Scientific visualization is a visual representation of results of scientific research by computer graphics. Thus, the numerical results of research can be presented in a more appropriate graphic visual form. This makes it possible to see something that is hidden in the numerical data, since the main goal of the research is not only a numerical measurement of any performance, but also an opportunity to make visible some processes. Often scientific visualization is tool that can be used only to explain the essence of the modeled process and to consider a process in dynamics. This proposal considers about possible approaches for the creation of visualization system designed for scientific experiments. Possible hardware solutions are represented with various combinations of numerous components, such as computing nodes which are based on graphics cards, visualization and virtualization servers. Peripheral devices are studied in details. Description of software solutions that can be used are also provided. The discribed system can be divided into three subsystems. First of them is subsystem of virtualiza-tion, based on NVIDIA GRID m60 cards. The main task of NVIDIA GRID graphics accelerators is to provide a highperformance graphics-intensive applications working with demanding graphics compu-tation in a virtual environment directly. This approach also avoids additional costs of purchase licenses for individual computers. Next subsystem is a computing cluster with NVIDIA k80 cards, which provide high rendering performance. This platform com-bines the fastest GPU-accelerators widely used model CUDA parallel programming and extensive ecosystem of application developers, application providers. The third subsystem is the subsystem of the image output. The company NVIDIA has a special NVS technology. It is the standard for business graphics with support for multiple displays.