2 Replies Latest reply on Nov 28, 2015 12:35 PM by sgoodrow

    Heightmap Problem with Radeon HD 6870

    sgoodrow

      Hi, I am the graphics programmer at Ilsanjo Studios; we are receiving reports of problems with a particular AMD graphics cards (XFX HD-687A-ZDFC) in the development of a PC game (The Wild Eternal) using the Unity3D game engine; we are using the latest Unity 4.6.7 version.

       

      The rendering of heightmap-traced fog is being corrupted on these cards, resulting in an inaccurate depiction of a dynamical fog surface that is critical for gameplay. I believe I have isolated the problem to be related to incorrectly transforming a texture into/out of gamma space when it should be in linear space. I am hoping with this inquiry to communicate with anyone who is familiar with any reports of problems with this card.

       

      I have included some source files that may help you to understand the problem. In the included .cs file, line 261 serves as a branch to avoid the heightmap texture noise perturbation process. This circumvents the problem. If anything is unclear, I can hopefully answer any particular questions that might get to the bottom of this.

       

      Unity/C# Script (renders shader as camera Image Effect): http://hastebin.com/qatewanoqo.cs

      Unity/CG Shader: http://hastebin.com/ipokijuqen.avrasm

      ilsanjo_radeon_problem.jpg

        • Re: Heightmap Problem with Radeon HD 6870
          pinform

          Hi Scott,

           

          Welcome and thanks for posting.

           

          I have white-listed you, so you should be able to directly post in the relevant forum. In the meantime, let me post this issue to the relevant dev forum.

           

          --Prasad

            • Re: Heightmap Problem with Radeon HD 6870
              sgoodrow

              Thank you Prasad. Is there any additional actions I can perform to receive a response to this, or is it just a waiting game and I should remain patient?

               

              We are getting very close to our release deadline and this issue remains outstanding. I am considering implementing a hack solution that reads the result of a test case and based on the result, disables the part of the effect that corrupts the heightmap data. So far, this is the only idea we have kicking around, and I'd much rather attack the source of the problem, not a symptom.

               

              Thank you.

               

              Scott