VOB performance

Does anyone know how much performance gain I can have from using Vertex Buffer Objects?
I have a test with AthlonXP1800+, GeforceTi4200: there are no textures, only vertexes and normal vectors (yet).
So, it’s about 40% faster. Is it ok?
The original program used glVector3fv and such functions.

I wouldn’t worry too much about it. VBO’s, in the long run, are going to be the optimized rendering path for OpenGL.

Hi,

I would like to post my result about VBO extension…

flat triangles with 1 buffer
on linux (GF 4 TI 4600, Athlon 1.2Ghz):
19Mtri/sec
on windows (GF 4 TI 4600, Athlon XP 1800):
35Mtri/sec

It seems that if we use more than 1 buffer, we loss a lots of fps on nvidia, anyone can confirm this fact ?
and is there the same problem on ATI ?

psc80

I would like to know if VBO is currently faster or slower than VAR or maybe it’s the same speed? Anyone done any benchmarking on this?

Originally posted by Adrian:
I would like to know if VBO is currently faster or slower than VAR or maybe it’s the same speed? Anyone done any benchmarking on this?

Some people say that the VAR is faster a little bit.
NVIDIA says that it should have the same performance.

Unless there’s a significant speed difference, you really shouldn’t worry. It’s best to get used to VBO as much as possible since it’s meant to replace similar vendor specific extensions.

I added VBO support to my engine just 3 days ago.
First i have to say, that VBO is MUCH easier to implement, than VAR. It takes only 5 to 10 function-calls to get well organized buffers. In VAR i had to do all this organization-stuff myself, which took me several days to get it as good as possible.

I compared the speed between VAR and VBO. Both were equally fast. But this was a “real-case” scenary, that means, i had a real level, with 20 to 40 texture-switches and only a few thousand triangles (less than 4000). So i don´t know if there is a difference, if you try to get as much triangles out as possible.

Jan.