VBOs are more powerful, since they can be combined and modified.

For example, a vbo with position data can be used by itself during a depth pass and together with another vbo that contains normals and tex coords when doing the actual rendering.

Display lists can't be modified at all so you have to build new ones, which is pretty slow.

If you're just drawing a static object - no I don't think VBO's are faster, they might actually be slower.

I'm not sure nested display lists are supported with good performance on modern drivers.

Stick with display lists if they work well for you - but I think most modern GL applications just use VBOs for all drawing.

Quote:Original post by slvmnd
If you're just drawing a static object - no I don't think VBO's are faster, they might actually be slower.

I have measured about 2% difference - it's not even enough to be a safe measurement. Cannot we just pass over that old myth? Lists were faster because they were the only way to write to server-side memory. Nowadays, I wouldn't be surprised to discover display lists as a layer over VBOs and a bit of driver unrolling.

Lists are dead and everybody using them should be warned that they'll be left out in the cold soon.

Quote:Original post by Dolf
What's the advantage over putting a vertex array in a display list?

Do not confuse vertex arrays with VBOs. Historically, putting client-arrays in a display list was an efficient way to gain near-VBO performance. In fact, in GL1.0 texture objects were simulated with TexImage calls inside display lists... pre-historical stuff.
Nowadays, I don't think there's any good reason to do so and it opens a can of worms.