Here is a talk and the presentation slides about how to optimize texture transferts in OpenGL. The presentation covers textures uploads in multithreaded applications, PBOs and synchronization (GL_ARB_sync). The talk is available HERE. You can download the presentation PDF HERE.
This presentation is also available in OpenGL Insights, a new OpenGL book, under the chapter Fermi Asynchronous Texture Transferts:
For an overview of the contents, just visit this link.
Too bad they won’t enable dual texture transfers like they have on Quadro (and Telsla for CUDA) on GeForce.
@Leith Blade
What do you mean with dual texture tranfers?
Probably he means asynchronous host transfers. NV introduced in Fermi two host controllers working in parallel.