VR2-0628
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Connecting GPUs in the Scale-Out InfiniBand Network - Line hint:
753
Context Before
Source: SemiAnalysis AI Networking Model ↗
Evidence
Context After
Source: SemiAnalysis AI Networking Model ↗
② Atomic Claim
因此，HGX Rubin NVL8 server 的 InfiniBand architecture 實質上是 single-plane、8-rail network，每顆 HGX Rubin NVL8 GPU 都以 1 條 800G uplink 連至 leaf switch。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"attribute": "COUNT",
"context_nodes": [
{
"id": "04_knowledge_base/Vera Rubin",
"label": "Rubin"
},
{
"id": "04_knowledge_base/NVL8",
"label": "NVL8"
},
{
"id": "04_knowledge_base/InfiniBand",
"label": "InfiniBand"
},
{
"id": "04_knowledge_base/GPU",
"label": "GPU"
},
{
"id": "04_knowledge_base/800G",
"label": "800G"
}
],
"entity": {
"id": "04_knowledge_base/HGX",
"label": "HGX"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [
"8",
"1",
"800"
],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [
"8",
"1",
"800"
],
"value_text": "因此，HGX Rubin NVL8 server 的 InfiniBand architecture 實質上是 single-plane、8-rail network，每顆 HGX Rubin NVL8 GPU 都以 1 條 800G uplink 連至 leaf switch。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | HGX | HGX |
| context_0 | Rubin | 04_knowledge_base/Vera Rubin |
| context_1 | NVL8 | NVL8 |
| context_2 | InfiniBand | InfiniBand |
| context_3 | GPU | GPU |
| context_4 | 800G | 800G |
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。