Parallel Technology and Development in the Gaming Industry

Zixuan Xu

School of Advanced Technology, Xi’an Jiaotong-Liverpool University,

111 Ren 'ai Road, Suzhou Industrial Park, Huqiu District, Suzhou, China

Keywords: Parallel Technology, Multi-Core, Multi-Thread, Game Industry.

Abstract: With the rapid development of the game industry, the traditional single-core simple logic games have been

unable to meet the needs of teenagers, so the addition of parallel technology has changed the status quo. The

use of multi-threading, multi-core, and the use of gpu and cpu improves graphics rendering, computing power,

and physics simulation in the game industry. This article provides an overview of the application and

development of parallel technical in the gaming industry. It gives readers a deeper understanding of

parallelism and its impact on the game industry. The focus is on the underlying concept of parallel technique,

for instance: multi-threading, multi-core, single instruction multiple data (SIMD), and the use of cpu and gpu.

Furthermore, the article will also discuss about several applications of parallel techniques in gaming industry,

for example: image rendering, route planning of Non-player characters (NPCS), and parallel processing of

some large online games and so on. Finally, the future development of parallel technology will also be

mentioned, such as how to combine with Artificial Intelligence (AI) to make AI more intelligent, and how to

combine with AR technology to enhance the authenticity of the virtual world and reduce the delay of

Augmented Reality (AR) devices.

1 INTRODUCTION

As the game industry rapidly evolves, traditional

single-core simple logic games are increasingly

inadequate in meeting the demands of teenagers; thus,

the integration of parallel technology has transformed

the existing landscape. Parallel computing represents

a computational paradigm that harnesses multiple

processing elements concurrently to address complex

problems with greater efficiency than traditional

serial computing methods. This approach is

fundamentally driven by the necessity to manage

large-scale computations and data processing tasks

that surpass the capabilities of single-threaded

processors. The origins of parallel computing can be

traced back to the formative years of computer

science; however, it gained substantial traction with

the emergence of multi-core processors and

distributed computing systems. In this framework,

tasks are partitioned into smaller subtasks executed

simultaneously across various processors or cores.

This division of labor not only accelerates

computation but also enhances application

https://orcid.org/0000-0001-5437-6650

scalability. Parallel computing encompasses diverse

architectures and models, including multi-core

executing, muti-thread model, the together usage of

CPU and GPU. With the continuous improvement of

the demand for game processing performance, the

demand for game picture quality improvement,

parallel computing has become essential across game

domains. As technology continues to advance,

innovations in parallel computing promise to further

augment performance while unlocking new potential

within computational capabilities. So, this article will

about the underlying definition of several parallel

techniques as well as their applications in the game

industry which will give readers a deeper

understanding of parallelism and its impact on the

game industry. Decades years ago, AlBahnassi,

Mudur and Goswami have shown that in recent years,

several existing games have begun to transition

towards supporting multiple processors, so it seems

that the game industry is moving toward

parallelization (Albahnassi, Mudur, Goswami, 2012).

The main reason why game manufacturers prefer to

develop parallel games is not only because parallel

232

Xu and Z.

Parallel Technology and Development in the Gaming Industry.

DOI: 10.5220/0013514900004619

In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning (DAML 2024), pages 232-238

ISBN: 978-989-758-754-2

technology can save a certain investment cost, but

also because it can greatly improve the performance

of the game and improve the processing efficiency of

the computer. And Venu has demonstrated that

various advanced techniques have been developed to

enhance performance, including parallel processing,

data-level parallelism, and instruction-level

parallelism, all of which have demonstrated

significant effectiveness (Venu, 2011). For example,

for multi-core processing venu has also shown that

multi-core processors were available for over a

decade; however, their significance has increased

recently due to the technological limitations faced by

single-core processors, such as challenges in

achieving high throughput and prolonged battery life

with enhanced energy efficiency (Venu, 2011).

Additionally, as for multi-thread Tulip, Bekkema,

Nesbitte also found out that although current game

engines are extensively optimized for efficient

operation on single-processor architectures and have

largely avoided multi-threaded design approaches.

Nevertheless, the forthcoming generation of game

engines must tackle the intricacies of multi-threaded

programming to fully leverage the performance

capabilities of emerging PC and gaming console

platforms (Tulip, Bekkema, Nesbitte, 2006).

Furthermore, the game industry also applied the

parallel processing capabilities of GPUs, thereby

markedly enhancing graphical performance and

visual fidelity in gaming applications. In the

remaining parts of the article, it will be divided into

five small parts, firstly some Overview of parallel

techniques, then the detailed development of parallel

technology, after that some applications of parallel

technology in game development, fourthly the

prospect of future development and lastly the

conclusion.

2 OVERVIEW OF PARALLEL

TECHNIQUES

Nowadays, with the development of parallel

technology, many different parallel technologies have

been derived. Examples include the use of multiple

cores, the use of multiple threads, or the combination

of CPUs and GPUs. Most parallel techniques can

greatly improve the speed, speed, and quality of a

computer if they are used in the right amount which

can definitely improve the quality of the game

industry.

2.1 Multi-core

Multi-core technology refers to the integration of

multiple processing cores on a processor chip. Each

core is able to perform tasks independently, which

can greatly improve computational performance. And

nowadays, the processor may have two, four, six,

eight or even more core to accelerate the processing

quality. Each core can deal with the threads and tasks

independently. Figure 1 below shows the simple

architecture of a computer with 4 core processor.

Figure1: a computer with 4 core processor.（Photo/Picture

credit : Original ）

Memory control center (MCH), also known as

the North bridge chip. Input/output control hub (ICH),

also known as the South bridge chip. MCH mainly

provides support for CPU, memory and other devices.

ICH mainly supports peripheral devices such as

keyboards and interfaces. And MEM is the memory

which saves the data. The use of multi-core

processors can significantly improve the multitasking

ability and parallel computing performance. By

sharing the workload, the individual cores can operate

at a lower frequency, resulting in lower power

consumption. The underlying working principle is

that the single core in multi-core processor may not

be powerful than the single-core processor. But it can

enhance the overall performance by handling

multiple tasks at the same time. A single-core

processor will distribute different time slices to

different program when dealing with multi-program.

But once if one program takes longer time to

complete, other processed start to lag behind.

However, for multi-core, each task can be executed

by separate core at the same time which can largely

improve the performance as shown in Figure 2 (Geer,

2005).

Parallel Technology and Development in the Gaming Industry

233

Figure 2: Multicore chips perform better – based on Intel

tests using the SPECint2000 and SPECfp2000 benchmarks

– than single-core processors (Geer, 2005).

Furthermore, the frequency is not high for multi-

core in the chip, but the paralleling processing model

can make greater performance and reduce the power

consumption as shown in figure 3.

Figure 3: Accumulated energy consumption for different

numbers of cores during network average latency test with

a 5 ms limit (Oliveira, Xavier-De-souza, Silveira, 2021).

So, this explains why numerous game developers

favor multi-core processing.

2.2 Multi-threading

Multi-threading technology encompasses the

concurrent execution of multiple threads within a

single processor core. A thread represents an

autonomous path of execution for a program, while

multi-threading facilitates the sharing of resources—

such as memory and file descriptors—among multiple

threads operating under the same process. There are

two main models for multi-threading. One is multi-

threading working on a single processor. In a single-

core processor, the thread scheduler plays a crucial

role in distributing CPU time among multiple threads.

Given that there is only one core available, true

simultaneous execution of multiple threads is not

feasible; therefore, the scheduler rapidly alternates

between thread executions, ensuring that each thread

has an opportunity to utilize the CPU. Although this

model is not really working in parallel, it still improve

the usage of the cpu that is why a single processor

model can still accelerate the processing efficiency.

For the second model, multi-threading working on

multiple processor, it can realize parallel operation in

the real sense. Tulip, Bekkema and Nesbitt also found

that for a fixed number of cpu processor, the user can

increase the number of threads to further accelerate

the processing speed. However since each of the cpu

processor has maximum processing bottlenecks, the

user needs to increase the number of cpu to enhance

the processing speed and energy efficiency which

shown in Figure 4 and Figure 5 below (Zecena,

Burtscher, Jin et al, 2013).

Figure 4: Runtime of NBOMP with 100,000 bodies and 10

timesteps on System 1 (Zecena, Burtscher, Jin et al, 2013).

Figure 5: Energy consumption of NBOMP with 100,000

bodies and 10 time steps on System 1 (Zecena, Burtscher,

Jin et al, 2013)

2.3 CPU-GPU hybrid technology

CPU-GPU hybrid technology leverages the

synergistic collaboration between central processing

units (CPUs) and graphics processing units (GPUs) to

significantly enhance computational performance.

Due to their extensive parallel processing capabilities,

GPUs are particularly adept at handling tasks that

demand substantial parallel computation. In real life,

computers usually improve the processing speed of

computers by reasonably allocating computer tasks to

CPU and GPU, such as uploading some content with

large memory or for deep learning and scientific

DAML 2024 - International Conference on Data Analysis and Machine Learning

234

Figure 6: Sort benchmark. Faster GPUs are affected more by the memory transfer overhead. For instance, when sorting 64M

values, the application time on the GTX 480 is 3.6x slower than the kernel itself (Gregg, Hazelwood, .2011).

Table 1: GPUs tested.

GPU Type Compute

Capabilit

Cores Memory(MB) Clock(MHz) Host-Dev

BW(Mb/s)

DDev-Host

BW(Mb/s)

Tesla

C2025

2.0 480 3072 1150 2413.9 2359.2

GTX 480 2.0 480 1024 1401 1428.0 1354.2

9800 GT 1.1 112 1024 1500 2148.8 1502.5

330M 1.2 48 256 1265 2396.2 2064.7

All GPUs are from Nvidia and run the Cuda programming language. the 330m GPU is a laptop GPU and the others are

desktop GPUs. “host-dev” shows transfer times from main memory to GPU memory, and “dev-host” shows transfer times

from GPU memory to main memory (Gregg, Hazelwood, 2011).

computing. As shown in figure 6 and table 1 below,

they demonstrate that cpu and gpu work together and

show that gpu with more cores has higher processing

efficiency to further improve data processing speed.

3 APPLICATION OF PARALLEL

TECHNOLOGY IN GAME

DEVELOPMENT

There are numerous applications of parallel

technology in the game industry. This article will only

discuss three main states: graphic rendering, AI route

or instruction planning and massive online gaming.

3.1 Graphic Rendering

Various forms of parallelism can be employed in the

rendering process, including data decomposition, task

assignment and the acceleration of GPU. For data

decomposition, a prevalent strategy involves

partitioning the image into smaller segments. Each

segment is processed independently by distinct

processing units, such as CPU cores or GPU threads.

This approach effectively leverages computational

resources and mitigates the burden on individual

processing units. While for complex 3d scene, the

scene will be divided into several parts, each part will

be rendered in separated processing unit to achieve

the parallel processing. Then for the task assignment,

during the rendering process, geometric processing

tasks—such as transformations, clipping, and lighting

calculations — can be decomposed into multiple

parallelizable components. For instance, each task

associated with geometric processing for a given

model may be allocated to distinct threads or

processing units. The rasterization process transforms

geometric shapes into pixels, enabling parallel

execution across multiple threads. As for the GPU

acceleration. The developer usually applies GPU

parallel architectural CUDA or OpenCL to deal with

numerous threads at the same time which will largely

increase the speed of rendering. Figure 7 below shows

two different types of parallel rendering model and

Figure 8 demonstrates the frame improvement of

different number of GPUS.

Parallel Technology and Development in the Gaming Industry

235

Figure 7: The rendering time variation under different

rendering models (Zhang, Ma, Qiu, et al, 2023).

Figure 8: performance of the multi-GPU system under

different number of GPUS (Zhang, Ma, Qiu, et al, 2023).

3.2 AI Route or Instruction Planning

There are various parallel methods that can be applied

to Ai route or instruction planning. For example,

multi-thread and multi-progress. In addressing

complex path planning challenges, such as navigation

tasks within extensive environments, multi-threading

enables the simultaneous computation of multiple

potential paths to identify the optimal route. Each

thread can manage a specific sub-region or explore

various algorithms, thereby reducing overall

computational time. In decades years ago, Sanci has

also shown the result of the acceleration of route

algorithm by applying multi-threads in figure 9. In

systems requiring the simultaneous execution of

multiple instructions, such as automated production

lines, multi-process architectures can manage distinct

tasks or instructions in parallel, thereby enhancing the

overall operational efficiency of the system. For some

NPC in the game, the constructor usually applies

reinforce learning to train the ai to explore and study

independently. In the training of agents such as those

used in gaming or robotics, parallel computing

facilitates the simultaneous execution of multiple

environment instances, thereby expediting the

learning process. Each instance operates

independently to explore and learn, with subsequent

results aggregated for model updates.

Figure 9: Effect of Using Different Number of Threads for

the Genetic Algorithm in the GPU (Sanci, 2010).

3.3 Massive Online Games

For large online games, it usually requires handling

complex game logic, such as the analysis of players

methods, the update of players’ mission, or some

event trigger. This progress will be divided into

several processors or various threads in order to

improve the speed of game logic processing and

reduce the latency due to the large number of players.

Abdelkhalek and Bilas have found out that for more

threads or more processor server, the average time of

response is lower than those which have reach the

bottleneck of the peak value of player. But when both

servers not achieving the bottleneck, the respond time

will be almost the same (Abdelkhalek, Bilas, 2004).

And Figure 10 shows the response rate of sever due

to the player. Figure 11 demonstrates the average

respond time. In each figure different line means

different numbers of threads.

Figure 10: Total server response rate (Abdelkhalek, Bilas,

2004).

DAML 2024 - International Conference on Data Analysis and Machine Learning

236

Figure 11: Average server response time (Abdelkhalek,

Bilas, 2004).

Furthermore, it is also significant for massive

online games to make multiple data synchronization.

So when applying parallel processing technique, the

server can deal with numerous data packages at the

same time which can reduce the internet latency and

maintain the synchronization of the data. Additionally,

in a massive online game, the artificial intelligence

needs to process the behavior decisions of multiple

NPCS (non-player characters) at the same time. So

parallel computing is of great importance to evaluate

the behavior logic of multiple NPCS at the same time

to improve the intelligence level and response speed

of AI.

4 FUTURE DEVELOPMENT

The past decades years of this new kinds of

technology (parallel technology) was mainly applied

on some basic online games. And nowadays with the

development of artificial intelligence, the game

industry has a trend to develop smarter Ai and smart

computing. it is also realizable for users to play game

without the touch of their fingers with the

development of virtual reality in the future. There also

has been some progress in virtual reality (VR). By

applying DLoVe and other parallel systems,

Deligiannidis and Jacob have already found that the

application of DLoVe can largely improve the overall

performance of applications and increase the average

framework, it can also be used to provide mechanisms

for the implementation and transformation of single-

user programs into multi-user applications which is

useful when developing some large online game on

virtual world in the future (Deligiannidis, Jacob,

2005).

Moreover, this article only discussed about

several parallel technologies and some applications

on game industry, a more detailed description of these

contents will be improved and supplemented in the

future.

5 CONCLUSION

In conclusion, this article has discussed several basic

parallel technologies for instance: multi-core, multi-

threads and cpu-gpu hybrid technology. It also talked

about some applications in the game industry. For

example, graphic rendering, ai route or instruction

planning and large online games. Moreover, this

article also explores the future integration of artificial

intelligence, virtual reality, and parallel computing.

Ultimately, several drawbacks emerged that require

resolution in the future, including addressing

compatibility issues between parallel techniques and

hardware, as well as ensuring data consistency.

REFERENCES

Abdelkhalek, A., & Bilas, A. 2004. Parallelization and

performance of interactive multiplayer game servers.

In Proceedings of the 18th International Parallel and

Distributed Processing Symposium (pp. 72-). IEEE.

AlBahnassi, W., Mudur, S. P., & Goswami, D. 2012. A

design pattern for parallel programming of games.

In 2012 IEEE 14th International Conference on High

Performance Computing and Communication & 2012

IEEE 9th International Conference on Embedded

Software and Systems (pp. 1007-1014). IEEE.enu, B.

(2011). Multi-core processors: An overview. arXiv.

Deligiannidis, L., & Jacob, R. J. K. 2005. Improving

performance of virtual reality applications through

parallel processing. Journal of Supercomputing, 33(2),

155–173.

Geer, D. 2005. Chip makers turn to multicore processors.

Computer, 38(5), 11-13.

Gregg, C., & Hazelwood, K. 2011. Where is the data? Why

you cannot debate CPU vs. GPU performance without

the answer. In Proceedings of the IEEE International

Symposium on Performance Analysis of Systems and

Software (pp. 134-144). IEEE.

Oliveira, T. F., Xavier-De-souza, S., & Silveira, L. F.

2021. Improving energy efficiency on SDN control-

plane using multi-core controllers. Energies, 14(11).

https://doi.org/10.3390/en14113161

Sanci, S. 2010. A parallel algorithm for flight route

planning on gpu using cuda [M.S. - Master of Science].

Middle East Technical University.

Tulip, J., Bekkema, J., & Nesbitt, K. 2006. Multi-threaded

game engine design. In Proceedings of the 3rd

Australasian Conference on Interactive Entertainment

(pp. 9–14). Murdoch University.

Zecena, I., Burtscher, M., Jin, T., & Zong, Z. 2013.

Evaluating the performance and energy efficiency of

Parallel Technology and Development in the Gaming Industry

237

n-body codes on multi-core CPUs and GPUs. 2013

IEEE 32nd International Performance Computing and

Communications Conference (IPCCC), Performance

Computing and Communications Conference

(IPCCC), 2013 IEEE 32nd International, 1–8.

https://doi.org/10.1109/PCCC.2013.6742789

Zhang, H., Sheng, B., Bi, L., Kim, J., Magnenat-Thalmann,

N., & Thalmann, D. (Eds.). 2024. Multi-GPU parallel

pipeline rendering with splitting frame. In Advances in

computer graphics: CGI 2023 (Vol. 14496, pp. 225-

238). Springer.

DAML 2024 - International Conference on Data Analysis and Machine Learning

238