I don’t know that this definitively answers that question. It’s possible to use a different architecture based on cost/performance and keep a small population of Intel machines in service because you want access to their superior PMUs. Most of what you learn on the latter would still apply to the former.