I am studying the situation where power consumption causes CPU frequency reduction. I want to construct a workload that can generate the highest possible power consumption and trigger the frequency reduction. Considering that a CPU micro-architecture consists of multiple components, including instruction fetch, decoding, issue, execution, branch prediction, among others, as well as cache, memory controller, MMU, TLB, etc.

Generally speaking, which components have the largest power consumption during operation? How should I code specifically to target these components?