Media Player .NET (MPDN) - D3D HQ GPU Video Renderer [v2.49.0/v1.31.0 27 Dec 2018] - Page 103

Zachs · 8th May 2015, 06:48

Yeah looks like a bug in NV's newer drivers.

ryrynz · 8th May 2015, 10:39

Quote:

Originally Posted by Zachs

Yeah looks like a bug in NV's newer drivers.

Shame, it sped up NNEDI3 16 taps on my HD4000 by over 20%, getting similar performance gains on Nvidia would deliver better performance than madVR (at least on my card)
Still can't recommend it for Intel though it's almost twice as slow as the shader version 'prefer vector & small code'. I would quite like to see what NNEDI3 performance is like on Skylake's GT4e.

Zachs · 8th May 2015, 12:18

Ah it was never meant to speed up NV hardware!
Are you saying MPDN's OpenCL NNEDI3 is slower than madVR on your card? That's not the case on my 560GTX...

Btw I've uploaded the changes to GitHub, reverting only NV's code path.

I'd like to get some feedback on whether AMD users are gaining/losing performance from the changes at all. As for Intel, it is ~20% faster.

Anime Viewer · 8th May 2015, 13:23

Quote:

Originally Posted by Zachs

Hmm, might have encountered an NVIDIA OpenCL compiler bug there. I don't have access to my 560 GTX at the moment so I think I'll revert the changes until I can fully test it on NV's latest drivers.

EDIT: I tested this on an NV 8400GS with driver v341.44 and it's fully working... What version are you running by the way?

I'm running v350.12 (which uses OpenCL 1.2). Your v341.44 runs OpenCL 1.1 correct? There probably is a good chance that the problem is related to how Nvidia's OpenCL 1.2 and 1.1 differ as opposed to something else in the driver. I wouldn't be surprised if the other people who are encountering the same problem as me are running Nvidia's OpenCL 1.2 as well.

Running it with my Intel 4000 gpu NNEDI3 was still unusable (running at 246ms by itself, and 300ms as part of a script chain with debanding and superchromares).

Zachs · 8th May 2015, 13:44

Well it won't do 720p on my Intel HD4600 but for lower res upscaling it works just fine. It's those lower res sources that arguably need NNEDI3 more than the higher res ones.

flacattack · 9th May 2015, 02:35

Starting with the newer builds (around April 30th), I get a blank screen when trying to use SuperRes with my r9 290x. The other scripts work perfect.

Code:

InvalidArgument=Value of '0' is not valid for 'SelectedIndex'.
Parameter name: SelectedIndex

Shiandow · 9th May 2015, 03:03

You probably need to reset SuperRes's settings. There should be a "SuperRes.config" file somewhere, you can just delete that. Or if it's in a script chain you can just remove it and then add it back again.

flacattack · 9th May 2015, 04:26

Worked like a charm, thank you Shiandow!

ryrynz · 9th May 2015, 05:16

Hey Zach, feature request! Reset to defaults!

Zachs · 9th May 2015, 12:50

That was always in the plan but it's not that easy. There are a few 'reset to defaults' that I'd like to see - within each render script / player extension, and a global one that resets all of MPDN's config to default as well as one that wipes both MPDN, render scripts and player extensions config. Haven't figured out a way to do that yet...

Zachs · 9th May 2015, 12:52

Just a quick heads up. I've added OpenCL NNEDI3 as a prescaler of SuperRes.
You'll need to delete "%localappdata%\MediaPlayerDotNet\RenderScripts.x??\SuperRes.config" again for this option to appear after you've downloaded the new scripts from GitHub.

nekromantik · 9th May 2015, 14:25

To upscale anything to 4k is this chain good enough? if not any suggestions.

NNEDI -> SuperChromaRes + NNEDI

Zachs · 10th May 2015, 04:18

I'm not sure I understand your script chain at all.
Specifically what does NNEDI -> SuperChromaRes + NNEDI mean?

nekromantik · 10th May 2015, 11:25

Quote:

Originally Posted by Zachs

I'm not sure I understand your script chain at all.
Specifically what does NNEDI -> SuperChromaRes + NNEDI mean?

Sorry ignore the + NNEDI.

It goes NNEDI image doubling and then SuperChromaRes upscaling.
Or is SuperRes with NNEDI and then SuperChromaRes better?

Zachs · 10th May 2015, 12:27

You have to scale chroma before you can use any render script upscalers. SuperChromaRes has to come before NNEDI in your case. I usually do deband right after chroma upscaling, then followed by NNEDI3 with SuperRes (very low strength). Even when running NNEDI3 with 64 neurons on my gtx680, I still find SuperRes to increase image quality both in sharpening and reducing aliasing.

nekromantik · 10th May 2015, 16:26

Quote:

Originally Posted by Zachs

You have to scale chroma before you can use any render script upscalers. SuperChromaRes has to come before NNEDI in your case. I usually do deband right after chroma upscaling, then followed by NNEDI3 with SuperRes (very low strength). Even when running NNEDI3 with 64 neurons on my gtx680, I still find SuperRes to increase image quality both in sharpening and reducing aliasing.

Thanks will give that a go.

Shiandow · 11th May 2015, 00:20

Hi there! I made some changes to the renderscripts again, based on the feedback on the MadVR thread and some of things madshi noticed while implementing them.

A short summary:
- Tried to improve the quality of the debanding, it should have less artifacts around edges and might be slightly faster.
- SuperChromaRes now works in rgb, this should improve quality and be slightly faster, although the anti-ringing still needs some work, it'll work correctly, but the performance isn't too good at the moment.
- Some minor performance tweaks / fixes to SuperRes and SuperChromaRes.
- Better defaults for SuperRes and SuperChromaRes.

Zachs · 11th May 2015, 00:50

Hmm just found out madVR is in violation of the LGPL license (section 4d and/or 4e) - i.e. including the source isn't enough. Section 4d requires that the user be able to modify the LGPL source and have the app run the modified version and/or link to it. MadVR is also in violation of section 4c which requires that it displays a proper copyright notice. The ancillaries required to run those shaders have also been converted to closed source, which is also in violation of LGPL.

On a similar note, it would appear that its NNEDI3 implementation is also in violation of LGPL, as changing the included source in the zip file won't affect madVR's implementation. The original NNEDI3 OpenCL source code by SEt is LGPL and is in full compliance with LGPL.

Anima123 · 11th May 2015, 04:03

I hate to mention it, especially in such a circumstance, but, using NEDI with SuperRes via madVR does not suffer from rendering time increasing for my laptop.

Zachs · 11th May 2015, 05:29

Seeing as no one else with Optimus has seen this problem, could it be something to do with your .NET installation?

8th May 2015, 12:18	#2043 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	Ah it was never meant to speed up NV hardware! Are you saying MPDN's OpenCL NNEDI3 is slower than madVR on your card? That's not the case on my 560GTX... Btw I've uploaded the changes to GitHub, reverting only NV's code path. I'd like to get some feedback on whether AMD users are gaining/losing performance from the changes at all. As for Intel, it is ~20% faster. Last edited by Zachs; 8th May 2015 at 13:15.

9th May 2015, 02:35	#2046 \| Link
flacattack Registered User Join Date: Jan 2013 Posts: 12	Starting with the newer builds (around April 30th), I get a blank screen when trying to use SuperRes with my r9 290x. The other scripts work perfect. Code: InvalidArgument=Value of '0' is not valid for 'SelectedIndex'. Parameter name: SelectedIndex

11th May 2015, 00:50	#2058 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	Hmm just found out madVR is in violation of the LGPL license (section 4d and/or 4e) - i.e. including the source isn't enough. Section 4d requires that the user be able to modify the LGPL source and have the app run the modified version and/or link to it. MadVR is also in violation of section 4c which requires that it displays a proper copyright notice. The ancillaries required to run those shaders have also been converted to closed source, which is also in violation of LGPL. On a similar note, it would appear that its NNEDI3 implementation is also in violation of LGPL, as changing the included source in the zip file won't affect madVR's implementation. The original NNEDI3 OpenCL source code by SEt is LGPL and is in full compliance with LGPL. Last edited by Zachs; 11th May 2015 at 02:10.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

8th May 2015, 06:48	#2041 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	Yeah looks like a bug in NV's newer drivers.

8th May 2015, 13:44	#2045 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	Well it won't do 720p on my Intel HD4600 but for lower res upscaling it works just fine. It's those lower res sources that arguably need NNEDI3 more than the higher res ones.

9th May 2015, 03:03	#2047 \| Link
Shiandow Registered User Join Date: Dec 2013 Posts: 753	You probably need to reset SuperRes's settings. There should be a "SuperRes.config" file somewhere, you can just delete that. Or if it's in a script chain you can just remove it and then add it back again.

9th May 2015, 04:26	#2048 \| Link
flacattack Registered User Join Date: Jan 2013 Posts: 12	Worked like a charm, thank you Shiandow!

9th May 2015, 05:16	#2049 \| Link
ryrynz Registered User Join Date: Mar 2009 Posts: 3,650	Hey Zach, feature request! Reset to defaults!

9th May 2015, 12:50	#2050 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	That was always in the plan but it's not that easy. There are a few 'reset to defaults' that I'd like to see - within each render script / player extension, and a global one that resets all of MPDN's config to default as well as one that wipes both MPDN, render scripts and player extensions config. Haven't figured out a way to do that yet...

9th May 2015, 12:52	#2051 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	Just a quick heads up. I've added OpenCL NNEDI3 as a prescaler of SuperRes. You'll need to delete "%localappdata%\MediaPlayerDotNet\RenderScripts.x??\SuperRes.config" again for this option to appear after you've downloaded the new scripts from GitHub.

9th May 2015, 14:25	#2052 \| Link
nekromantik Registered User Join Date: Feb 2013 Posts: 96	To upscale anything to 4k is this chain good enough? if not any suggestions. NNEDI -> SuperChromaRes + NNEDI

10th May 2015, 04:18	#2053 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	I'm not sure I understand your script chain at all. Specifically what does NNEDI -> SuperChromaRes + NNEDI mean?

10th May 2015, 12:27	#2055 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	You have to scale chroma before you can use any render script upscalers. SuperChromaRes has to come before NNEDI in your case. I usually do deband right after chroma upscaling, then followed by NNEDI3 with SuperRes (very low strength). Even when running NNEDI3 with 64 neurons on my gtx680, I still find SuperRes to increase image quality both in sharpening and reducing aliasing.

11th May 2015, 00:20	#2057 \| Link
Shiandow Registered User Join Date: Dec 2013 Posts: 753	Hi there! I made some changes to the renderscripts again, based on the feedback on the MadVR thread and some of things madshi noticed while implementing them. A short summary: - Tried to improve the quality of the debanding, it should have less artifacts around edges and might be slightly faster. - SuperChromaRes now works in rgb, this should improve quality and be slightly faster, although the anti-ringing still needs some work, it'll work correctly, but the performance isn't too good at the moment. - Some minor performance tweaks / fixes to SuperRes and SuperChromaRes. - Better defaults for SuperRes and SuperChromaRes.

11th May 2015, 04:03	#2059 \| Link
Anima123 Registered User Join Date: Jun 2005 Posts: 504	I hate to mention it, especially in such a circumstance, but, using NEDI with SuperRes via madVR does not suffer from rendering time increasing for my laptop.

11th May 2015, 05:29	#2060 \| Link
Zachs Suptitle, MediaPlayer.NET Join Date: Nov 2001 Posts: 1,721	Seeing as no one else with Optimus has seen this problem, could it be something to do with your .NET installation?