Co-Fabric：跨主机域统一 xPU 互连架构，加速 DeepSeek R1 推理 30%-80%
Co-Fabric: Breaking Host-Domain Boundaries for Unified xPU Interconnection
一篇体系结构论文，做了个跨主机的总线互连，64 卡系统上把 RoCE 的延迟砍一半，跑 DeepSeek R1 快 30%-80%，做 AI 集群的你看看。
arXiv 论文提出 Co-Fabric，一种打破主机域边界的总线型互连架构，面向 scale-up 超级节点。相比基于 Ethernet 的 RoCE，Co-Fabric 采用精简四层协议栈、基于包头端口标识符的跨域 P2P 路由和 shadow-device 自动枚举的统一地址空间。在 64-xPU 3D-Mesh 系统上，节点间通信延迟降低超过 50%，带宽提升 2-5 倍，DeepSeek R1 推理加速 30%-80%。互连本身的成本和功耗最高分别降低 80% 和 5%。
Co-Fabric: Breaking Host-Domain Boundaries for Unified xPU Interconnection
Large-model parameters have grown beyond the capacity of a single xPU, dispersing across multiple xPUs spanning distinct host domains, where xPU-to-xPU communication dominates overall system efficiency. Existing scale-up interconnect remains inadequate: network-based solutions built on Ethernet--such as RoCE (RDMA over Converged Ethernet)--introduce specific message-semantics and protocol-stack characteristics, and rely on fragmented per-host addressing, while conventional host-based fabrics are confined to a single host domain and lack cross-host unified addressing. This paper presents Co-Fabric, a bus-based interconnect that, unlike conventional bus designs, breaks host-domain boundaries to deliver unified xPU interconnection for scale-up superpods. Co-Fabric makes three contributions: a streamlined four-layer protocol stack achieving nanosecond-scale processing latency with native reliability; a cross-domain scaling and P2P mechanism that routes using port identifiers embedded in the packet header; and a unified address space built on shadow-device auto-enumeration. On a 64-xPU 3D-Mesh system, Co-Fabric cuts inter-node communication latency by over 50% and improves bandwidth by 2-5x over RoCE, accelerating DeepSeek R1 inference by 30%-80%. Moreover, since its streamlined four-layer protocol stack and higher data-communication efficiency reduce protocol and processing overhead relative to the Ethernet-based RoCE stack, Co-Fabric cuts the cost and power of the interconnect itself by up to 80% and 5%, respectively. These results demonstrate Co-Fabric's advantage for AI computing centers.