TYVM for the detailed explanations! Maybe down the road you could start compiling this in the OP or into a forthcoming user manual.

So I cannot get NNEDI to upscale 360*288 to 1440*576? I was under the impression that NNEDI allowed either 2X horizontal and/or vertical upscale? This would increase performance I presume?

And you recently said that NNEDI was kinda overkill for chroma, but for clarity sakes's what did you use for chroma in this screenshots comparison please?

Originally Posted by 6233638

Error handling is definitely better with software decoding.
Someone posted a really good sample video here (probably years ago at this point) which shows errors with the hardware decoders that do not exist with LAV's software decoding

Yeah, you can always find nasty samples that will make some decoders output junk, the big difference here is that the libavcodec ppl are very responsive to this kind of problem.

[Ver Greeneyes]

Originally Posted by Asmodian

As I understand their preferences 6233638 prefers less ringing to a bit sharper which is why they prefer Mitchell-Netravali to Bicubic 75. This is of course perfectly reasonable. I hope I am correct as to your reasoning, 6233638.

Thanks, but yeah, I think my preferences are more in line with those of 6233638 I hate artifacts much more than a 'soft' look.

[Werewolfy]

@madshi

Thanks for all your work, I was very busy this week end so I haven't followed the recent updates. I just can say that I have no issues with your latest build except the OpenCl features because I have a Nvidia GPU.

But the profiles are great, more flexible than I have imagined! I was a little bit afraid at the beginning, it seemed a little bit complicated but after some testings, it's not so hard. I'm sure this is the price to pay for this high flexibility.

I juste have one question about NNEDI3. I understand it doubles the resolution so if a video is 720*576 I get 1440*1152. I have a 1920*1080 TV screen so horizontal resolution is smaller and vertical resolution is bigger. Does it imply a upscale or a downscale or both in that case?

I hope you'll find a solution for Nvidia GPU owners, your're making a great job (I really love the debanding feature ) so I wish you the best of luck to fix this issue.

[vivan]

Originally Posted by Werewolfy

I juste have one question about NNEDI3. I understand it doubles the resolution so if a video is 720*576 I get 1440*1152. I have a 1920*1080 TV screen so horizontal resolution is smaller and vertical resolution is bigger. Does it imply a upscale or a downscale or both in that case?

Both (if NNEDI3 is used). First, video is upscaled with NNEDI3 (720x576 -> 1440x1152), then downscaled (1440x1152 -> 1350x1080 (square pixels) or 1440x1080 (4:3)).
I'm not sure about anamorphic 16:9, however. It should be 1440x1152 -> 1920x1080, means upscaling in one direction and downscaling in the other.

[Werewolfy]

Thanks, I forgot that for a 1080p Tv screen, the resolution is only 1440*1080 for a 4:3 video so after NNEDI3 doubles the resolution to 1440*1152, madVR has only to downscale it to the appropriate resolution with Catmull-Rom or whatever you've chosen.

Yes, there is still anamorphic 16:9 content who needs to be upscaled horizontally and downscaled vertically, I wonder what happens in that particular case, if Jinc (for example) and Catmull-Rom work together.

[pie1394]

@madshi

Confirm the DXVA deinterlacing's performance is restored with 0.87.3! Appreciate for your hard work!

[720x480i60]
0.86.10 : 1.20ms(0.83 ms **)
0.87.1: 1.64 ms
0.87.1b: 1.38 ms
0.87.3: 1.27 ms

**Somehow I no longer can reproduce this number during today's verification... Not quite sure if the CPU's TurboBoost among several frequencies affects the timing measurement's accuracy

[1440x1080i60]
0.86.10 : 1.44 ms
0.87.1: 4.21 ms
0.87.1b: 2.12 ms
0.87.3: 1.38 ms

--
Core i5-3570K + Z77 + dual-ch DDR3-2400 + HD7970@925MHz Catalyst 13.12 (forced VSync ON, Vector-adaptive deinterlace mode) + Win7x64SP1 + MPC-BE 1.3.0.3 + LavFilter 0.60.1 (DXVA2) + madVR FSE (deband,angle-detect, all Jin3AR for up-scale)

[AGKnotUser]

With 87.3 when I try to play a 1.33:1 MKV file in fullscreen mode with MPC-HC the video stops, displays the frame it stops at but the audio continues. Using Windows 8.1 64 bit, MPC-HC 1.7.2. Going back to 86.11 cures the problem.

Edit: I notice this happens only with videos with no top or bottom black bars.

[michkrol]

Originally Posted by AGKnotUser

With 87.3 when I try to play a 1.33:1 MKV file in fullscreen mode with MPC-HC the video stops, displays the frame it stops at but the audio continues. Using Windows 8.1 64 bit, MPC-HC 1.7.2. Going back to 86.11 cures the problem.

Works here
On a more serious note, mkv is a container format, this tells us nothing important. What codec is used, is the video interlaced, are you using software or hardware decoding (which exactly dxva-n/dxva-cb/cuvid/quicksync), what resolution are you scaling to/from?
If you're on MPC-HC select File->Properties and paste the MediaInfo here. Information about operating system, GPU, drivers, player and codecs used might also help.
Are you using any of the new features? Perhaps try restoring settings to default?

[AGKnotUser]

Originally Posted by michkrol

Works here
On a more serious note, mkv is a container format, this tells us nothing important. What codec is used, is the video interlaced, are you using software or hardware decoding (which exactly dxva-n/dxva-cb/cuvid/quicksync), what resolution are you scaling to/from?
If you're on MPC-HC select File->Properties and paste the MediaInfo here. Information about operating system, GPU, drivers, player and codecs used might also help.
Are you using any of the new features? Perhaps try restoring settings to default?

Video: MPEG4 Video (H264) 956x710 (171:127) 23.976fps [Video]
Audio: Dolby AC3 48000Hz stereo 640kbps [Audio]

General
Unique ID : 201361505293360179762353311071887896198 (0x977CD0B57A6965EABFF11649609F2E86)
Complete name : C:\MKV\A Christmas Carol.mkv
Format : Matroska
Format version : Version 4 / Version 2
File size : 1.99 GiB
Duration : 1h 26mn
Overall bit rate : 3 289 Kbps
Encoded date : UTC 2014-01-13 00:48:38
Writing application : mkvmerge v6.7.0 ('Back to the Ground') 64bit built on Jan 8 2014 15:10:52
Writing library : libebml v1.3.0 + libmatroska v1.4.1
Attachment : Yes

Video
ID : 1
Format : AVC
Format/Info : Advanced Video Codec
Format profile : High@L4.1
Format settings, CABAC : Yes
Format settings, ReFrames : 4 frames
Codec ID : V_MPEG4/ISO/AVC
Duration : 1h 26mn
Bit rate : 2 584 Kbps
Width : 956 pixels
Height : 710 pixels
Display aspect ratio : 4:3
Frame rate mode : Constant
Frame rate : 23.976 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Bits/(Pixel*Frame) : 0.159
Stream size : 1.56 GiB (79%)
Writing library : x264 core 140 r2377 1ca7bb9
Encoding settings : cabac=1 / ref=3 / deblock=1:-1:-1 / analyse=0x3:0x113 / me=hex / subme=7 / psy=1 / psy_rd=1.00:0.15 / mixed_ref=1 / me_range=16 / chroma_me=1 / trellis=1 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-3 / threads=12 / lookahead_threads=2 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=3 / b_pyramid=2 / b_adapt=1 / b_bias=0 / direct=1 / weightb=1 / open_gop=0 / weightp=2 / keyint=250 / keyint_min=23 / scenecut=40 / intra_refresh=0 / rc_lookahead=40 / rc=crf / mbtree=1 / crf=20.0 / qcomp=0.60 / qpmin=0 / qpmax=69 / qpstep=4 / ip_ratio=1.40 / aq=1:1.00
Language : English
Default : Yes
Forced : No

Audio
ID : 2
Format : AC-3
Format/Info : Audio Coding 3
Mode extension : CM (complete main)
Format settings, Endianness : Big
Codec ID : A_AC3
Duration : 1h 26mn
Bit rate mode : Constant
Bit rate : 640 Kbps
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 48.0 KHz
Bit depth : 16 bits
Compression mode : Lossy
Stream size : 396 MiB (19%)
Language : English
Default : Yes
Forced : No

Using cuvid with nvidia GT 220 card:

NVIDIA System Information report created on: 01/27/2014 11:29:28
System name: USER-PC

[Display]
Operating System: Windows 8.1 Pro with Media Center, 64-bit
DirectX version: 11.0
GPU processor: GeForce GT 220
Driver version: 327.02
Direct3D API version: 10.1
CUDA Cores: 48
Core clock: 615 MHz
Shader clock: 1335 MHz
Memory data rate: 1580 MHz
Memory interface: 128-bit
Total available graphics memory: 4095 MB
Dedicated video memory: 1024 MB DDR3
System video memory: 0 MB
Shared system memory: 3071 MB
Video BIOS version: 70.16.3C.00.05
IRQ: 16
Bus: PCI Express x16 Gen2
Device Id: 10DE 0A20 19101462

[Components]

easyDaemonAPIU64.DLL 1.10.8.0 NVIDIA Update Components
WLMerger.exe 1.10.8.0 NVIDIA Update Components
daemonu.exe 1.10.8.0 NVIDIA Update Components
ComUpdatus.exe 1.10.8.0 NVIDIA Update Components
NvUpdtr.dll 1.10.8.0 NVIDIA Update Components
NvUpdt.dll 1.10.8.0 NVIDIA Update Components
nvui.dll 8.17.13.2702 NVIDIA User Experience Driver Component
nvxdsync.exe 8.17.13.2702 NVIDIA User Experience Driver Component
nvxdplcy.dll 8.17.13.2702 NVIDIA User Experience Driver Component
nvxdbat.dll 8.17.13.2702 NVIDIA User Experience Driver Component
nvxdapix.dll 8.17.13.2702 NVIDIA User Experience Driver Component
NVCPL.DLL 8.17.13.2702 NVIDIA User Experience Driver Component
nvCplUIR.dll 7.4.740.0 NVIDIA Control Panel
nvCplUI.exe 7.4.740.0 NVIDIA Control Panel
nvWSSR.dll 6.14.13.2702 NVIDIA Workstation Server
nvWSS.dll 6.14.13.2702 NVIDIA Workstation Server
nvViTvSR.dll 6.14.13.2702 NVIDIA Video Server
nvViTvS.dll 6.14.13.2702 NVIDIA Video Server
nvDispSR.dll 6.14.13.2702 NVIDIA Display Server
NVMCTRAY.DLL 8.17.13.2702 NVIDIA Media Center Library
nvDispS.dll 6.14.13.2702 NVIDIA Display Server
PhysX 09.10.0514 NVIDIA PhysX
NVCUDA.DLL 8.17.13.2702 NVIDIA CUDA 5.5.1 driver
nvGameSR.dll 6.14.13.2702 NVIDIA 3D Settings Server
nvGameS.dll 6.14.13.2702 NVIDIA 3D Settings Server

Resetting render-er to defaults had no effect.

Hope this helps.

[mindbomb]

are you using error diffusion or image doubling?

[djfred93]

Originally Posted by mimi123

Very thanks for your reply, the tread you tell me has been very useful.

I read these tread and I understand there is minus to none improvement on image quality, at least for the normal BD content, playing at 10 bits, is my understanding correct?

Anyway, can you tell me a good card, or some, to get the full potential of Madvr?

The only improvement is deep color who is a smooth gradation.
I think that your projector works in 12bit but can only display 8 bit. Madvr works in 16bit but can only output 8bit.

Maybe the next physical disk will do more than 8bit.
If you want to use OpenCL, you have to choose an ati card until madshi resolve all the problem with the Nvidia card.

My card is a ati radeon hd 5770 OC and i can't do all the OpenCL stuff. I can do Jinc3 AR/Jinc3 AR/ Catmull-Rom AR+SL with debanding. I think a ATI Radeon R9 280X is good. You can compare GPU here.

[AGKnotUser]

Originally Posted by mindbomb

are you using error diffusion or image doubling?

I had in madVR settings image upscaling and image downscaling set to DXVA2. If I change the settings to Bilinear the video plays.

[vomanci]

I can confirm setting image upscaling and/or downscaling to DXVA2 freezes the image whether it's up/downscaled. On an Nvidia 8200 onboard (that's right; I use bilinear and works )

[antonyfrn]

No joy with madVR v0.87.3 + NNEDI3 + MPC-HC 1.7.2 + NVidia 334.67 Beta still a black screen.

[madshi]

Originally Posted by Deim0s

My configuration in this message.

Here I post screenshots of memory consumption. v0.87.* consumes less RAM.

Your quote of my question does not include the ">" arrow, so I can't look back at my complete post you're refering to. Please always keep the quote logic working, otherwise you're making it hard for me to look back at our old discussion.

Can you please check whether the problem with 4K playback still occurs in v0.87.4?

Originally Posted by mindbomb

I have a question about image doubling(which also applies to the scaling factor requirement). Does doubling mean doubling the total amount of pixels, or doubling both the height and width (quadrupling the total pixels)?

Image doubling looks at X and Y separately. Meaning that the image could be doubled in X direction, but not in Y. Or in both. Or only in Y but not in X. It all depends on the X and Y scaling factors and your settings.

Originally Posted by AngelGraves13

Is the problem with dragging the screen to another monitor fixed? It would stop playback and you'd have to open the file again with the player on the second screen for it to play, or it would just stay stopped no matter matter. Pausing and Playing did nothing.

I don't know. Why don't you give it a try and report back?

Originally Posted by Qotscha

I think the performance of error diffusion is quite poor on my computer (Radeon HD 6850 with Catalyst 13.12). Here is a little comparison:

720p50 to 1080p50:

• Jinc 4 + AR (Image + chroma), medium deband, random dithering: 91% GPU load, no dropped frames
• Bilinear (Image + chroma), no deband, error diffusion: ~90% GPU load, a lot of dropped frames (result is about the same with any scaling algorithm)

That is rather weird. It shouldn't be that slow. With D3D9 <-> OpenCL interop taken out of the picture, my HD7770 can do error diffusion on 1080p images with > 300fps. Measured in madVR it costs about 12ms rendering time, when talking about 1080p24 playback.

Originally Posted by Asmodian

I the process of making this list I found something very odd.

Queues 13/13/8 works (good picture) but 13/12/8 gives black screen. 14/13/8 and 15/13/8 give black but 16/13/8 gives a good picture. This seems to be true with LL on or off, 24/22/8
LL Off (2295 MB) works but 23/22/8 doesn't, 23/20/8 does. Very odd; and these exact queue sizes seem to change. Sorry for the lack of good data.

That really is odd. I rather think it's random. Try again with v0.87.4. And if the problem still occurs, please double check with v0.86.11 to make sure it's a new bug and not an old one. I'm only interested right now in bugs which were introduced in v0.87.x.

Originally Posted by James Freeman

Can you please add [...]

Originally Posted by James Freeman

Have you ever thought about adding [...]

Do you have plans to further add [...]

I'm not accepting feature requests, anymore, until I've reached v1.0.

If you have problems with forced film mode, you can provide me with samples, so that I can improve forced film mode, when I find time to look at it again (probably not too soon).

Originally Posted by sdancer75

1) My app is not free, so can I ask for a free license or do I have to pay license fee ?

I think we should better discuss this via email. Please contact me via madshi (at) gmail (dot) com with a description of your app (maybe a homepage link).

Originally Posted by sdancer75

2) Still I can not make it to work : Here the code I use. I can see the madvr window inside my CView pane, and something that is moving (like a video playback) with flickering.

I don't have that much experience with media player development. But if it helps here's how madTPG initializes madVR for playback:

  if (VideoWindow.put_Owner(MadVRPanel.Handle) = 0) and
     (FilterGraph.AddFilter(DummySource, 'source') = 0) and
     (FilterGraph.AddFilter(madVR,       'madVR' ) = 0) then
  begin
    DummySource.EnumPins(enumPins);
    if enumPins <> nil then
      enumPins.Next(1, outPin, @c1);
    madVR.EnumPins(enumPins);
    if enumPins <> nil then
      enumPins.Next(1, inPin, @c1);
    if (inPin <> nil) and (outPin <> nil) and
       (FilterGraph.ConnectDirect(outPin, inPin, nil) = 0) and
       (VideoWindow.put_MessageDrain(MadVRPanel.Handle) = 0) and
       (VideoWindow.SetWindowPosition(0, 0, MadVRPanel.Width, MadVRPanel.Height) = 0) and
       (VideoWindow.get_WindowStyle(style) = 0) and
       (VideoWindow.put_WindowStyle((style and (not WS_CAPTION)) or WS_CHILD) = 0) and
       (VideoWindow.put_AutoShow(true) = 0) and
       ((FilterGraph as IMediaControl).Pause = 0) then

If that doesn't help, maybe you could look at the MPC-HC code to find out how it makes madVR work?

Originally Posted by mimi123

Anyway, can you tell me a good card, or some, to get the full potential of Madvr?

That's hard to so. It all depends on which madVR features you want to use, and at which resolutions and framerates. Some algorithms are very hard on the GPU. But the question is if you have to use them. I'd suggest that you first play with the various madVR settings with your current GPU to find out which algorithms you want to use. Then maybe we can recommend a GPU that can do what you need.

Originally Posted by leeperry

So I cannot get NNEDI to upscale 360*288 to 1440*576?

Might be possible, depending on the exact target resolution. Look at all the options in the image doubling settings page, use your head, and you should be able to figure it out yourself. Just keep in mind that X and Y are handled completely separately by madVR.

What you cannot do is tell madVR to do exactly hard coded 4x in X direction and 2x in Y direction. You can only tell madVR whether to do 2x or 4x (or no change) depending on which scaling factor is needed to achieve the target resolution. I think my solution makes more sense because not every 360x288. My settings should make sense with all kinds of source resolutions, not just one.

Originally Posted by leeperry

And you recently said that NNEDI was kinda overkill for chroma, but for clarity sakes's what did you use for chroma in this screenshots comparison please?

Don't remember, something cheap, I think.

Originally Posted by Werewolfy

But the profiles are great, more flexible than I have imagined! I was a little bit afraid at the beginning, it seemed a little bit complicated but after some testings, it's not so hard. I'm sure this is the price to pay for this high flexibility.

What I found important was to not make the settings dialog more complicated - as long as you don't activate profiles. If you do activate them, you have to live with a bit more complicated setup. I'm sure it would have been possible to make the profiles slightly less complicated, but then, as you say, it wouldn't have been as flexible then...

Originally Posted by Werewolfy

I juste have one question about NNEDI3. I understand it doubles the resolution so if a video is 720*576 I get 1440*1152. I have a 1920*1080 TV screen so horizontal resolution is smaller and vertical resolution is bigger. Does it imply a upscale or a downscale or both in that case?

Originally Posted by vivan

Both (if NNEDI3 is used). First, video is upscaled with NNEDI3 (720x576 -> 1440x1152), then downscaled (1440x1152 -> 1350x1080 (square pixels) or 1440x1080 (4:3)).
I'm not sure about anamorphic 16:9, however. It should be 1440x1152 -> 1920x1080, means upscaling in one direction and downscaling in the other.

Originally Posted by Werewolfy

Thanks, I forgot that for a 1080p Tv screen, the resolution is only 1440*1080 for a 4:3 video so after NNEDI3 doubles the resolution to 1440*1152, madVR has only to downscale it to the appropriate resolution with Catmull-Rom or whatever you've chosen.

Yes, there is still anamorphic 16:9 content who needs to be upscaled horizontally and downscaled vertically, I wonder what happens in that particular case, if Jinc (for example) and Catmull-Rom work together.

NNEDI3 image doubling is completely separate from the other scaling. Think of NNEDI3 as manipulating the source video into a different resolution. The other scaling algorithms don't care what the original source video res was, they just look at the resolution of the image doubling output, and go from there.

Jinc currently doesn't support downscaling, and it always handles both X and Y. So if either X or Y need to be downscaled, Jinc can't be used (and should silently be disabled by madVR).

Originally Posted by AGKnotUser

I had in madVR settings image upscaling and image downscaling set to DXVA2. If I change the settings to Bilinear the video plays.

Originally Posted by vomanci

I can confirm setting image upscaling and/or downscaling to DXVA2 freezes the image whether it's up/downscaled. On an Nvidia 8200 onboard (that's right; I use bilinear and works )

I can confirm the problem. Will look at that for v0.87.5. But not today, anymore...

[hannes69]

Today I played around with my AMD 4550 GPU to test all the new features of madvr. Debanding and the profile stuff is working great.
Just for info concerning the OpenCL stuff: I don´t use it (no need for the features in my setup) and it doesn´t seem to work properly. That´s no problem because this ancient video card could certainly not handle this stuff, even if the implementation of these functions would be correct.
NNEDI3 chroma upscaling: GPU load rises, but the load is independent of the number of neurons ...
image doubling: GPU load stays the same, it doesn´t matter which settings are made (tested with lowres video)
OpenCL DXVA NV12 surfaces: no effect when using DXVA native or copyback decoder (GPU load the same)
OpenCL error diffusion: GPU load stays the same.

I doubt that any of the OpenCL stuff is working with my good old card. GPU-Z reports full OpenCL support for this GPU.
Maybe the old generation cards HD2000 - HD4000 series with legacy drivers don´t support all the OpenCL features.

All no problem for me, just wanted to notify.

Beside that the recent version of madvr works stable for me.

General idea: it would be really nice in a mid-term future that all the computing of video could be freely allocated to CPU/GPU. My CPU is bored at the moment with just decoding the video. I thought the idea of OpenCL was to use CPU and GPU interchangeable for all purposes...
Of course it is too time consuming to make code implementations for GPU and CPU for every little feature, but isn´t OpenCL comparable to a sort of cross-compiler (have no better comparison) so code can rather easily be converted?
Sorry for my non-expert point of view, just interested.

[madshi]

madVR v0.87.4 released

http://madshi.net/madVR.zip

* workaround added: NNEDI3 upscaling failed/froze with newer NVidia GPUs
* fixed: NNEDI3 chroma upscaling produced wrong colors with 10bit sources
* got rid of some unnecessary texture sharing

All OpenCL features (except DXVA NV12 processing) should now also work with newer NVidia GPUs. However, the current official NVidia drivers are broken and always produce zero/black output when used with D3D9 interop inside of madVR. I've tried to make it work somehow, but failed. You have to downgrade to NVidia 327.23 drivers (or older) to make the OpenCL features work with NVidia. Would be nice if someone could report the OpenCL driver bug with the current official drivers to NVidia.

Maybe the 4K downscaling performance issue could be fixed, too, but I don't know. Give it a try and let me know.

[antonyfrn]

Originally Posted by madshi

madVR v0.87.4 released

http://madshi.net/madVR.zip

* workaround added: NNEDI3 upscaling failed/froze with newer NVidia GPUs
* fixed: NNEDI3 chroma upscaling produced wrong colors with 10bit sources
* got rid of some unnecessary texture sharing

All OpenCL features (except DXVA NV12 processing) should now also work with newer NVidia GPUs. However, the current official NVidia drivers are broken and always produce zero/black output when used with D3D9 interop inside of madVR. I've tried to make it work somehow, but failed. You have to downgrade to NVidia 327.23 drivers (or older) to make the OpenCL features work with NVidia. Would be nice if someone could report the OpenCL driver bug with the current official drivers to NVidia.

Maybe the 4K downscaling performance issue could be fixed, too, but I don't know. Give it a try and let me know.

Sent a bug report to NVidia ill try poke one of the reps over at guru3d