Objectives Blood is a typical non-Newtonian fluid, and poor fluidity of blood is key to the pathogenesis of thrombus. The dynamic viscosity (high blood viscosity) of blood is a key factor in blood vessel embolization, especially cardiovascular and cerebrovascular embolisms. Computational fluid dynamics (CFD) is an efficient method for Newtonian or non-Newtonian fluid flow simulation. The finite volume method (FVM) is also an efficient numerical method for fluid dynamics simulation. This study aimed to investigate blood flow characteristics in arteries, and the dynamic viscosity is taken into account via the non-Newtonian fluids and CFD theory. Methods A series of numerical computations are presented to reproduce the process of blood flowing in arteries using the FVM code CFX. Blood flow simulation will be accelerated using the GPU parallel computation scheme with the CUDA-enabled GPU coprocessor. Results The dynamic viscosity and flow field of blood will be computed and shown by the contours. The velocity, shear strain rate and dynamic viscosity of blood can be calculated and monitored for different ages of people. Conclusions The dynamic viscosity and flow characteristics of blood in arteries can be reproduced using CFD and FVM. The mechanism of blood vessel embolization can be analyzed via the non-Newtonian fluids and CFD theory. Thrombus occurs when the rheology of blood is enhanced. Acknowledgements This research was financially supported by Shaanxi Hundred People Plan Foundation (Grant No. 302-253051601).