Embedded systems design combines software implementations running on an on-chip processor and dedicated hardware components. It also introduces IP-components (Intellectual Property) to be reused and integrated in Systems-on-a-Chip (SoCs). This means a tremendous paradigm shift from the traditional system design. This paper introduces an embedded systems design flow in which the major challenge is the exploration of the design space for optimal architecture configurations. We show that automation of this architecture exploration phase heavily relies on fast and relatively accurate performance estimates for both hardware and software implementations simultaneously. For performance estimation of hardware, we advocate the introduction of a priori interconnect estimations in architecture exploration tools and show how such estimates can be used beneficially.