View Issue Details

IDProjectCategoryView StatusLast Update
0000399madVRbugpublic2018-01-14 15:31
Reporterdan Assigned Tomadshi  
PrioritynormalSeverityminorReproducibilityalways
Status closedResolutionunable to reproduce 
OSWindows 10 x64OS Version10.0.10586 
Summary0000399: Adaptive Sharpen makes image blocky
DescriptionActivating "Adaptive Sharpen" (under either "Image enhancements" or "Upscale refinement") causes the image to become blocky (or screwed up). See ss for example.

Cheers!
Steps To ReproduceJust activate Adaptive Sharpen.
Additional Information> Happens on my desktop PC. This is a new build and I have not run old madvr versions or graphics drivers in it.
> Does not happen on my laptop using either nvidia or intel gpu (tested both).

More info on my desktop pc:
Running on Intel's onboard graphics (HD 530) of an i5-6500 CPU (Skylake). I suspect this to be driver related (either a bug or an incompatibility with something Intel intentionally changed) since it doesn't happen on my laptop (with older gen cpu).
In the "graphics control panel" (driver options) from Intel I have all possible settings set to "Use application settings" and disabled specific applications optimizations (though activating it didn't change anything).
TagsNo tags attached.
madVR Version0.90.18
Media Player (with version info)MPC-HC x64 1.7.10
Splitter (with version info)LAV 0.68.1
Decoder (with version info)LAV 0.68.1
DecodingSoftware
Deinterlacingnone (progressive)
DXVA2 Scaling Activeno
Aero / Desktop CompositionOn
Problem occurs with modeall modes
GPU ManufacturerIntel
GPU ModelIntel HD Graphics 530
GPU Driver Version20.19.15.4424

Activities

dan

2016-05-10 22:36

reporter  

madvrbug.png (2,526,280 bytes)

madshi

2016-05-11 09:13

administrator   ~0001332

Does that occur with all videos? Could be a GPU driver bug, but I'm not sure. Have you tried with madVR default settings (+ AdaptiveSharpen, of course)?

dan

2016-05-11 20:12

reporter  

mad1.png (569,362 bytes)

dan

2016-05-11 20:12

reporter  

mad2.png (1,557,953 bytes)

dan

2016-05-11 20:13

reporter  

mad3.png (2,091,460 bytes)

dan

2016-05-11 20:13

reporter   ~0001333

> Does that occur with all videos?
From what I can tell, yes. I've attached some more screenshots, with different types of videos formats.

> Have you tried with madVR default settings (+ AdaptiveSharpen, of course)?
Yes. For the new screenshots, they were all taken using default settings.
Do notice that when just using Adaptive Sharpen under "Image enhancements" screws up a little less than when using it as "Upscale Refinement". Still bad though.

A few more observations:
- I tried using another GPU driver (latest beta 20.19.15.4444), but it didn't help.
- I rolled back madVR to 0.90.17 to avoid the bug where image doubling is activated on downscale (I didn't report this one because I've seen it reported on doom9 forum).
- I noticed my rendering stats skyrocketing with Adaptive Sharpen on and videos being played (like, from <2ms to >130ms). See "mad3.png".

Let me know if and how I can generate any more useful data.

Cheers!

madshi

2016-05-11 20:48

administrator   ~0001334

Did it ever work correctly? If so, what changed?

dan

2016-05-11 21:19

reporter   ~0001335

Last edited: 2016-05-11 21:22

Never worked correctly on this computer, I built it just this week.

I've been using Adaptive Sharpen since it was introduced on my previous rig and laptops without issues, however.

This is why I suspect it to be a Skylake specific incompatibility/problem.

edit: I don't have any other Skylake pcs to test it, so maybe you could ask for someone (i.e. on doom9 forum or something) to test it (using Intel graphics).

madshi

2016-05-11 23:16

administrator   ~0001336

That would be my best guess, too. Do any of the other shaders produce problems? Maybe you could report this to Intel customer support? I don't know if they have such a thing, though.

dan

2016-05-12 00:30

reporter   ~0001337

No other shaders produce problems.

Apparently they have a forum for developers to report issues:
https://software.intel.com/en-us/forums/graphics-driver-bug-reporting

And another for customers:
https://communities.intel.com/community/tech/graphics

Seeing how slow they are to fix anything (even major problems reported by dozens of users) I just don't think that will be of much help.
I've always had all sorts of problems with Intel graphics on my laptops, and now on this desktop too.
One such problem is that madVR with Direct3D 11 is just unusable across all my devices with Intel graphics (get all sorts of hiccups, like out of order frames, especially when seeking, and black screens after switching files without leaving fullscreen). I guess they just can't make decent drivers.

So, even if this is Intel's fault, I doubt they'll fix it, especially not within a year. You'd have to work around it, but then, I wonder if it's even worth the hassle :/

Thanks for the attention though :D
If you ever get the chance to test a Skylake device, you might just figure out some silly way to fix these issues.

Cheers!

madshi

2016-05-12 08:54

administrator   ~0001338

Yeah, I won't be able to do much without being able to reproduce these issues some how. My Intel HD4000 doesn't show these issues, don't have anything newer atm, unfortunately...

dan

2016-05-16 03:59

reporter   ~0001340

Last edited: 2016-05-16 04:02

I have some new info!

Through some testing, I found out that up to version 0.89.12 this bug does not occur. From 0.89.13 onward it does.

Maybe Intel is at fault, but if you figure out what changed between those versions you may be able to work around it.

I'm willing to do testing if you need it

madshi

2016-05-16 09:46

administrator   ~0001341

In 0.89.13 I updated AdaptiveSharpen to the latest version. Seems that the Intel GPU driver likes the old version but not the new one. No other GPU seems to have problems with the new version, though.

You could double check by testing the AdaptiveSharpen HLSL shaders directly in MPC-HC as custom shaders, using EVR.

dan

2016-05-18 04:30

reporter   ~0001345

Last edited: 2016-05-18 04:33

>You could double check by testing the AdaptiveSharpen HLSL shaders directly in MPC-HC as custom shaders, using EVR.

It renders a completely green screen using it like that (EVR-CP with Adaptive Sharpen HLSL shaders). If I just add "pass one" it doesn't screw it up, but when I add pass two it does.
I wonder if it's worth it contacting bacondither about this.

On his Doom9 thread, he does mention "If the shader passes are added in the wrong order or the alpha channel is truncated, it will return a green screen."
Just in case, I tried adding in both orders and get the same result. So maybe something related to the alpha channel? (whatever that is)

madshi

2016-05-18 09:01

administrator   ~0001346

You may have to switch EVR into high precision / floating point mode to make it work.

dan

2016-05-19 04:17

reporter   ~0001347

Last edited: 2016-05-19 04:20

>You may have to switch EVR into high precision / floating point mode to make it work.

Now I get a black screen with blinking artifacts. I guess it really is on Adaptive Sharpen's side.

edit: I wouldn't even mind if you want to run some tests yourself on my machine via remote desktop or something. Let me know.

madshi

2016-05-19 09:26

administrator   ~0001348

Oh well.

Since this is *probably* a GPU driver problem, I don't think it makes sense for me to spend time trying to debug it via remote desktop. I don't really have that much resources for madVR development. I need to spend the resources wisely.

Issue History

Date Modified Username Field Change
2016-05-10 22:36 dan New Issue
2016-05-10 22:36 dan File Added: madvrbug.png
2016-05-11 09:13 madshi Note Added: 0001332
2016-05-11 20:12 dan File Added: mad1.png
2016-05-11 20:12 dan File Added: mad2.png
2016-05-11 20:13 dan File Added: mad3.png
2016-05-11 20:13 dan Note Added: 0001333
2016-05-11 20:48 madshi Note Added: 0001334
2016-05-11 21:19 dan Note Added: 0001335
2016-05-11 21:22 dan Note Edited: 0001335
2016-05-11 23:16 madshi Note Added: 0001336
2016-05-12 00:30 dan Note Added: 0001337
2016-05-12 08:54 madshi Note Added: 0001338
2016-05-16 03:59 dan Note Added: 0001340
2016-05-16 04:00 dan Note Edited: 0001340
2016-05-16 04:02 dan Note Edited: 0001340
2016-05-16 09:46 madshi Note Added: 0001341
2016-05-18 04:30 dan Note Added: 0001345
2016-05-18 04:33 dan Note Edited: 0001345
2016-05-18 09:01 madshi Note Added: 0001346
2016-05-19 04:17 dan Note Added: 0001347
2016-05-19 04:20 dan Note Edited: 0001347
2016-05-19 09:26 madshi Note Added: 0001348
2018-01-14 15:31 madshi Status new => closed
2018-01-14 15:31 madshi Assigned To => madshi
2018-01-14 15:31 madshi Resolution open => unable to reproduce