ANALISIS KINERJA GALAXY AI PADA SAMSUNG S24 UNTUK TUGAS PEMROSESAN BAHASA ALAMI HARIAN BERBASIS DEEP LEARNING
PERFORMANCE ANALYSIS OF GALAXY AI ON THE SAMSUNG S24 FOR DAILY DEEP LEARNING–BASED NATURAL LANGUAGE PROCESSING TASKS
DOI:
https://doi.org/10.52972/hoaq.vol17no1.p182-186Keywords:
Galaxy AI, Pemrosesan Bahasa Alami, Deep Learning, Analisis Kinerja, Neural Processing UnitAbstract
Penelitian ini mengevaluasi sejauh mana Galaxy AI mengelola tugas pemrosesan bahasa alami sehari-hari pada seri Samsung S24 berdasarkan deep learning. Fokus utama penelitian ini mencakup tiga area: pengelompokan teks, evaluasi sentimen, dan pengambilan informasi. Pendekatan penelitian menggunakan teknik transfer learning dengan model berbasis transformer yang disesuaikan untuk unit pemrosesan neural (NPU) seri S24. Pengujian komprehensif dilakukan dengan mengukur indikator seperti akurasi model, waktu respons inferensi, penggunaan energi, dan pemanfaatan sumber daya sistem selama proses inferensi di perangkat. Hasil penelitian menunjukkan kinerja yang luar biasa: tingkat akurasi mencapai 94,3% untuk pengelompokan teks, 89,7% untuk evaluasi sentimen, dan 86,2% untuk pengambilan informasi. Waktu respons inferensi rata-rata adalah 0,8 detik per 100 kata dengan penggunaan energi efisien berkisar antara 285 hingga 320 mW. Analisis lebih lanjut menunjukkan pemanfaatan NPU optimal hingga 78% dengan manajemen panas yang efektif yang menjaga stabilitas kinerja selama operasi berkelanjutan. Desain canggih sistem ini mengintegrasikan perangkat keras dan perangkat lunak melalui pembagian beban komputasi yang efisien antara NPU dan CPU. Hasil ini membuktikan bahwa Galaxy AI pada Samsung S24 memiliki kemampuan superior dalam menangani tugas pemrosesan bahasa alami.
This study evaluates how well Galaxy AI manages everyday natural language processing tasks on the Samsung S24 series based on deep learning. It focuses on three main areas: text clustering, sentiment evaluation, and information retrieval. The research approach employs transfer learning techniques with a transformer-based model tailored for the S24 series's neural processing unit (NPU). Comprehensive testing was conducted by measuring indicators such as model accuracy, inference response time, energy usage, and system resource utilization during inference on the device. The research findings reveal outstanding performance: accuracy rates reached 94.3% for text clustering, 89.7% for sentiment evaluation, and 86.2% for information retrieval. The average inference response time was 0.8 seconds per 100 words with efficient energy usage ranging from 285 to 320 mW. Further analysis showed optimal NPU utilization of up to 78% with effective heat management that maintains performance stability during continuous operation. The system's advanced design integrates hardware and software through efficient computational load sharing between the NPU and CPU. These results prove that the Samsung S24's Galaxy AI has superior capabilities in handling natural language processing tasks.
References
J. Hanhirova, T. Kämäräinen, S. Seppälä, M. Siekkinen, V. Hirvisalo, and A. Ylä-Jääski, “Latency and throughput Characterization of convolutional Neural Networks for mobile Computer Vision,” MMSys '18: Proceedings of the 9th ACM Multimedia Systems Conference. MMSys 2018, pp. 204–215, 2018, doi: 10.1145/3204949.3204975.
A. Ignatov., et.al., “AI Benchmark: Running deep Neural Networks on android Smartphones,” Lecture Notes in Computer Science, vol. 11133 LNCS, pp. 288–314, 2019, doi: 10.1007/978-3-030-11021-5_19.
A. Marchisio., M. A. Hanif., F. Khalid., G. Plastiras., C. Kyrkou., and T. Theocharides., “Deep Learning for Edge Computing: Current Trends, Cross-Layer Optimizations, and Open Research Challenges,” in 2019 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), 2019, pp. 553–559. doi: 10.1109/ISVLSI.2019.00105
A. Howard., et.al, “Searching for MobileNetV3 Accuracy vs MADDs vs model Size,” 6th International Conference on Computer Vision., pp. 1314–1324, 2019.
Q. V. Le Mingxing Tan, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks Mingxing,” Canadian Journal of Emergency Medicine., vol. 15, no. 3, p. 190, 2013.
C.-J. Wu et al., “Machine Learning at Facebook: Understanding Inference at the Edge,” in 2019 IEEE International Symposium on High Performance Computer Architecture (HPCA), 2019, pp. 331–344. doi: 10.1109/HPCA.2019.00048.
H. Cai, C. Gan, T. Wang, Z. Zhang, and S. Han, “Once-for-All: Train One Network and Specialize It for Efficient Deployment,” 8th International Conference. Learning. Representation. ICLR 2020, pp. 1–15, 2020. doi: 10.48550/arXiv.1908.09791
H. Wang et al., “HAT: Hardware-Aware Transformers for Efficient Natural Language Processing,” Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 7675–7688, 2020, doi: 10.18653/v1/2020.acl-main.686.
Vaswani et al., “Attention Is All You Need,” NeurIPS Proc., 2017.
Yang, T. J., et al. “Netadapt: Platform-Aware Neural Network Adaptation for Mobile Applications,” European Conference on Computer Vision., pp. 285–300, 2020. doi: 10.1007/978-3-030-01249-6_18.
“Snapdragon® 8 Gen 3 Mobile Platform,” pp. 2–3, [Online]. Available: https://docs.qualcomm.com/bundle/publicresource/87-71408-1_REV_G_Snapdragon_8_gen_3_Mobile_Platform_Product_Brief.pdf?
N. K. Manaswi, et.al. "RNN and LSTM. In: Deep Learning with Applications Using Python". Apress, Berkeley, CA., pp. 115–126, 2018, doi: 10.1007/978-1-4842-3516-4_9
V. J. Reddi et al., “MLPerf Inference Benchmark,” 2020 ACM/IEEE 47th Annual International Symposium on Computer Architecture (ISCA)., pp. 446–459, 2020, doi: 10.1109/ISCA45697.2020.00045.
S. Raschka, “Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning,” 2020, [Online]. Available: http://arxiv.org/abs/1811.12808
R. D. Peng, “Reproducible Research in Computational Science,” Science, vol. 334, no. 6060, pp. 1226–1227, 2011, doi: 10.1126/science.1213847.
Y. Hang and H. Kabban, “Thermal Management in Mobile Devices: Challenges and Solutions,” in 2015 31st Thermal Measurement, Modeling & Management Symposium (SEMI-THERM), 2015, pp. 46–49. doi: 10.1109/SEMI-THERM.2015.7100138.
C. W. Ramya et al., “Sustainable AI: Environmental Implications, Challenges and Opportunities,” 2022.
M. Sandler, A. Howard, M. Zhu, and A. Zhmoginov, “Sandler_MobileNetV2_Inverted_Residuals_CVPR_2018_paper.pdf,” arXiv, pp. 4510–4520, 2018.
S. Han, H. Mao, and W. J. Dally, “Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding,” 4th International Conference on Learning Representations, ICLR 2016., pp. 1–14, 2016.
A. Zhou, A. Yao, Y. Guo, L. Xu, and Y. Chen, “Incremental Network Quantization: Towards Lossless CNNS with Low-Precision Weights,” 5th International Conference on Learning Representations., pp. 1–14, 2017.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 HOAQ (High Education of Organization Archive Quality) : Jurnal Teknologi Informasi
This work is licensed under a Creative Commons Attribution 4.0 International License.
Jurnal HOAQ (High Education of Organization Archive Quality) : Jurnal Teknologi Informasi diterbitkan berdasarkan lisensi Creative Commons Attribution 4.0 International License (CC BY 4.0). Lisensi ini memungkinkan setiap orang untuk Berbagi: menyalin dan mendistribusikan kembali materi ini dalam format atau bentuk apapun; Adaptasi: merombak, mengubah, dan membuat turunan dari materi ini untuk keperluan apa pun, termasuk keperluan komersial, asalkan mereka memberikan pengakuan kepada Penulis Asli atas hasil karya aslinya.
