Neural Methods for Adaptive Voltage and Frequency Scaling in Energy Aware Processors

Yihan Wang; Suchuan Xing

doi:10.54097/pkgmn637

Authors

Yihan Wang
Suchuan Xing

DOI:

https://doi.org/10.54097/pkgmn637

Keywords:

Adaptive voltage and frequency scaling, Deep reinforcement learning, Energy-aware computing, Neural network workload prediction, Processor power management, On-chip intelligence

Abstract

Dynamic Voltage and Frequency Scaling (DVFS) represents one of the most effective runtime mechanisms for balancing computational performance against energy consumption in modern processors. Traditional rule-based and heuristic-driven DVFS governors struggle to adapt to increasingly heterogeneous and unpredictable workload patterns characteristic of contemporary computing environments, including mobile devices, data centers, and edge deployments. This review examines the growing body of research applying neural methods—including recurrent neural networks, convolutional neural networks, deep reinforcement learning (DRL), and federated learning—to the adaptive DVFS problem. We survey the evolution from classical prediction-based approaches to policy-learning frameworks capable of jointly optimizing energy efficiency and Quality of Service (QoS) constraints. Our analysis reveals that DRL-based approaches consistently achieve energy reductions of 25–35% over fixed-frequency baselines while maintaining QoS violation rates below 4%, outperforming both traditional governors and supervised prediction models. We conclude by identifying open research directions in multi-objective optimization, on-device continual learning, and heterogeneous processor support.

Downloads

Download data is not yet available.

References

[1] Basireddy, K. R. (2019). Runtime energy management of concurrent applications for multi-core platforms (Doctoral dissertation, University of Southampton).

[2] Li, R., Pu, C., Tao, J., Li, C., Fan, F., Xiang, Y., & Chen, S. (2023). LLM-based frameworks for power engineering from routine to novel tasks. arXiv preprint arXiv:2305.11202.

[3] Lin, W., Shi, F., Wu, W., Li, K., Wu, G., & Mohammed, A. A. (2020). A taxonomy and survey of power models and power modeling for cloud servers. ACM Computing Surveys (CSUR), 53(5), 1-41.

[4] Kaveripakkam Raghuraman, A. K. (2025). An adaptive QoE-driven DVFS: A comprehensive approach to enhancing user experience and optimizing power across concurrent application workloads (Doctoral dissertation, St. Francis Xavier University).

[5] Yuan, H., Bi, J., Li, S., Zhang, J., & Zhou, M. (2024). An improved LSTM-based prediction approach for resources and workload in large-scale data centers. IEEE Internet of Things Journal, 11(12), 22816-22829.

[6] Kuzminsky, M. B. (2025). Around the conditional 4th generation of modern server processors AMD and Intel: their microarchitecture and the performance of the corresponding computing systems. Программные системы: теория и приложения, 16(5 (68)), 43-514.

[7] Sartor, A. L., Becker, P. H. E., Wong, S., Marculescu, R., & Beck, A. C. S. (2019, July). Machine Learning-Based Processor Adaptability Targeting Energy, Performance, and Reliability. In 2019 IEEE Computer Society Annual Symposium on VLSI (ISVLSI) (pp. 158-163). IEEE.

[8] Sarma, W., Dey, S., & Tiwari, S. (2022). Autonomous IoT: AI-Driven Edge Computing to Power Intelligent Decision-Making. International Journal of AI, BigData, Computational and Management Studies, 3(2), 52-61.

[9] Wan, J., Li, J., Imran, M., & Li, D. (2019). A blockchain-based solution for enhancing security and privacy in smart factory. IEEE Transactions on Industrial Informatics, 15(6), 3652-3660.

[10] Hou, X., Tang, X., Liu, J., Li, C., Liang, L., & Cheng, K. T. (2024). Wasp: Efficient power management enabling workload-aware, self-powered aiot devices. IEEE Transactions on Parallel and Distributed Systems, 35(8), 1400-1414.

[11] Althoubi, A., Alshahrani, R., & Peyravi, H. (2020, November). Tail latency in datacenter networks. In Symposium on Modelling, Analysis, and Simulation of Computer and Telecommunication Systems (pp. 254-272). Cham: Springer International Publishing.

[12] Bhuiyan, A., Liu, D., Khan, A., Saifullah, A., Guan, N., & Guo, Z. (2020). Energy-efficient parallel real-time scheduling on clustered multi-core. IEEE Transactions on Parallel and Distributed Systems, 31(9), 2097-2111.

[13] Wang, H., & Cao, Y. (2019). Event driven power consumption optimization control model of GPU clusters. Cluster Computing, 22(3), 965-979.

[14] Radhakrishnan, K., Swaminathan, M., & Bhattacharyya, B. K. (2021). Power delivery for high-performance microprocessors—Challenges, solutions, and future trends. IEEE Transactions on Components, Packaging and Manufacturing Technology, 11(4), 655-671.

[15] Feng, J., Fu, R., Song, Y., Wang, Q., Chen, C., & Cao, L. (2024). The Analytical Model of Hotspot Temperature and the Effects of Different Factors in 3-D Integration. IEEE Transactions on Components, Packaging and Manufacturing Technology, 14(10), 1761-1770.

[16] Dong, Z., Luo, Y., Meng, Y., Yan, C., Bi, Z., Zhu, K., ... & Zeng, X. (2025, October). LCTMwalk: GPU-Accelerated Transient Thermal Simulation for Liquid-Cooled 2.5 D/3D ICs via Random Walks on Circuit Networks of Modified Compact Thermal Models. In 2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD) (pp. 1-9). IEEE.

[17] Park, J. G., Dutt, N., & Lim, S. S. (2021). An interpretable machine learning model enhanced integrated cpu-gpu dvfs governor. ACM Transactions on Embedded Computing Systems (TECS), 20(6), 1-28.

[18] Lv, L., Wu, Z., Zhang, L., Gupta, B. B., & Tian, Z. (2022). An edge-AI based forecasting approach for improving smart microgrid efficiency. IEEE Transactions on Industrial Informatics, 18(11), 7946-7954.

[19] Guan, W., & Ababei, C. (2020, October). Algorithm-Hardware Co-Design in Computing Systems: From Embedded Systems to the Cloud. In 2020 11th International Green and Sustainable Computing Workshops (IGSC) (pp. 1-3). IEEE.

[20] Jiang, W., Yu, H., Zhang, J., Wu, J., Luo, S., & Ha, Y. (2020). Optimizing energy efficiency of CNN-based object detection with dynamic voltage and frequency scaling. Journal of Semiconductors, 41(2), 022406.

[21] Wen, H., Yu, J., Pan, G., Chen, X., Zhang, S., & Xu, S. (2022). A hybrid CNN-LSTM architecture for high accurate edge-assisted bandwidth prediction. IEEE Wireless Communications Letters, 11(12), 2640-2644.

[22] Martınez, J. F., Ipek, E., Mutlu, O., Caruana, R., & Redmond, W. A. RETROSPECTIVE: Self-optimizing Memory Controllers: A Reinforcement Learning Approach.

[23] Panda, P., Tripathy, A., & Bhuyan, K. C. (2024, April). Reinforcement learning-based dynamic voltage and frequency scaling for energy-efficient computing. In 2024 Third International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE) (pp. 1-6). IEEE.

[24] Wali, S., Khan, M. I., & Imran, M. (2025). Semantic-aware reinforcement and ensemble learning for signal management and anomaly detection in IoT systems. Scientific Reports, 15(1), 42594.

[25] Terven, J. (2025). Deep reinforcement learning: A chronological overview and methods. Ai, 6(3), 46.

[26] Song, S., & Das, A. (2020). A case for lifetime reliability-aware neuromorphic computing. arXiv preprint arXiv:2007.02210.

[27] Peluso, V., Rizzo, R. G., & Calimera, A. (2019). Performance profiling of embedded convnets under thermal-aware dvfs. Electronics, 8(12), 1423.

[28] Atienza, D., Zhu, K., Huang, D., & Costero, L. (2025). A 20-year retrospective on power and thermal modeling and management. IEEE Design & Test.

[29] Monil, M. A. H., Belviranli, M. E., Lee, S., Vetter, J. S., & Malony, A. D. (2020, September). Mephesto: Modeling energy-performance in heterogeneous socs and their trade-offs. In Proceedings of the ACM International Conference on Parallel Architectures and Compilation Techniques (pp. 413-425).

[30] Donyanavard, B., Mück, T., Rahmani, A. M., Dutt, N., Sadighi, A., Maurer, F., & Herkersdorf, A. (2019, October). Sosa: Self-optimizing learning with self-adaptive control for hierarchical system-on-chip management. In Proceedings of the 52nd annual IEEE/ACM international symposium on microarchitecture (pp. 685-698).

[31] Joshi, S., Pandey, N. K., Diwakar, M., & Mishra, A. K. (2024, December). Reinforcement learning-driven auto scaling for sla enhancement in cloud environment. In 2024 Eighth International Conference on Parallel, Distributed and Grid Computing (PDGC) (pp. 634-639). IEEE.

[32] Osypanka, P., & Nawrocki, P. (2022). QoS-aware cloud resource prediction for computing services. IEEE Transactions on Services Computing, 16(2), 1346-1357.

[33] Yang, T., Feng, X., Cai, S., Niu, Y., & Pen, H. (2024). A privacy-preserving federated reinforcement learning method for multiple virtual power plants scheduling. IEEE Transactions on Circuits and Systems I: Regular Papers, 72(4), 1939-1950.

[34] Rojahn, M., & Grum, M. (2025). Green AI: A systematic review and meta-analysis of its definitions, lifecycle models, hardware and measurement attempts. arXiv preprint arXiv:2511.07090.

[35] Krzywaniak, A., Czarnul, P., & Proficz, J. (2022). DEPO: A dynamic energy‐performance optimizer tool for automatic power capping for energy efficient high‐performance computing. Software: Practice and Experience, 52(12), 2598-2634.

[36] Eeckhout, L. (2022). Performance Metrics. In Computer Architecture Performance Evaluation Methods (pp. 5-14). Cham: Springer International Publishing.

[37] Dey, S. (2023). Optimizing Power & Reliability in Mobile Computing with DVFS. Somdip Dey.

[38] Lu, Y. (2022). Workload Prediction and Resource Management for Energy Efficiency in Cloud Data Centres (Doctoral dissertation, University of Leicester).

[39] Zhang, M., Li, L., Wang, H., Liu, Y., Qin, H., & Zhao, W. (2019). Optimized compression for implementing convolutional neural networks on FPGA. Electronics, 8(3), 295.

[40] Zhou, T., & Lin, M. (2023). CPU frequency scheduling of real-time applications on embedded devices with temporal encoding-based deep reinforcement learning. Journal of Systems Architecture, 142, 102955.

[41] Zafar, R., Huang, P., & Sun, Y. (2025). Enhancing electric vehicle charging load prediction in data-scarce scenarios: A hybrid deep learning-based approach integrating clustering analysis and transfer learning. Energy and AI, 21, 100545.

[42] Chu, Z., Li, B., & Ying, C. (2025). Transformer-based latency prediction for stream processing task. Journal of King Saud University Computer and Information Sciences, 37(5), 106.

[43] Chandrasiri, S., & Meedeniya, D. (2026). Multi-Objective Cloud Workflow Scheduling with Two-Stage Deep Reinforcement Learning and Graph Neural Networks. IEEE Access.

[44] Zhang, H., Peng, J., Tan, H., Dong, H., & Ding, F. (2020). A deep reinforcement learning-based energy management framework with Lagrangian relaxation for plug-in hybrid electric vehicle. IEEE Transactions on Transportation Electrification, 7(3), 1146-1160.

[45] Ma, H., Liu, Q., Liu, H., Sun, Z., Gu, C., & Huang, H. (2025, July). Energy-Aware Task Scheduling Using DVFS and On/Off Switching in Data Center. In International Conference on Intelligent Computing (pp. 517-528). Singapore: Springer Nature Singapore.

[46] Samadi, N., Yadollahi, F., & Shah-Mansouri, H. (2025). iDEAS: Intelligent DVFS for Energy-Efficient Task Scheduling in Mobile Devices With big. LITTLE Computing Architecture. IEEE Access, 13, 200711-200724.

[47] Zhang, S., Qiu, L., & Zhang, H. (2025). Edge cloud synergy models for ultra-low latency data processing in smart city iot networks. International Journal of Science, 12(10).

[48] Machmudi, M. A., Putra, Y. W. S., & Naim, A. G. (2025). Proximal Policy Optimization for Adaptive Resource Allocation in Mobile OS Kernels: Enhancing Multitasking Efficiency. Jurnal Teknik Informatika (Jutif), 6(6), 5621-5631.

[49] Du, Y., Liu, J., Wang, M., Li, W., & Zeng, X. (2025). DS-HPPO Deep Reinforcement Learning for Optimal DVFS Control on an Image Signal Processor. IEEE Access.

[50] Bi, H., & Gelenbe, E. (2019). Emergency management systems and algorithms: a comprehensive survey. arXiv preprint arXiv:1907.04136.

[51] Benila, S., & Devi, K. (2025). Federated synergy: Hierarchical multi-agent learning for sustainable edge computing in IIoT. IEEE Access.

[52] Vega, D., Arvidsson, A., & Saïah, F. (2023). Resilient supply management systems in times of crisis. International Journal of Operations & Production Management, 43(1), 70-98.

[53] Li, Z. (2024). Investigation of reconfigurable hardware acceleration for low-power embedded neural networks (Doctoral dissertation, Université Côte d'Azur).

[54] Xu, Z., Chen, X., Tang, W., Lai, J., & Cao, L. (2021). Meta weight learning via model-agnostic meta-learning. Neurocomputing, 432, 124-132.

[55] Kim, S., Kim, S., Um, S., Kim, S., Kim, K., & Yoo, H. J. (2023). Neuro-CIM: ADC-less neuromorphic computing-in-memory processor with operation gating/stopping and digital–analog networks. IEEE Journal of Solid-State Circuits, 58(10), 2931-2945.

[56] Kulkarni, U., Meena, S. M., Gurlahosur, S. V., & Bhogar, G. (2021). Quantization friendly mobilenet (qf-mobilenet) architecture for vision based applications on embedded platforms. Neural Networks, 136, 28-39.

[57] Wong, J., Chen, P. W., Zhang, T., Gonzalez, J., Tian, Y., & Seshia, S. A. (2023). Ashera; Neural Guided Optimization Modulo Theory. Tech. rep. UCB/EECS-2023-103. EECS Department, University of California, Berkeley.

[58] Batra, N., Kukunuri, R., Pandey, A., Malakar, R., Kumar, R., Krystalakos, O., ... & Parson, O. (2019, November). Towards reproducible state-of-the-art energy disaggregation. In Proceedings of the 6th ACM international conference on systems for energy-efficient buildings, cities, and transportation (pp. 193-202).

[59] Ranjbar, B., Singh, A. K., Sahoo, S. S., Dziurzanski, P., & Kumar, A. (2023). Power management of multicore systems. In Handbook of Computer Architecture (pp. 1-33). Singapore: Springer Nature Singapore.

[60] Daini, C. (2025). Predictable CNN Inference and Real-Time Scheduling on Multicore Embedded Platforms (Doctoral dissertation, Université de Lille).

[61] Radhoush, S., Whitaker, B. M., & Nehrir, H. (2023). An overview of supervised machine learning approaches for applications in active distribution networks. Energies, 16(16), 5972.

[62] Hebbar, R., & Milenković, A. (2022). PMU-events-driven DVFS techniques for improving energy efficiency of modern processors. ACM Transactions on Modeling and Performance Evaluation of Computing Systems, 7(1), 1-31.

[63] Kazemi, F., Barzegar, B., Motameni, H., & Yadollahzadeh-Tabari, M. (2025). An energy-aware scheduling in DVFS-enabled heterogeneous edge computing environments. The Journal of Supercomputing, 81(9), 1078.