Analyzing the CUDA Applications with its Latency and Bandwidth Tolerance