Research on Model Rendering Performance in Blender 3D Using Massively Parallel Systems
Olesia Barkovska, Oleksandr Shulinus, Dmytro Rosinskiy, Yevhen Lebodkin, Vitalii Serdechnyi · 2023
Object rendering is currently used in a wide range of applications, from movie and game production to manufacturing, architecture and medicine. Companies of various levels use rendering to promote, produce and present their products. This trend is only increasing, which indicates the relevance and demand for research in the relevant problem area. The transformation of models and scenes into a realistic image using specialized software is accompanied by complex calculations, which requires the use of high-performance computers to reduce rendering time. That is why our work is mainly focused on the rendering execution time for various numbers of passes (calculation operations) required by the renderer to achieve a high-quality picture (with minimal graininess) on computers utilizing different graphics processors — Nvidia GeForce GTX 1050 Ti and AMD Radeon Vega 8 Graphics. The results of the experiments showed that a larger number of samples leads to less noise and a smoother display of lighting and materials, but at the same time, the time required for rendering increases. Comparing the performance of graphics cards, we can say about the advantages of Nvidia graphics processor over the AMD GPU, which, with relatively equal technical characteristics, shows an acceleration of almost 6% (for 4096 samples) and 36% (for 512 samples).