With emerging many-core architectures, using on-chip shared memories is an interesting approach because it provides high bandwidth and high throughput data exchange. Such a feature is usually implemented as a multi-bus multi-banked memory. Since predicting timing behavior is key to efficient design and verification of embedded real-time systems, the question that arises is how to evaluate the access time for one memory access of a given task while others may concurrently access the same memory-bank at t the same time. In this paper, we give the answers for a subset of streaming applications modeled like CSDF Model of Computation and implemented in Kalray’s MPPA chip.