Hi guys, I plan on switching from OpenGL to D3D, and the switching goes well, but I have a few questions: - What is the fastest way to send static geometry to the accelerator. statci geometry means: the vertices never change for this kind of geometry. The transform matrix can change, of course. Thus I am looking for a mechanism like OpenGL display lists or compiled vertex arrays. - What is the fastest way to send dynamic geometry to the accelerator? This is on-the-fly generated geometry of which the vertices are not known in advance - What is the best way to threat a large amount of small triangles? Triangle strips? (does not matter too much with OpenGL) Indexed primitives? Something else? - I am using D3DX to set up D3D Immediate mode. D3DX is a wonderful library to make the life of a former OpenGL programmer bearable. But, do I use the modern T&L enabled HW to it''s fullest extent with this? Or should I go through all the mess and ask for a T&L HAL device myself? Thanks in advance, DaBit

Optimized vertex buffers are best for static data. Just call VB->Optimize(), these are similar to compiled vertex arrays.

You can use VB''s as a data stream to the accelerator. Check a recent reply of mine in "How many vertex buffers?"

This really depends on the hardware, and isn''t OpenGL or Direct3D-specific. Indexed triangle lists are usually pretty good, some hardware can attain their best performance with strips (like most 3dfx cards and the GeForce.)

D3DXCreateContext() with a hardware level of D3DX_DEFAULT will create a context that has support for hardware TnL, if available.