I have two machines, Master and slave. It can distribute the render as expected and I can set the slave to either Cuda or CPU. For the master it does not seem to matter whether I set it to Cuda or CPU it will only render on the cpu? Also, the combined render is much slower then an individual render ?

I am planning on building a renderfarm using crowdrender and I have a fundamental& basic question to start with. Does each GPU need to have the memory to render the scene or is that distributed somehow? I guess I am wondering whether I can use cheap-ish GPU´s or If I need at least an 8 GB Gpu for each computer. If it distributes the memory for a single image does it work differently for rendering animations, that is , when rendering an animation does it render each frame on a different computer or does it distribute each frame and render the one frame on all computers ? Also, would it be worth running 5 computers on cpu only and would all of them need to have enough memory to load the whole scene or this also somehow distributed ? Thanks, excited to test this out.