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.
Comments
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.