... how? There aren't any generally-available CPUs that are substantially faster today than were available ten years ago. Maybe double the speed per core, trip

... how? There aren't any generally-available CPUs that are substantially faster today than were available ten years ago. Maybe double the speed per core, triple at best. After that, throwing more cores at it also rapidly runs out of steam because parallel code has its own overheads. Any shared state instantly kills performance, no matter the language. Very clever tricks have to be used to get decent scaling past 64 hardware threads (32 cores), and going past 256 is surprisingly difficult. You start having to worry about NUMA, IRQ steering, and core pinning. Bandwidth gets to be an issue, even to L3 and L4 cache, let alone out to main memory. This notion that you can just "dial up" hardware performance to infinity as a fix for any amount of developer laziness needs to die.