Shadows visual artifacts


I’m getting strange visual artifacts with the standard shadow demos (TestDirectionalLightShadow & TestPssmShadow).
I am using the versions “SDK stable”, “SDK nightly”, “SVN jme3 head”.
Here is how it looks (ground texture disabled using the M key so that you can see the error better):

My Hardware (with latest drivers):

Primary Adapter
Graphics Card Manufacturer Powered by AMD
Graphics Chipset ATI Radeon HD 4800 Series
Device ID 9460
Vendor 1002

Subsystem ID 2282
Subsystem Vendor ID 1787

Graphics Bus Capability PCI Express 2.0
Maximum Bus Setting PCI Express 2.0 x16

BIOS Version
BIOS Part Number 113-BA900
BIOS Date 2009/06/12

Memory Size 1024 MB
Memory Type GDDR5

Core Clock in MHz 900 MHz
Memory Clock in MHz 1000 MHz
Total Memory Bandwidth in GByte/s 128,0 GByte/s

Software information:

Driver Packaging Version 8.861-110524a-119602C-ATI
Catalyst™ Version 11.6
Provider AMD.
2D Driver Version
2D Driver File Path System/CurrentControlSet/Control/Video/{163DEFC1-498B-4F33-96AA-0CEEBC05F76A}/0000
Direct3D Version
OpenGL Version
Catalyst™ Control Center Version 2013.0424.1155.19491
AIW/VIVO WDM Driver Version
AIW/VIVO WDM SP Driver Version

And by the way…
Your forum search is broken somehow… Here is how it looks in my Firefox browser:

Yet another ATI quirk…
I would normally recommend to update your card drivers, but ATI is the only vendor making things worst with new driver versions…
so…rollback to previous version…?
I’ll look into it next week, but not sure I’ll be able to do something I’m sorry.

Thanks for the try,

Just some thoughts:

  1. Maybe we could have a standard shadow mapping as a workaround?
  2. I heard that NVIDIA is not so good either, because they give poor Linux support (Linus said that).
  3. Well, GPUGems is by NVIDIA. Maybe there is a version of the algorithm that works better with ATI cards?
  4. Consoles will use AMD hardware and some users will propably have their GCN cards soon. I don’t see a way around this vendor for the future.

I’ve had similar problems during my master thesis - had to write a code switch for the big two card vendors, because their GLSL code differed in some details…

The thing is…it’s not only ati versus nvidia, it’s some combinations of ati cards and ati drivers on some os versus everything else.
That’s why it’s hard to debug.
This code was developed on an ati card (my former one).
I still got it somewhere at worst I’ll re-install it and see if it fail…
But last time i checked it was working.
But I had to keep drivers from febuary 2011, or any latest drivers had other issues…

@Ogli said: Thanks for the try, 2) I heard that NVIDIA is not so good either, because they give poor Linux support (Linus said that).

They have closed source binary only drivers, you know like every windows & OSX driver :slight_smile: But if you are OK with that on linux then they are pretty good I think. On par with windows drivers when it comes to OpenGL version (not like Apple that are lagging several years).

@Ogli said: 2) I heard that NVIDIA is not so good either, because they give poor Linux support (Linus said that).

Linus says this because the drivers are not open source. They are plenty fine (for Linux, I’d say excellent even).

A history of ATI is pretty abysmal regarding OpenGL support. For most of the years around 2000 it was pretty much a joke among the OpenGL-using community. OpenGL was only barely supported and it was a buggy piece of garbage. Things improved once AMD bought them but they are still climbing out of a hole.

I have one persistent problem on some ATI card where users get a “Shaders failed to link” error with no additional information. Makes me want to kick someone. :wink:

The basic sample works, but also looks strange:
The grey values look like dithering (when you look closely at the full size image)