In the field of artificial intelligence and game theory, GTS is a computational problem. Fast GTS algorithm is crucial in computer games. In this paper, to enhance the speed of game tree search and utilize a capability of parallel processing in game tree search using GPU, we concentrate on how to grip extensive parallelism capabilities of GPU. The system works on the real time game called Tic-Tac-Toe. This game is also verifies the effectiveness and efficiency of MINIMAX algorithm. It doesnt allow one player to succeed all the time and a significant proportion of games played result in draw. The focus is on the advance of no-loss strategies in game using decision tree algorithms and comparing them with existing methodologies. The motive of this paper is to consult compares and examine various parallel algorithms of gaming tree and improve the acceleration of game tree search. The main focus of our system is on the implementing the game using the MINIMAX algorithm. NVIDIA™ made CUDA™ programming language is used and implemented by (GPU) to accomplish the game theory. Toget better performance of GTS algorithms GPU is widely used in game. The MINIMAX approach is the best method to locate best move in a computer game and GPU works on it. The perception of the work is using GPU is the most feasible way for improving the performance of GTS.