In the context of the rapid growth of the mobile gaming industry, the issue of energy efficiency and sustainable development of digital technologies is becoming particularly relevant. Modern game projects are increasingly using complex three-dimensional models with a high number of polygons and detailed textures, which provides a high level of realism, but at the same time leads to a significant load on the computing resources of mobile devices. This is reflected in increased energy consumption, accelerated wear of hardware components and a shorter device lifecycle, which, in turn, contradicts the goals of sustainable development related to responsible consumption and reducing the carbon footprint. This study is aimed at analyzing methods for optimizing 3D models in the Blender environment for subsequent use in game projects based on the Unity engine, as well as identifying their contribution to the development of the Green IT concept. Approaches such as polygon reduction, texture baking, and resource compression are considered. The aim of the study is to demonstrate that optimizing a polygonal grid can not only improve the performance of mobile applications, but also reduce energy consumption, thereby contributing to the implementation of the principles of sustainable development. The paper proposes a methodology for comparative analysis of the original and optimized models in terms of file size, frame rate, processor load and power consumption. It is expected that the results obtained will justify the need for system integration of optimization methods into the practice of mobile game development. Thus, optimization of 3D models is considered not only as a technical technique, but also as a tool to reduce the environmental impact of the gaming industry.
In recent decades, the digital entertainment industry has experienced rapid growth, encompassing
not only console and computer games, but also, primarily, the mobile segment. According to industry
statistics, mobile games currently account for the largest share of the global market, attracting
hundreds of millions of users every day. The video game market has been growing rapidly in recent
years: its volume will increase from $310.97 billion in 2024 to $343.22 billion in 2025, corresponding to
a compound annual growth rate (CAGR) of 10.4% [
Modern mobile devices are limited in terms of computing power, battery capacity, and cooling capabilities. With gamers expecting console-like quality and developers striving to keep their audience engaged with graphics, there is an increasing load on central and graphics processors. This, in turn, leads to increased energy consumption, accelerated equipment wear, and a shorter device life cycle. Therefore, enhancing the energy efficiency of games represents a critical challenge for mobile devices with limited battery capacity. An additional factor is the growth of “digital waste”: users are increasingly forced to upgrade their smartphones due to incompatibility with increasingly resourceintensive applications. Taken together, this creates a problem of unsustainable consumption that directly contradicts the Sustainable Development Goals (SDGs), among which the tasks of responsible production and consumption (SDG 12) and combating climate change (SDG 13) stand out.
3D models, which form the basis of the vast majority of games, play a special role in this context. High-polygon models with a large number of textures and complex materials provide visual realism, but at the same time place a significant load on the system. In most cases, the level of visual quality achieved in movies cannot be fully replicated within games. This is because 3D models must be optimized for real-time rendering as players navigate and interact with the environment. Processing geometry with millions of polygons and multi-layered material maps requires significant computing resources. On stationary systems, this leads to increased energy consumption and heat dissipation, and in the case of mobile devices, to accelerated battery discharge and overheating. Thus, heavy 3D models are one of the key factors shaping the energy and environmental footprint of the gaming industry.
Optimizing polygonal meshes and related resources is an effective tool for reducing hardware
load. Methods such as decomposition, polygon reduction, implementation of levels of detail (LOD),
and material baking can significantly reduce model weight without critical losses in visualization
quality. Once computed and transferred onto the low-poly mesh, the normal map reproduces the
light reflection properties of the original high-resolution model. This process allows the simplified
geometry to approximate the appearance of the non-decimated version: in essence, by modifying the
shading of the reduced-detail mesh, the normal map enables the low-poly model to visually resemble
its high-poly counterpart [
Thus, it can be argued that the task of optimizing 3D models goes beyond purely technical or aesthetic considerations. It takes on a socio-environmental dimension, directly affecting issues of energy efficiency, rational use of resources, and sustainable development of the digital economy. At a time when the global agenda is focusing on reducing carbon footprints and increasing environmental responsibility, computer graphics optimization methods can be seen as an important component of a broader “green” game development strategy.
The purpose of this study is to demonstrate the potential of optimizing the polygonal mesh of game models as a tool for improving energy efficiency and reducing the environmental impact of the mobile gaming industry. The paper proposes to consider methods for simplifying geometry and managing levels of detail, evaluate their impact on performance and energy consumption, and discuss their contribution to the concept of Green IT. Particular attention is paid to practical scenarios for implementing optimizations in mobile projects, where the balance between visualization quality and efficiency is critical.
This work is structured as follows. The “Literature Review” section analyzes current research in the field of 3D model optimization and its connection to the concept of sustainable development. The “Methodology” section describes the research objects, the tools used, and the criteria for evaluating the effectiveness of optimization. The “Results” section provides data on the impact of various methods on file size, frame rate, energy consumption, and visual quality. The “Discussion” section examines the interpretation of the results in the context of sustainable development goals and the concept of Green IT. Finally, the conclusion formulates the main conclusions, limitations of the study, and directions for future work.
In recent years, sustainable development has become a key framework for evaluating digital
technologies. The COVID-19 pandemic has accelerated digitalization for sustainability. There is
growing interest among businesses and regulators in using digitalization for sustainability, but
academic literature on the subject lags behind. Technologies such as AI, IoT, blockchain, big data
analytics, cloud computing, 5G, remote sensing, and GIS make a significant contribution to the
Sustainable Development Goals (SDGs). Each of them supports several goals. AI optimizes resources
(SDGs 2, 3, 13), blockchain provides transparency (SDGs 12, 16), IoT monitors in real time (SDGs 11,
7), big data helps analyze trends (SDGs 13, 3), cloud computing scales (SDGs 9, 17), 5G improves
connectivity (SDG 11, 3), and remote sensing/GIS monitors the environment (SDG 14, 15). Thus, the
impact of digital technologies on sustainable development is complex and multifaceted [
Sustainable development in the digital environment resonates with the theme of Green IT. Green
IT is the discipline that encompasses the study and implementation of methods for designing,
manufacturing, operating, and disposing of computers, servers, and related subsystems, such as
monitors, printers, storage units, and networking or communication devices, in a manner that
ensures efficiency and effectiveness while minimizing, or ideally eliminating, adverse environmental
impacts. The systematic review notes that the key factors in battery discharge are user profile (23.53%
of articles — habits, usage time), mobile networks/Wi-Fi (17.65%), background applications/services
(11.76%), old hardware and standby mode (11.76% each), and background activity can increase
consumption by 160% in some scenarios [
Display adaptation – dark themes for OLED.
A study based on life cycle assessment identified video streaming as the type of content that creates
the largest carbon footprint in the atmosphere [
There are numerous studies on the topic of effective smartphone energy resource management. A
new approach to energy management when working with gaming applications on mobile devices has
been proposed: System-level Energy-optimization for Game Applications (SEGA) [
One strategy for efficient energy consumption is to offload processor-intensive tasks such as 3D
object rendering. This is also critical for increasing the performance of video games. In this article,
this phenomenon is examined from the perspective of optimizing polygons while maintaining the
quality of the viewer's perception. The main problem with optimizing 3D models is the balance
between model quality and performance, which stems from a paradoxical requirement: a 3D model
must look good but have a minimal impact on computer resources. Overly detailed models overload
the computer, while simplified models lose scientific accuracy [
Admittedly, there are limitations to this optimization method. First, normal maps do not compensate for geometric accuracy. When looking at an object from an acute angle, the effect of realism can be quite lost. So this method may not be suitable for tasks that require geometric precision from the developer. Displacement map, in turn, requires competent retopology and UV scan preparation, and errors at this stage are critical. Secondly, there is a general problem of accuracy with any geometric simplifications. If you do not control this moment, it reduces the metric reliability of the model, and it cannot be used for scientific measurements. The third is the unexplored topic. A review of the literature shows that there is a significant amount of work on optimizing 3D graphics and on Green IT. There is also a rich literature on sustainable development and digital technologies. However, there are almost no studies that directly link the methods of optimizing polygonal grids of game models with their contribution to sustainable development through reducing energy consumption, extending the life cycle of devices, and reducing e-waste. The present work is aimed at filling this gap. In addition to limitations, there are also opportunities. First, the optimization approach can be used not only in the gaming industry. It is also applicable to museums, architecture, online tours, educational VR/AR platforms i.e. in places where reusability and real-time rendering is very important. Secondly, cross-platform engines like Unity allow you to run such projects in the browser and on mobile devices. This, in turn, leads to a wider audience reach. The third is the democratization of games. With proper optimization, players with less powerful machines benefit as their devices become capable of playing games. The fourth is the concept of the "golden mean" of polygons. There is an optimal number of polygons, in which the visual quality is practically indistinguishable from highly polygonal models. In the gaming industry, such models are defined as mid-poly. Thus, improving performance by optimizing 3D models helps to work directly with the energy efficiency of the device. This method is not universal in all cases, but in most cases it helps to achieve the optimal effect.
The purpose of this study is to evaluate the impact of 3D model optimization on performance and stability. Using an optimized model as an example, we want to show how to find a balance between performance and visual quality of an object. The study is applied in nature and focused on practical results. As part of this study, we used Unity Editor version 2022.3.62f1 and Blender version 3.6. An experiment was chosen as the method, as it helps to empirically reveal the research hypothesis. The original and optimized versions of the 3D model will be compared. A single 3D barrel will be created as a representative example of low-polygon assets.
The standardized process of developing 3D models in the gaming industry is called a pipeline. In our study, the pipeline is divided into several stages: 1. High poly model simulation. This version is as detailed as possible and is used to bake maps and chamfers into a low poly model. As a rule, this model contains round and smooth shapes. 2. Low poly model simulation. This version is used in the final game and is optimized to the maximum extent possible. Before baking, this model does not have smooth shapes, and all polygons on it serve to preserve the overall silhouette of the object. 3. Creating a UV map. This process helps the 3D package recognize how to apply textures to the 3D model.
4. Baking textures. Transferring detail from the high poly model to the low poly model.
Several optimization methods will be applied to this model. First, polygon reduction using lowpolygon modeling of the 3D object. This method helps reduce the number of geometric shapes, making the object “lighter” and its rendering less costly. Second, texture optimization using the atlas method. This involves combining several materials into a single texture map. Our model will have two types of materials: wooden boards and metal products. They will be baked into a single texture file. The third is baking normal maps. This method is designed to add visibility of detail and relief where they should be, without adding new geometry. All these methods together help to create a low-polygon but visually pleasing picture of a 3D object.
The equipment used in this experiment includes several key aspects. First, the software. Blender version 3.6 was used for low poly and high poly modeling, Adobe Substance 3D Painter was used for texturing, Marmoset Toolbag was used for baking normal maps for low poly, and the Unity game engine was used for testing. Second, the hardware environment. The models were tested using an ASUS TUF Gaming A15 laptop and a Xiaomi Redmi Note 11S mobile device with the specifications listed in the Table 1.
To conduct the study, we collected a set of data reflecting the key parameters of the 3D model before and after its optimization. Data collection was carried out in two stages: first, the characteristics of the initial high-poly version of the barrel were recorded, and then the same metrics were determined for the optimized low-poly version prepared according to the pipeline described above. First of all, the geometric characteristics of the model were collected, including the number of vertices, edges, polygons, and triangles. These indicators are one of the basic ones for analysis, because they directly determine the complexity of the scene, the load on the GPU, and the overall performance when integrating the model into a game application. For ease of comparison, the values before and after optimization were recorded in the table, which makes it possible to visually show the percentage of reduction in the complexity of the polygonal grid. The next group of data was the size of the model files and texture resources. At this stage, the weight of the exported format files was measured .fbx, as well as the total size of the texture folder. This is necessary to assess the impact of optimization on the total weight of the game asset and, consequently, on the amount of memory that will be occupied in the mobile application. The third component is performance metrics collected directly in the Unity game engine. In the experiment, the average FPS value was recorded when visualizing a scene using high-poly and low-poly models, as well as the loading time of the scene and objects. For greater objectivity, the measurements were repeated several times, after which the average value was calculated. In addition, the GPU load was monitored, which allows you to indirectly assess the device's power consumption when working with each version of the model. Finally, the visual quality assessment stage was included in the study. To do this, screenshots were taken of both versions of the model in identical lighting conditions and camera position. Visual differences were recorded in a descriptive form, as well as potential distortions associated with the use of normal maps or a decrease in texture resolution. Despite the subjective nature of this analysis, it is extremely important because it allows you to correlate the technical indicators of optimization with the actual perception of the quality of the model by the end user. Thus, the data collection process included quantitative metrics (geometric parameters, file size, FPS, download time, GPU load) and qualitative observations (visual assessment). This approach provides an idea of the effectiveness of the proposed optimization method, which allows it to be linked not only to the technical aspects of game development, but also to sustainable development goals related to energy efficiency and prolonging the life cycle of devices.
An additional criterion for evaluating the developed model is its impact on sustainability. The fact is that the optimization method is designed to reduce the resource consumption of the user's hardware. This, in turn, leads to a lower load on the GPU, which helps the game application work more efficiently while maintaining quality. less GPU load leads to lower power consumption, which is one of the goals of sustainable development. Reducing the total weight of the model on the device leads to savings in traffic and memory, which prolongs its service life. This leads to the sustainability and environmental friendliness of the game process.
There are several limitations to this study. One of the key ones is sampling. During the experiment, only one barrel model is used, which may reduce its validity. It is also worth recognizing that energy consumption metrics are taken indirectly through indicators such as FPS and GPU utilization. In addition, to show the optimization results comprehensively, scaling is required on more complex game scenes. However, this is only a pilot study and a more elaborate design of the experiment is possible in the future.
Testing on a desktop computer confirmed a significant difference in computational efficiency between the high-polygon (HP) and low-polygon (LP) versions of the model. The number of vertices in the HP model was 2,159,592, while in the LP model it was only 1,891, which is a reduction of more than a thousand times. Similar differences are observed in the number of edges (5,756,516 vs. 5,088) and faces (3,596,994 vs. 3,216). The file size was reduced from 113,424 KB to 143 KB, which reduces storage and download speed requirements. At the same time, the average FPS remained virtually unchanged (144 FPS for HP and 147 FPS for LP), but the load on the graphics processor decreased from 3.7 ms to 2 ms, indicating a more efficient use of resources. Presumably, this is due to the small size of the game scene, which consisted only of a 3D model of a barrel. There is a hypothesis that as the number of models increases, the difference may become more significant. Thus, mesh optimization provides a radical reduction in geometric complexity and data volume while maintaining high visualization quality and stable frame rates.
Visual assessment showed that the low-poly model obtained using baked textures satisfactorily reproduces the geometric and material characteristics of the high-poly version. The main elements of the wood and metal details are preserved and visually distinguishable, despite a significant reduction in the number of polygons. The presence of normal maps and texture maps compensates for the loss of small geometric details, providing a comparable level of perception at medium and long distances. Thus, the use of Low Poly with baking can be considered as the optimal balance between visualization quality and computational load.
To collect data on the mobile device, we used the built-in Unity Profiler tool, through which we measured performance data. The measurement was carried out using a USB cable that connects the mobile device to Unity Profiler running on a computer. First, we made two builds containing the HP and LP model versions. Then we configured the Unity environment and the mobile device. Once the preliminary work was done, we launched the game app versions on the mobile device and recorded the metrics in a table.
Test results showed significant differences in performance between the high-polygon (HP) and low-polygon (LP) versions of the model. The LP model file size was 39.7% smaller than the HP model (107,302 KB vs. 177,825 KB), which reduces the load on the system when loading resources. The average FPS when rendering the LP model was 30 frames per second, which is significantly higher than the 18 FPS of the HP model and provides a smoother user experience. The CPU load also shows a sharp contrast: 0.96 ms for LP versus 53.03 ms for HP, confirming the effectiveness of mesh optimization. Similarly, RAM usage was reduced by almost a third (198.2 MB vs. 297.6 MB). Taken together, these metrics indicate that the optimized model delivers significant performance improvements without critical visual losses, making it a more sustainable solution for mobile gaming applications.
A comparative analysis of high-polygon and low-polygon models showed that geometry optimization followed by texture baking provides a radical reduction in the number of vertices, edges, and faces (by a factor of thousands), as well as a reduction in file size by up to hundreds of times. This leads to a reduction in the use of RAM and computing resources on both desktop and mobile devices: the load on the GPU and CPU is reduced several times over, while the frame rate either increases significantly (on mobile devices) or remains stable (on PCs). Visual evaluation confirmed that when normals and texture maps are used correctly, the low-polygon model retains a comparable level of detail and shape readability. Thus, model optimization allows a balance to be achieved between visualization quality and efficient use of computing resources, which is in line with the principles of sustainable development and Green IT.
Our results demonstrate that using polygonal mesh optimization and texture baking methods in Blender can significantly reduce the computational load when integrating models into the Unity engine, especially in the context of mobile games. A comparative analysis showed that low-polygon versions of models require orders of magnitude less geometric data and disk space, while providing either an increase in frame rate (on mobile devices) or maintaining stable FPS (on desktop systems). A significant reduction in CPU and GPU load, as well as a reduction in RAM consumption, indicates a more efficient use of computing resources.
In the context of SDG 7 (“Affordable and Clean Energy”), the study demonstrates that optimizing 3D models helps reduce the energy consumption of computing devices by reducing the load on the processor and graphics accelerator. Improving energy efficiency is especially important for mobile platforms, where battery savings directly affect the accessibility of technology for users with limited access to stable power sources. In addition, reducing energy consumption in conjunction with optimized use of hardware resources can be seen as a contribution to reducing energy consumption across the gaming industry. Thus, the development of optimized models not only improves application performance, but also supports the achievement of SDG 7, which aims to ensure cleaner and more affordable energy.
In the context of SDG 9 (“Industry, Innovation and Infrastructure”), the study demonstrates that the optimization of 3D models is an example of a technological solution aimed at improving the efficiency of digital infrastructure. Reducing the computational load and file resource volume allows for the development of lighter and more productive applications, which contributes to the accessibility of gaming and educational technologies even on devices with modest hardware specifications. In addition, such methods can be integrated into industrial and educational software, contributing to the formation of more sustainable and innovative digital ecosystems. Thus, the optimization of graphic models in the gaming industry not only increases productivity but also contributes to the development of a digital infrastructure that is consistent with the principles of sustainable industrialization and innovation.
Within the framework of SDG 12 (“Responsible Consumption and Production”), the results of the study show that optimizing 3D models can significantly reduce the digital resources required for their storage and use. Reducing file sizes and the number of polygons lowers the requirements for memory, network bandwidth, and disk storage, which is in line with the principles of more efficient use of computing resources. For developers, this means the ability to create games and applications with lower storage and distribution costs, and for users, it means reduced device load and more sustainable software use. Thus, polygon mesh optimization can be seen as an example of responsible digital production that minimizes excessive consumption of computing power and contributes to the more environmentally friendly development of the industry.
From a sustainability perspective, these results confirm the importance of optimizing digital assets. The energy efficiency of gaming applications is directly related to the duration of mobile device operation without recharging and the reduction of total energy consumption by data centers and end devices. Thus, the optimization of 3D models can be viewed as part of a broader Green IT paradigm aimed at reducing the carbon footprint of the digital industry. Taken together, this indicates that technical solutions in the field of 3D graphics can not only enhance the user experience, but also contribute to sustainable development by reducing the gaming industry's impact on the environment.
Our research has shown that the use of polygonal mesh optimization and texture baking techniques in Blender for integration into Unity allows for a significant reduction in computational and resource costs without a critical loss of visualization quality. The results of the experiment, obtained on desktop and mobile devices, confirmed that low-polygon models have significantly lower memory and computational resource requirements, while providing stable frame rates and comparable visual detail. This makes this approach an effective solution for developing mobile games, where limited hardware resources are particularly critical.
From a sustainable development perspective, the proposed methodological approach demonstrates the potential to reduce energy consumption, optimize digital resources, and increase the accessibility of technologies. The contribution of the study is directly related to SDG 7 (“Affordable and Clean Energy”), SDG 9 (Industry, Innovation and Infrastructure), and SDG 12 (Responsible Consumption and Production), as the optimization of 3D content contributes to more efficient use of resources and the development of innovative digital infrastructure. Thus, the optimization of game models can be seen not only as a technical improvement, but also as a step towards a more sustainable and environmentally responsible gaming industry.
One limitation of this study is that only one model was used, which was tested on a single mobile device and a single computer. It is difficult to scale its results to the entire gaming industry. It should also be added that energy consumption indicators are obtained from indirect sources. The applicability of the method in real games requires scaling to more complex and detailed scenes. Promising areas for future research include expanding the set of models and testing optimization on a larger number of assets and game scenes. The game scene can be tested on iOS, Android, and VR devices, and a comparison of the popular Unity and Unreal Engine game engines can be performed.
The authors have not employed any Generative AI tools. [13] J. G. Park, N. Dutt, H. Kim, and S. S. Lim, "HiCAP: Hierarchical FSM-based dynamic integrated CPU-GPU frequency capping governor for energy-efficient mobile gaming", in Proc. Int. Symp. Low Power Electron. Design (ISLPED), Aug. 2016, pp. 218–223.