Comment by kimixa

Realistically the models will always be resident in the GPU during processing - the CPU and it's memory would probably more be used for related development tasks (as this is a workstation not a cloud server).

If it's just running the same model all day that CPU and it's RAM would be pretty wasted.