I found this question which talks around the question I'm interested in. In particular I focused on the line where the guys says that GraphicsDevice object is literally a representation of the underlying graphics card.
Now, can I assume that XNA automatically uses the memory directly from the underlying graphics card (to perform the most expensive calculations like updates and drawing)?
If so, is everything (like thread management) already optimized or there are way to configure the device depending my various cases? If not, do I have to tell the GraphicsDevice object whether to use or not the GPU memory rather than the CPU?