ChipLight: Cross-Layer Optimization of Chiplet Design with Optical Interconnects for LLM Training
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914494238687232 |
|---|---|
| author | Bai, Kangbo Zhu, Zhantong Ding, Yifan Jia, Tianyu |
| author_facet | Bai, Kangbo Zhu, Zhantong Ding, Yifan Jia, Tianyu |
| contents | In large-scale distributed LLM training, communication between devices becomes the key performance bottleneck. Chiplet technology can integrate multiple dies into a package to scale-up node performance with higher bandwidth. Meanwhile, optical interconnect (OI) technology offers long-reach, high-bandwidth links, making it well suited for scale-out networks. The combination of these two technologies has the potential to overcome communication bottlenecks within and across packages. In this work, we present ChipLight, a cross-layer multi-objective design and optimization method for training clusters leveraging chiplet and OI. We first abstract an architecture model for such complex clusters, co-optimizing chiplet architecture, training parallel strategy, and OI network topology. Based on such models, we tailor the design space exploration flow by combining both black-box and white-box methodologies. Evaluated by our experimental results, ChipLight achieves significantly improved training efficiency and provides valuable design insights for the development of future training clusters. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_18909 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ChipLight: Cross-Layer Optimization of Chiplet Design with Optical Interconnects for LLM Training Bai, Kangbo Zhu, Zhantong Ding, Yifan Jia, Tianyu Hardware Architecture In large-scale distributed LLM training, communication between devices becomes the key performance bottleneck. Chiplet technology can integrate multiple dies into a package to scale-up node performance with higher bandwidth. Meanwhile, optical interconnect (OI) technology offers long-reach, high-bandwidth links, making it well suited for scale-out networks. The combination of these two technologies has the potential to overcome communication bottlenecks within and across packages. In this work, we present ChipLight, a cross-layer multi-objective design and optimization method for training clusters leveraging chiplet and OI. We first abstract an architecture model for such complex clusters, co-optimizing chiplet architecture, training parallel strategy, and OI network topology. Based on such models, we tailor the design space exploration flow by combining both black-box and white-box methodologies. Evaluated by our experimental results, ChipLight achieves significantly improved training efficiency and provides valuable design insights for the development of future training clusters. |
| title | ChipLight: Cross-Layer Optimization of Chiplet Design with Optical Interconnects for LLM Training |
| topic | Hardware Architecture |
| url | https://arxiv.org/abs/2604.18909 |