clm_9654cd64695147409ff8e56a79f6eb4e_r1

IX2-0217

① SA Source

• Source: 開啟完整 SA 文章
• Section: Nvidia Disagg Prefill and WideEP
• Line hint: 372

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

previous generations (with only 8-GPU NVLink domains) could only do EP across 8 GPUs at NVLink speed before hitting the slower IB/Eth networks

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

上一代只有 8-GPU NVLink domains，因此 EP 最多只能在 8 顆 GPUs 內以 NVLink 速度運作，超過後就必須進入較慢的 IB/Eth networks。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "comparison_expression": "上一代只有 8-GPU NVLink domains，因此 EP 最多只能在 8 顆 GPUs 內以 NVLink 速度運作，超過後就必須進入較慢的 IB/Eth networks。",
  "entities": [
    {
      "id": "04_knowledge_base/GPU",
      "label": "GPU"
    },
    {
      "id": "04_knowledge_base/NVLink",
      "label": "NVLink"
    }
  ],
  "frame_type": "COMPARISON",
  "metric": "COUNT",
  "operator": "UNSPECIFIED_COMPARISON",
  "qualifiers": {
    "condition_text": null,
    "numeric_mentions": [
      "8"
    ],
    "temporal_mentions": []
  }
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
comparison_entity_0	GPU	GPU
comparison_entity_1	NVLink	NVLink

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。