It's not a dumb question, but it's so broad that it would take at least an entire textbook to fully answer it.
Just off the top of my head, some possibilities are:
* More cores
* More numerous/powerful execution units that can do e.g. vectorized math
* Architectural features such as virtualization and security defenses
* Bigger and more flexible caches
* Complex pipelining/out-of-order/speculation logic that allows more instructions to be executed (on average) per clock cycle
* Special-purpose functional units that dramatically accelerate particular applications (e.g. AES encryption, video encoding/decoding) and are idle the rest of the time
* Replacing "deep" networks of logic gates with equivalent "wider" ones, which occupy more die area but have a shorter critical path, enabling faster clock speeds
Just off the top of my head, some possibilities are:
* More cores
* More numerous/powerful execution units that can do e.g. vectorized math
* Architectural features such as virtualization and security defenses
* Bigger and more flexible caches
* Complex pipelining/out-of-order/speculation logic that allows more instructions to be executed (on average) per clock cycle
* Special-purpose functional units that dramatically accelerate particular applications (e.g. AES encryption, video encoding/decoding) and are idle the rest of the time
* Replacing "deep" networks of logic gates with equivalent "wider" ones, which occupy more die area but have a shorter critical path, enabling faster clock speeds