The Impact of GPU DVFS on the Energy and Performance of Deep Learning: an Empirical Study
Abstract
Over the past years, great progress has been made in improving the computing power of general-purpose graphics processing units (GPGPUs), which facilitates the prosperity of deep neural networks (DNNs) in multiple fields like computer vision and natural language processing. A typical DNN training process repeatedly updates tens of millions of parameters, which not only requires huge computing resources but also consumes significant energy. In order to train DNNs in a more energy-efficient way, we empirically investigate the impact of GPU Dynamic Voltage and Frequency Scaling (DVFS) on the energy consumption and performance of deep learning. Our experiments cover a wide range of GPU architectures, DVFS settings, and DNN configurations. We observe that, compared to the default core frequency settings of three tested GPUs, the optimal core frequency can help conserve 8.7%~23.1% energy consumption for different DNN training cases. Regarding the inference, the benefits vary from 19.6%~26.4%. Our findings suggest that GPU DVFS has great potentials to help develop energy efficient DNN training/inference schemes.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The default frequency settings on tested GPUs, such as NVIDIA's Pascal P100 and Volta V100, are often not optimized for energy efficiency in DNN training and inference.
“Moreover, the authors show that the default frequency settings on tested GPUs, such as NVIDIA's Pascal P100 and Volta V100, are often not optimized for energy efficiency in DNNtraining and inference.”
From Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models (2020) · cites this paper as Tang et al., 2019
DVFS can be applied to GPUs to help conserve about 8.7% to 23.1% energy consumption for training different DNNs and about 19.6% to 26.4% for inference.
“Tang et al. (2019) show that DVFS can be applied to GPUs to help conserve about 8 . 7 %to 23 . 1 %energyconsumptionfortrainingdifferent DNNs and about 19 . 6 %to 26 . 4 %for inference.”
From Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models (2020) · cites this paper as Tang et al., 2019
Power management techniques like dynamic voltage and frequency scaling (DVFS) can help conserve energy consumption and for some models even reduce time to reach convergence.
“Power management techniques like dynamic voltage and frequency scaling (DVFS) can further help conserve energy consumption (Li et al., 2016) and for some models even reduce time to reach convergence (Tang et al., 2019).”
From Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models (2020) · cites this paper as Tang et al., 2019
No citation lists for this paper
No citation lists for this paper yet — checked again daily.
References 19 · Cited by 88
- Authors
- Tang, Z., Wang, Y., Wang, Q., & Chu, X.
- Affiliations
- Hong Kong Baptist University (HK)
- Venue
- Proceedings of the Tenth ACM International Conference on Future Energy Systems
- Publisher
- ACM
- Published
- 2019-06-15
- Pages
- 315-325
- OpenAlex
- W2945148926
- Cited by
- 88
- References
- 19
- Citations by year
- 2026: 17 · 2025: 14 · 2024: 11 · 2023: 14 · 2022: 7 · 2021: 10 · 2020: 11 · 2019: 4
- FWCI
- 6.30
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM International Conference on Future Energy Systems (e-Energy) (publishedVersion)
- Rare & Special e-Zone (The Hong Kong University of Science and Technology) (submittedVersion)
- Funders
- Hong Kong RGC (HKBU 12200418)
- Topics
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Physical Sciences › Computer Science › Hardware and Architecture › Energy Efficiency in Computing
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- energy efficiency, deep neural network training, deep neural network inference, energy consumption, energy-performance trade-off, GPGPU
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
