GB2-0007
① SA Source
- Source: 開啟完整 SA 文章
- Section:
The 4 Rack Scale Form Factors of Blackwell - Line hint:
157
Context Before
The GB200 NVL72 racks consists of 18 1U compute trays and 9 NVSwitch trays. Each compute tray is 1U in height and contains 2 Bianca boards. Each Bianca board is 1 Grace CPU and 2 Blackwell GPUs. The NVSwitch trays have two 28.8Tb/s NVSwitch5 ASICs.
With the exception of one hyperscaler who plans to deploy this as the primary variant, we believe that this version will be rarely deployed until Blackwell Ultra as most datacenter infrastructure cannot support this high of a rack density even with direct-to-chip liquid cooling (DLC).
Evidence
Between the 2 racks, it still maintains being non-blocking all-to-all between all the 72 GPUs found in NVL72.
Context After
Source: SemiAnalysis
② Atomic Claim
GB200 NVL36x2 的兩個 rack 之間維持 72 顆 GPU 的 non-blocking all-to-all 連線。
- Epistemic Mode:
ASSERTED - Mapping Status:
PARTIAL
③ Semantic Frame
{
"attribute": "scale_up_connectivity",
"context_nodes": [
{
"id": "04_knowledge_base/GPU",
"label": "GPU"
}
],
"entity": {
"id": "04_knowledge_base/NVL36",
"label": "NVL36"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": "NVL36x2 pair",
"numeric_mentions": [
"2",
"72"
],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [
"2",
"72"
],
"value_text": "兩個 rack 間對 72 顆 GPU 維持 non-blocking all-to-all"
}
}④ Canonical Entity Mapping
⑤ V2 Draft（需要修正時直接編輯）
Inline repair
只編輯以下三個 Draft 區塊。完成後將
review_action設為REQUEST_REPAIR，再手動執行 Repair Preview。
執行腳本前，Draft 不會寫入 Claim Store 或 Graph。
V2 Atomic Claim Draft
GB200 NVL36x2 的兩個 rack 之間維持 72 顆 GPU 的 non-blocking all-to-all 連線。
V2 Semantic Frame Draft
{
"attribute": "scale_up_connectivity",
"context_nodes": [
{
"id": "04_knowledge_base/GPU",
"label": "GPU"
}
],
"entity": {
"id": "04_knowledge_base/NVL36",
"label": "NVL36"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": "NVL36x2 pair",
"numeric_mentions": [
"2",
"72"
],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [
"2",
"72"
],
"value_text": "兩個 rack 間對 72 顆 GPU 維持 non-blocking all-to-all"
}
}V2 Revision Metadata Draft
{
"epistemic_mode": "ASSERTED",
"mapping_status": "PARTIAL"
}⑥ 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。