Unigine Heaven Dx11

Lite info för dem med dx10 kort.

Due to recent confusion regarding DX11, I’m going to try to clarify some things.

First of all DX11 is a superset of DX10, what does that mean? That means that when a developer codes a “DX11 Path” into the game, as in supports DX11 in the game, DX10 is automatically supported as well. So a game that says “Supports DX11/9” also includes 10 by default.

Secondly, another confusing factor is how tessellation works on older hardware. Short answer, it doesn’t. Long answer: In the recent unigine demo, tessellation appears as an option that can be enabled on DX10 hardware, it actually can’t be enabled. Tessellation requires dedicated hardware. An interesting fact is that ATi cards including the 2xxx series, 3xxx series and 4xxx series actually have hardware tessellation, but it’s an ATi proprietary standard that isn’t compatible with the DX11 version, thus won’t be supported by default, unless a developer specifically adds support for it. In the future, it’s more than likely that features specific to DX11 will be grayed out for non DX11 hardware.

Tessellation isn’t magic. A lot of people seem to think that tessellation should be applied to every surface as if it works some kind of magic that suddenly makes everything look better, it doesn’t. What tessellation does, is subdivide an object into more pieces based on the viewpoint of the camera. This doesn’t automatically make objects better looking though, in fact if you applied this to an object vs not applying it at the same distance from the object, the object will look EXACTLY the same. So what advantages does it bring? A few actually - one is in dealing with displacement mapping. Ok so basically a displacement map works by applying a height map on the material, physically changing the vertex's position on the object. The more vertex's you have the more precise the height map can be. Tessellation splits subdividing an object into more pieces based on view settings. With more polygons you have more vertex's and can ultimately build a higher quality displacement map. The benefits are that tessellation is based on view settings, so you can effectively control the level of detail, instead of using various LOD replacements at certain distances. Another advantage is that because it's physically altering the object itself, it's much easier to have lighting properties effect the object. The disadvantage is that it's extremely performance sensitive on the vertex side. You're creating tons of new polygons that need to be rendered. It also can be performance sensitive on the shader side, if you apply complex shaders to the material, as lighting will fully affect the object.

Tessellation is good for physics. Because it can subdivide objects into smaller sections, physics calculations can be made more realistic. Physics works by taking an object and animating it based on the defined physics parameters for that object and how that object is broken up. For instance, if you took a solid plane and applied a cloth physics tag on it and simulated it, it would behave like a piece of wood. A cloth plane needs tons of subdivisions in order properly simulate cloth like flex effects. Because tessellation does this, you can apply tessellation to a solid plane have it be broken up into tons of smaller pieces, apply the cloth simulation now and it will behave correctly. Another advantage once again is LOD properties of tessellation. A physics object in the distance doesn’t actually need to be dead accurate as oppose to one right in the players face. Because tessellation can scale based on the viewpoint of the camera, you can increase the efficiency of physics calculations based on distance.

Hope that clears some stuff up.

Source: Guru3D