EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

BridgedRing: A Cost-Effective Hardware-Software Co-Design to Overcome the UPI Bottleneck in GPU Servers

Wanfeng Huang, Yihao Zhao, Yakun Zhang, Jianghong Ma, Yunming Ye

ACM Transactions on Architecture and Code Optimization · 30.09.2026 · 23 (3) · 1-25

DOI: 10.1145/3822179 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ACM Transactions on Architecture and Code Optimization, online erschienen am 24.07.2026. Verfasst von Wanfeng Huang und 4 weiteren Autor:innen. Die Arbeit nennt 35 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Shenzhen Science and Technology Program, National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Distributed training on standard dual-socket GPU servers suffers from the “UPI Wall” — a bottleneck stemming from the interplay between the low physical bandwidth of the Ultra Path Interconnect (UPI) and the severe bidirectional contention induced by NCCL’s ring-based collectives. Software-only optimizations cannot overcome this hardware constraint. We present BridgedRing , a hardware-software co-design that bypasses the UPI Wall using an off-the-shelf NVLink Bridge costing approximately $200 — less than 0.5% of the total cost of a typical 8-GPU server. Our topology-aware algorithm redirects all cross-NUMA traffic through this high-bandwidth bypass while preserving intra-NUMA communication over PCIe. Evaluated on 8×A6000 servers against NCCL and hierarchical algorithms: BridgedRing achieves up to 1.97× higher collective bandwidth and delivers 1.31× end-to-end throughput for GPT-13B training in Megatron-LM (TP=8, SP=8), 1.36× for GPT-6.7B, and 1.40× for Qwen3-1.7B in PyTorch DDP . Crucially, these gains require no framework modifications , offering immediate, cost-effective acceleration for the vast installed base of standard GPU servers.

Förderung

  • Shenzhen Science and Technology Program
  • National Natural Science Foundation of China

Zitieren

APA

Huang, W., Zhao, Y., Zhang, Y., Ma, J., & Ye, Y. (2026). BridgedRing: A Cost-Effective Hardware-Software Co-Design to Overcome the UPI Bottleneck in GPU Servers. ACM Transactions on Architecture and Code Optimization, 23(3), 1-25. https://doi.org/10.1145/3822179

Vancouver

Huang W, Zhao Y, Zhang Y, Ma J, Ye Y. BridgedRing: A Cost-Effective Hardware-Software Co-Design to Overcome the UPI Bottleneck in GPU Servers. ACM Transactions on Architecture and Code Optimization. 2026;23(3):1-25. doi:10.1145/3822179

BibTeX
@article{Huang2026,
  author = {Huang, Wanfeng and Zhao, Yihao and Zhang, Yakun and Ma, Jianghong and Ye, Yunming},
  title = {BridgedRing: A Cost-Effective Hardware-Software Co-Design to Overcome the UPI Bottleneck in GPU Servers},
  journal = {ACM Transactions on Architecture and Code Optimization},
  year = {2026},
  volume = {23},
  number = {3},
  pages = {1--25},
  doi = {10.1145/3822179}
}
Kostenlos starten →