In　the　very　large　scale　integration　(VLSI)　global　routing　stage,　one　of　the　most　critical　considerations　is　how　to　obtain　a　proper　balance　between　routing　quality　and　runtime.　In　this　paper,　we　adopt　a　multi-threading　technique　to　accelerate　various　subphases　of　global　routing.　To　ensure　the　robustness　of　routing　results,　we　add　a　stage　called　congestion　and　wire　length　aware　breaking　circles　algorithm　while　ensuring　to　remove　the　most　congested　or　longest　wire　segment　in　the　loops.　Since　routing　congestion　is　the　most　critical　constraint　in　the　global　routing　phase,　we　present　two　accurate　cost　models　to　evaluate　the　routing　resources,　which　can　effectively　reduce　the　congestion　of　routing　results.　The　experiment　shows　that　our　proposed　method　reduces　the　violations　by　28.3%　and　achieves　a　2x　faster　runtime　in　benchmarks　of　the　ISPD＇19　contest　compared　with　CUGR.　©　2022　IEEE.