IX2-0215

① SA Source

Context Before

InfiniBand/RoCEv2 Ethernet (outside of the NVL72 rack): Typically 400-800 Gbit/s per GPU uni-directional (50-100 GB/s). Note that all our testing for Nvidia was conducted on InfiniBand based clusters.

TP shards every layer’s weight matrices across GPUs. This means that every single token at every single layer requires up to two all-reduce communications (one after the column-parallel GEMM, one after the row-parallel GEMM). For EP, all-to-all is done only at MoE layers. Each GPU sends only the tokens routed to each expert. This means cheaper comms across all layers for EP vs TP.

Evidence

Because EP’s all-to-all communication bandwidth requirements scale with the number of participants, staying within the high-bandwidth NVLink domain before having to cross the slower IB/Eth fabric is better

Context After

SemiAnalysis InferenceX is free open source software and reader-supported. To receive new posts and support our work, consider becoming a free or paid subscriber.

Subscribed

② Atomic Claim

由於 EP 的 all-to-all communication 頻寬需求會隨參與者數量增加,因此最好盡可能留在高頻寬 NVLink domain 內,避免跨到較慢的 IB/Eth fabric。

  • Epistemic Mode: ASSERTED
  • Mapping Status: PARTIAL

③ Semantic Frame

{
  "comparison_expression": "由於 EP 的 all-to-all communication 頻寬需求會隨參與者數量增加,因此最好盡可能留在高頻寬 NVLink domain 內,避免跨到較慢的 IB/Eth fabric。",
  "entities": [
    {
      "id": "04_knowledge_base/NVLink",
      "label": "NVLink"
    }
  ],
  "frame_type": "COMPARISON",
  "metric": "BANDWIDTH",
  "operator": "UNSPECIFIED_COMPARISON",
  "qualifiers": {
    "condition_text": null,
    "numeric_mentions": [],
    "temporal_mentions": []
  }
}

④ Canonical Entity Mapping

RoleSurface LabelCanonical Target
comparison_entity_0NVLinkNVLink

⑤ Human Review

請在 Properties 逐項確認:

  • 原文 → Atomic Claim 是否忠實
  • Atomic Claim → Semantic Frame 是否忠實
  • Canonical Entity mapping 是否正確
  • Epistemic mode 是否保留原文語氣
  • 最後選擇 review_action

Review state

Markdown 內文不是正式 approval。只有 Apply bridge 寫入的 Decision Ledger event 才是正式決策。