This is called TurboCache for nVidia and HyperMemory for ATI. They have a small frame buffer just for the GPU to get around latency to RAM.
So for notebook #1 you actually have the full 512 available if you're only running the GPU with 64MB (barely enough to run low-resolution Aero). You can assign more and get the benefits of more video memory, but then you'll be competing with the rest of the system for RAM and lowering the amount of available system RAM. Probably better to let it act like a discreet card for now. You're still skimming above bare Vista usability.
For notebook #2 you must be using some additional system RAM because that 32MB won't cut it. Since you are now acting like integrated graphics you will get a performance hit due to both the system and the GPU contending for memory resources.
#1 - ATI 64 onboard 64 shared (of 512)
#2 - Intel 32 onboard 256 shared (of 2048)