Anti-Alias, trilinear, Anisotropic ?

HI,
her is a new thrad because the other was also about light/benchmark and tha is more advanced

1)Which of AA is best: FSAA, HRAA, Oversampling, Supersampling, TMAA, TSAA, Multisampling ?

  1. How do I use it with OpenGL ?

  2. How can I set the level to 2x, 2xQ, 4x, 4xS, 6xS,… ?

  3. I heared that Supersampling changes the quality of the textures, the others don’t ?

  4. Is that right :

// normal
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MAG_FILTER,GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MIN_FILTER,GL_NEAREST);

// bilinear
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MAG_FILTER,GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MIN_FILTER,GL_LINEAR);

// trilinear
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MAG_FILTER,GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MIN_FILTER,GL_LINEAR_MIPMAP_LINEAR);
  1. What does Anisotropic if I don’t use trilinear ?

  2. What is the difference between
    1.blilinear

  1. trilinear with AAx16
  1. If I cange the resolution, are the edges/textures of my model better (do I need other models/textures)

  2. What does texture compession ?

  3. Do you know a page where I can get the performance of all nVidia cards (and which is good for AA / good quality)

I would be glad if you could help me…

Sorry again for this long post :o

Can’t you help me?
It would be enough if you could answer one of my questions…

1: Some of these are synonyms for the other. Some of them aren’t even techniques in and of themselves (FSAA is a term for any anti-aliasing technique that opperates on the full scene, rather than on certain parts of the scene). And I’m not sure that some of these even exist (TMAA, TSAA).

2: How do you use what with OpenGL?

  1. How can I set the level to 2x, 2xQ, 4x, 4xS, 6xS,… ?

The level of what? Are you rendering to the default framebuffer or an FBO? I don’t know what the “Q” and “S” refer to, but if you want multisampling in OpenGL, there are two ways to go about it.

The first is to create a multisampled default framebuffer. This requires using WGL_ARB_pixel_format and WGL_ARB_multisample to get a multisampled framebuffer.

The alternative involves the use of ARB_framebuffer_object (or, failing that, EXT_framebuffer_object with EXT_framebuffer_blit and EXT_framebuffer_multisample) to create multisampled renderbuffers that you render to. Then you blit to the non-multisampled default framebuffer.

Creating renderbuffers with multisampling (glRenderbufferStorageMultisample) requires that you specify the number of samples to use per-pixel. The renderbuffer may have more samples than this, but it must have at least that many.

  1. I heared that Supersampling changes the quality of the textures, the others don’t ?

No. Supersampling cannot improve the quality of textures by itself. All it does is allow textures to be rendered at higher resolution; if the textures don’t have the pixels for the higher resolution, they won’t look any different.

  1. Is that right :

Yes.

  1. What does Anisotropic if I don’t use trilinear ?

Implementation-dependent. According to the spec, “Additionally, implementations are free to consider the current texture minification and magnification modes to control the specifics of the anisotropic filtering scheme used.”

If you don’t use mipmap minification filtering (GL_LINEAR_MIPMAP_NEAREST or GL_LINEAR_MIPMAP_LINEAR), then the implementation cannot access other mipmaps. If you do use mipmap filtering, then it can do so.

  1. What is the difference between
    1.blilinear
  1. trilinear with AAx16

I’m not sure what you’re asking for there.

  1. If I cange the resolution, are the edges/textures of my model better (do I need other models/textures)

If it were me, I’d say yes. Most people do not agree with my views on this subject. They seem perfectly fine with using textures and meshes with resolutions that are much larger than they were designed for.

It probably isn’t worthwhile to have lots of texture and meshes for different resolutions.

  1. What does texture compession ?

It allows textures to be… compressed. They take up less room, at the cost of some visual quality.

  1. Do you know a page where I can get the performance of all nVidia cards (and which is good for AA / good quality)

The performance for them at doing what?

  1. I think TSAA means Transparence multisample anti aliasing or something like that^^

  2. I’m using the Lesson 36 of NEHE, that uses multisampling, but I don’t see a difference between 4x AA and 8x

Q means Quincunx…

I want the best possible visual quality, and in all these 3-D Games the differences of AA are higher…

6)I can’t see any differnece between 0x and 16x, so what can I use it for ?

7)I thougt it looks the same ^^

  1. So I need one model/texture for each resolution ???

It looks the same if I use another resolution, only with some the models get resized…

Or can I use one high resolution model (without loss of performace) for lower resolutions ?

And what’s about the texture ? One high resolution (512x512)?
10) The performance for … ALL I don’t know what it is necessary (CPU clock, graphic memory, shader clock,…)

  1. I’m using the Lesson 36 of NEHE, that uses multisampling, but I don’t see a difference between 4x AA and 8x

6)I can’t see any differnece between 0x and 16x, so what can I use it for ?

7)I thougt it looks the same ^^

What exactly is your test case where you can’t see any difference? Is there a lot of aliasing or lack of texture smoothness in that scene, such that you would expect to notice a difference?

  1. I heared that Supersampling changes the quality of the textures, the others don’t ?

Super sampling can improve texture mignification/aniso filtering as it takes multiple shading samples. MSAA does not, only improves egdes.

For ultra high quality AA texture sampling there’s a nice demo of EWA filtering in GPU Gems 2.

Basically what you’re trying to do is approximate and average the sample area of a back projected pixel in the scene (or a polygon in our case). The idea with EWA then is to fit an ellipse to an arbitrary quadrilateral (pixel footprint), then average the samples that fall inside.

Sorry, but I don’t understand you, that’s too complicated for me, I’m german …

I still don’t know what Anisotropic does (with trilinear it makes the texture better…)

So what do I have to do if I want to use high resolution like in all these 3D-Games?

EDIT: Do high resolution models have more triangles / better textures ?

?

Wikipedia is your friend. It can tell you how anisotropic filtering works.