IX2-0212
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Nvidia Disagg Prefill and WideEP - Line hint:
370
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.
Evidence
For EP, all-to-all is done only at MoE layers
Context After
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. With NVL72, EP across 72 GPUs is possible without ever leaving NVLink, whereas 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.
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.
② Atomic Claim
EP 則只在 MoE layers 執行 all-to-all。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"attribute": "LAYER_COUNT",
"context_nodes": [],
"entity": {
"id": "04_knowledge_base/Mixture of Experts",
"label": "MoE"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [],
"value_text": "EP 則只在 MoE layers 執行 all-to-all。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | MoE | 04_knowledge_base/Mixture of Experts |
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。