View Full Version : ATI's pbuffer bug + NV's IBO bug ?

09-08-2004, 09:55 AM
A couple of questions/remarks:

1. On an ATI R9700 Pro, Catalyst 4.8. Any texture that was created before a wglShareLists call when creating a pbuffer, is getting invalidated/corrupted. Any texture created in the main context _after_ the wglShareLists call can correctly be accessed by the PBuffer, though. I'm 99% sure it's a driver bug as testing on an NVidia card showed the correct behavior. Reported to ATI's devrel a few days ago, no answer yet. It's a critical problem since to avoid it, i need to create all my PBuffers as soon as possible, at initialization time - but at that time i have no idea how many PBuffers i'll need (it's for a texture manager with a user-variable amount of textures).

2. On an NV 6800, latest drivers, updating a dynamic IBO (created with GL_STREAM_DRAW_ARB or GL_DYNAMIC_DRAW_ARB) cuts the framerate by a factor of ~4, even for a small polycount (from 150+ fps to 40). Disabling IBOs (but keeping VBOs as usual), and rendering the index buffers from system memory, makes the framerate go back to 150. The update is done by calling glBufferSubDataARB. Any idea on what could be causing this ? Or anybody had a similar experience ?


09-08-2004, 05:32 PM
EDIT: This was for STATIC_DRAW indices. I'll try dynamic if I have the time.
Indices in buffer objects on a 6800 are working for me.