With　the　craft　and　application　technology　continue　to　evolve,　progress,　on-chip　multi-core　technology　becomes　more　mature,　however,　with　the　size　of　the　on-chip　processor　becomes　larger　and　larger,　the　number　of　integrated　processor　cores　becomes　more　and　more　large,　which　leads　to　the　technology　of　NoC(Network　On　Chip)　consisted　by　the　processor　cores　and　other　components　having　greater　influence　on　overall　performance　of　multi-processor　system.　Different　network　topologies　and　routing　scheduling　will　affect　the　overall　performance,　such　as　execution　time,　network　latency,　power　consumption,　etc.　So　in　this　paper,　we　referred　a　new　topology　between　cores　and　other　components,　adding　more　interconnections　and　setting　different　weight　for　every　links.　Then　adjusted　communication　scheduling　of　router,　changing　port　sequence　being　checked　dynamically,　which　improved　the　situation　of　communication,　reduced　the　latency　of　network　and　running　time,　increased　the　throughput　of　network　and　improved　the　instruction　per　cycle.　Through　the　improvement　the　network　latency　reduced　9%-10%　compared　with　present　scheme,　and　as　the　number　of　cores　increasing　the　procedure　execution　time　reduced　8%-9%,　instruction　per　cycle　increased　3%-5%.　Effect　will　be　more　pronounced　with　the　cores　and　tasks　increasing.