The post is old, but still somewhat true. The most important change is how memory carriers can be moved between schedulers/cores nowadays which improves the TLB miss rate and makes for better system locality.