Computer designers utilize the recent huge advances in Very Large Scale Integration (VLSI) to get Chip Multiprocessor (CMP) by placing several processors on the same chip die. The CMP is the dominant architecture to improve the performance of the current computing systems. However, accessing a shared data by several processors is a primary challenge in CMP. The data consistency must be reached among all memory hierarchies to ensure correct behavior and higher performance. This paper, proposed a CMP with an efficient multilevel cache system, which enhances miss rate and latency (penalty) by designing and implementation of different write policies with two levels of cache. The proposed system is implemented and tested using Hardware Description Language (VHDL) on Altera’s FPGA chip. The results show that a combination of write-through without buffer for the first level and write-back for the second level offers a clear improvement on the multilevel cache system performance.