Recently, the downsizing of engine using turbocharger attracts more and more attention. Generally speaking, a turbocharger is usually designed based on its steady performance curve. However, the operating point of a turbocharger turbine does not match the steady operating point: instead it shows hysteresis behavior because of the pulsating flow generated by the engine valves. Unfortunately, turbine efficiency drops under pulsating flow conditions, but the loss mechanisms of the turbine under these conditions are not understood. Internal flow measurements under pulsating flow are actually very difficult. In this study, the internal flow under pulsating conditions was measured using a high speed PIV (Particle Image Velocimetry) system. The loss mechanisms were investigated by experimental investigation and computational fluid dynamics (CFD). The instantaneous pressure, velocity and torque were measured using a turbine experimental apparatus at WASEDA University. To generate the pulsating flow, a pulse generator was placed upstream of the turbine: a rotational disk with holes that only lets the flow through periodically. The pulsating frequency could be changed freely by changing the rotational speed of the disk. The visualization using PIV was performed at a frequency of 1 kHz at the turbine outlet. Many fine vortices which rotate in various directions were observed under pulsating flow. Such vortices mix in the exhaust diffuser and under low frequency flow, mixing of vortices took a long time. It was observed that one loss mechanism under unsteady conditions is the mixing of vortices at the turbine outlet. CFD was performed using ANSYS-CFX, with approximately 10 million nodes. Turbulent flows were treated by using the Reynolds-averaged Navier-Stokes (RANS) and Detached Eddy Simulation (DES) with the SST k-ω turbulence model. It was confirmed that the wheel and exhaust diffuser total pressure loss under pulsating flow was higher under steady flow conditions. In addition, the total pressure loss is proportional to the flow pulsation frequency. The analysis with DES agreed with the PIV results qualitatively. On the other hand, the analysis with RANS could not simulate the flow pattern at the turbine outlet.