Characterizing Compute-Communication Overlap in GPU-Accelerated Distributed Deep Learning: Performance and Power Implications
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918082935521280 |
|---|---|
| author | Lee, Seonho Oh, Jihwan Kim, Junkyum Go, Seokjin Park, Jongse Mahajan, Divya |
| author_facet | Lee, Seonho Oh, Jihwan Kim, Junkyum Go, Seokjin Park, Jongse Mahajan, Divya |
| contents | This paper provides an in-depth characterization of GPU-accelerated systems, to understand the interplay between overlapping computation and communication which is commonly employed in distributed training settings. Due to the large size of models, distributing them across multiple devices is required. Overlapping strategies, which enable concurrent computation and communication, are critical for mitigating communication bottlenecks and maximizing GPU utilization. However, the current consensus is that we should always and aggressively overlap compute and communication to mitigate the overhead of distribution. By systematically evaluating state-of-the-art GPUs, this study investigates the impact of hardware features such as numeric precision, specialized cores, and power capping on distributed training workloads. Comprehensive experiments and studies showcase the effects of overlapping strategies on performance and power consumption across varying scenarios. We observe that overlapping computation and communication can result in an average computational slowdown of 18.9%, with a maximum of 40.0% slowdown. This slowdown is in comparison to the scenario when no communication was happening with the compute. We consider this an ideal execution scenario, where the communication in parallel has not impact on the compute time. However, performing computation and communication sequentially is, on average, 10.2% slower than overlapped execution, with a maximum slowdown of 26.6%. We further observe, while specialized datapath and optimized numeric precision mitigate certain slowdowns, overlapping execution can lead to resource contention and also increase power consumption under specific configurations. The analysis also uncovers trade-offs introduced by power and frequency capping, emphasizing the importance of balanced strategies to optimize energy efficiency and training throughput. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_03114 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Characterizing Compute-Communication Overlap in GPU-Accelerated Distributed Deep Learning: Performance and Power Implications Lee, Seonho Oh, Jihwan Kim, Junkyum Go, Seokjin Park, Jongse Mahajan, Divya Distributed, Parallel, and Cluster Computing This paper provides an in-depth characterization of GPU-accelerated systems, to understand the interplay between overlapping computation and communication which is commonly employed in distributed training settings. Due to the large size of models, distributing them across multiple devices is required. Overlapping strategies, which enable concurrent computation and communication, are critical for mitigating communication bottlenecks and maximizing GPU utilization. However, the current consensus is that we should always and aggressively overlap compute and communication to mitigate the overhead of distribution. By systematically evaluating state-of-the-art GPUs, this study investigates the impact of hardware features such as numeric precision, specialized cores, and power capping on distributed training workloads. Comprehensive experiments and studies showcase the effects of overlapping strategies on performance and power consumption across varying scenarios. We observe that overlapping computation and communication can result in an average computational slowdown of 18.9%, with a maximum of 40.0% slowdown. This slowdown is in comparison to the scenario when no communication was happening with the compute. We consider this an ideal execution scenario, where the communication in parallel has not impact on the compute time. However, performing computation and communication sequentially is, on average, 10.2% slower than overlapped execution, with a maximum slowdown of 26.6%. We further observe, while specialized datapath and optimized numeric precision mitigate certain slowdowns, overlapping execution can lead to resource contention and also increase power consumption under specific configurations. The analysis also uncovers trade-offs introduced by power and frequency capping, emphasizing the importance of balanced strategies to optimize energy efficiency and training throughput. |
| title | Characterizing Compute-Communication Overlap in GPU-Accelerated Distributed Deep Learning: Performance and Power Implications |
| topic | Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2507.03114 |