Which is more optimized?

Good question, I want to know too :b

Can't see how this is difficult to actually try and see for yourself. It's not like there's a universal answer which is always true.

Quote:

Original post by Brother Bob Can't see how this is difficult to actually try and see for yourself. It's not like there's a universal answer which is always true.

Actually I'd be terrified is there wasn't a universal answer, because that would men Nvidia and ATI's driver engineers don't really know what they're doing.

Well, I can't answer the question, but unless your grid of vertices is VERY close to the screen resolution you're going to get substantially different results for each. If you use colours the grid squares are large you'll get the old problem where which way the quad has been divided into two triangles will be obvious because the dividing line will only blend colours from two corners of the quad. On the other hand if the grid squares are smaller than a pixel you'll get aliasing, so I'd pick texturing for those reasons alone. And judging by the way nVidia seem to be going with their hardware you might not actually need to load the grid into vertex memory on post-GeForce 8 graphics cards because they can generate geometry internally (I'm not totally sure how this works though).

I don;t know about OpenGL, but in DirectX, you would need an explicit interaction with the vertex buffer, to modify it's contents (Lock), so that could be more costly than just assigning a texture, especially when the number of vertices is very large.

Quote:

Original post by soconne Quote: Original post by Brother Bob Can't see how this is difficult to actually try and see for yourself. It's not like there's a universal answer which is always true. Actually I'd be terrified is there wasn't a universal answer, because that would men Nvidia and ATI's driver engineers don't really know what they're doing.

So you're telling me things like passing the attributes by immediate mode, or vertex arrays from system memory, or vertex arrays using on board vertex buffers, or using potentially unsupported (by the hardware) non-power of two textures, or a texture with an unfriendly internal format does NOT affect the result?

There are several ways to draw something, and all of them doesn't have the same performance. So how you draw it does makes a difference, therefore there may not be a universal answer.