This dataset documents hardware power profiling and core isolation experiments conducted across three distinct processor architectures: the Rockchip RK3588 (Orange Pi), Allwinner A733 (Cubie A7A), and a modern Intel hybrid x86 platform. The methodology relies on platform-specific automation drivers to induce synthetic mathematical workloads while constraining task execution to designated core clusters via explicit CPU affinity locking. Telemetry is captured concurrently through localized monitoring scripts that poll hardware nodes, native Power Management Integrated Circuits (PMICs), and Intel Running Average Power Limit (RAPL) interfaces under both dynamic scaling and fixed userspace frequency paradigms. The resulting time-series metrics—stored in structured CSV and log formats—are processed using specialized convolution and statistical correlation engines to calculate Pearson (r) coefficients and bin-dependent statistical convolutions. This mathematical framework generates hardware coupling matrices and entanglement scores, allowing investigators to isolate dynamic throttling artifacts and map operational dependencies between independent processing islands.
