谷歌在Google Cloud Next大会上发布新型推理专用TPU，采用名为Broadfly的新型网络拓扑。利用高基数设计，单pod最多可扩展到1152个TPU，相比Ironwood，pod大小提升4.5倍，网络直径减小，任意两个芯片间最多7跳。
During their last Google Cloud Next conference in Las Vegas, Google unveiled their new inference-focused TPU, featuring a novel network topology called "Broadfly".
By leveraging a high-radix design, Google can scale up to 1,152 TPUs in a single pod.
Compared to Ironwood, this enables a 4.5x larger pod size while reducing network diameter and with a maximum of just 7 hops between any two chips. (1/3) 🧵
likes: 147 | retweets: 13 | replies: 5 | views: 28065