When theora decodes a frame, it gives us YUV pixels but we need RGB to use it in 3D applications.
Therefore, we need to convert each frame on the CPU. More on TheoraVideoPlugin HowItWorks.
On high resolution videos this becomes a problem, and could cause frame dropping on slower machines.
Converting YUV on the GPU accelerates the whole process, but has pros and cons, see How it works article for details.
How to use
Using your own shaders