The Advantage Analysis of GPUs Based on CUDA
DOI:
https://doi.org/10.54097/0ay1vd06Keywords:
CUDA, GPU, Parallel computing, Performance acceleration.Abstract
The development of computing technology has increasingly exposed the limitations of traditional central processing units (CPUs) and their architectures, especially handling data-intensive and highly paralleled computing tasks. This paper aims to analyze the advantages presented by graphic processing units (GPUs) based on Nvidia compute unified device architecture (CUDA). First, introduce the development background of GPUs, explaining the architecture conversion from specialized graphic processors to general-purposed parallel processors. The core research work takes a deep dive into the CUDA programming model, explaining its layered thread organization, memory structure, and software ecosystem. Then, a comparative analysis is conducted, detailing the outstanding advantages of CUDA-based GPUs over CPUs and other GPU methods in terms of performance, energy efficiency, programming flexibility, and ecosystem. In the end, the paper summarizes that the synergy between GPU hardware and CUDA software is the key driving force to break through modern computing bottlenecks and push forward the development of high-performance computing.
Downloads
References
[1] De Mol, Liesbeth. Turing Machines. The Stanford Encyclopedia of Philosophy, 2021.
[2] Armoni, Marco. Beyond Von Neumann. The Systolic Array Architecture. IEEE Computer Architecture Letters, 2020, 19(2): 45–48.
[3] Peddie, Jon. The History of the GPU—New Developments. Springer, 2022.
[4] Hennessy, John L., and David A. Patterson. Computer Architecture: A Quantitative Approach. 6th ed., Morgan Kaufmann, 2019.
[5] NVIDIA. SIGGRAPH 1999 Keynote. Proceedings of SIGGRAPH, 1999.
[6] NVIDIA. GeForce 256: the World's first GPU. Press Release, 1999.
[7] Peddie, Jon. The History of Visual Magic in Computers. Springer, 2022.
[8] ACM Transactions on Graphics. The Evolution of GPU Architectures: From Fixed-Function to Unified Shaders. ACM TOG, 2021, 40(4): 38.
[9] NVIDIA. CUDA: A Parallel Computing Platform and Programming Model. IEEE Micro, 2008, 28(2): 50–57.
[10] NVIDIA. CUDA Toolkit Documentation. Version 11.0, 2020.
[11] NVIDIA. NVIDIA H100 Tensor Core GPU Architecture. White Paper, 2022.
[12] Vaswani, A., et al. Attention Is All You Need. Advances in Neural Information Processing Systems (NeurIPS), 2017, 30.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Academic Journal of Science and Technology
This work is licensed under a Creative Commons Attribution 4.0 International License.
