I agree with that - but if you take an unchanged core and manufacture it at a different node, then you won't see a change in IPC, which in my book makes it questionable to attribute IPC gains to the process node.