Virtual reality is an important tool for architects, allowing them to present the future object to the client from the early stages of design. A scene in virtual reality can be presented in two ways: by a VR headset connected to a computer or in standalone mode. With high-quality optimization, standalone VR systems can demonstrate graphics that are not inferior to VR devices connected to a computer, while remaining mobile, allowing the architect to flexibly organize meetings with the client. An effective scene optimization system is also necessary for stable and high FPS, which directly affects the comfort of the user experience and immersion in VR. Since models for architectural visualization are often not intended for use in VR, optimization of their geometry is necessary. Using manual optimization gives more control over preserving the visual quality of a 3D object and does not violate the UV-map structure, unlike the automatic optimization method. Creating retopology and using normal maps allows to preserve the level of detail of objects without increasing the number of polygons and thus reducing the load on the graphics card. The experimental part of this work examines in detail the effect of scene optimization on its performance. As part of the study, 2 scenes were created: the first included 3D models of interior elements downloaded from archviz sites, and the second - their optimized versions created as part of this work. In the next step, the performance of both scenes is analyzed and the impact of 3D object optimization on the overall efficiency of the scene is assessed. The ability to work with optimization allows the architect-visualizer to create high-quality realistic projects that enable the future structure to be presented favorably and increases efficiency in working with the customer.