We have five decades of very fine-grained analysis of CPU compute engines in the datacenter, and changes come at a steady but glacial pace when it comes