It comes down to what you define as a core.
Like Intel's HTT, AMD's CMT is a form of SMT. Essentially, SMT is a way for cores to process two threads at the same time, without needing to duplicate all the resources of a full core.
From a HW perspective: CPU cores have two main parts: The parts necessary to maintain information (Registers), and the parts necessary to "do something" (ALU's, schedulers, etc).
AMD's CMT duplicates the parts needed to maintain information and most of the resources needed to perform Integer based functions, but does not duplicate the resources needed to handle floating point math. As a result, when doing FP workloads, each module has the effective throughput of only a single core. As a result, Windows treats every second core as a "logical" processor, which is just a way to try and avoid scheduling multiple threads on a single module if possible, in order to reduce the chance of two threads needing access to one set of execution resources.
intel's HTT works in a similar way, except NONE of the execution resources are duplicated, making it a very light form of SMT.
In both cases, the "physical" and "logical" processors are IDENTICAL in HW; they just share parts of the same execution backend. The reason software calls cores "physical" and "logical" is so they have some way to avoid scheduling threads on shared resources. In Hardware, all cores are the same.