Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2512.01644 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915647050481664 |
|---|---|
| author | Wang, Haonan Xiao, Xuxin Yan, Mingyu Zhu, Zhuoyuan Han, Dengke Wang, Duo Li, Wenming Ye, Xiaochun Hu, Cunchen Chen, Hongyang Sun, Guangyu |
| author_facet | Wang, Haonan Xiao, Xuxin Yan, Mingyu Zhu, Zhuoyuan Han, Dengke Wang, Duo Li, Wenming Ye, Xiaochun Hu, Cunchen Chen, Hongyang Sun, Guangyu |
| contents | This work presents a systematic characterization of Large Language Model (LLM) inference to address fragmented understanding. Through comprehensive experiments, we establish a four-dimensional analytical framework: (1) Two-Phase Heterogeneity Observation; (2) Microarchitectural Root Cause Analysis; (3) System Scaling Principles; and (4) Emerging Paradigm Boundaries. Our investigation progresses systematically from observation to foresight: identifying performance phenomena, revealing hardware causes, validating system behavior, and exploring new paradigms. This study not only consolidates a reliable empirical foundation for existing research but also provides new discoveries and practical optimization guidance for LLM inference. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_01644 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A Systematic Characterization of LLM Inference on GPUs Wang, Haonan Xiao, Xuxin Yan, Mingyu Zhu, Zhuoyuan Han, Dengke Wang, Duo Li, Wenming Ye, Xiaochun Hu, Cunchen Chen, Hongyang Sun, Guangyu Hardware Architecture This work presents a systematic characterization of Large Language Model (LLM) inference to address fragmented understanding. Through comprehensive experiments, we establish a four-dimensional analytical framework: (1) Two-Phase Heterogeneity Observation; (2) Microarchitectural Root Cause Analysis; (3) System Scaling Principles; and (4) Emerging Paradigm Boundaries. Our investigation progresses systematically from observation to foresight: identifying performance phenomena, revealing hardware causes, validating system behavior, and exploring new paradigms. This study not only consolidates a reliable empirical foundation for existing research but also provides new discoveries and practical optimization guidance for LLM inference. |
| title | A Systematic Characterization of LLM Inference on GPUs |
| topic | Hardware Architecture |
| url | https://arxiv.org/abs/2512.01644 |