VR2R-0544_r1

VR2R-0544

① SA Source

• Source: 開啟完整 SA 文章
• Section: Scale-Up Network - NVLink 6
• Line hint: 458

Context Before

NVIDIA’s GB200 scale-up architecture consists of 18 NVLink 5 Switch chips in a single rack for the NVL72 system.

Evidence

Although the NVLink 6 Switch used in the VR NVL72 system delivers the same per switch 28.8T aggregate bandwidth as NVLink 5 Switch, the SerDes speed on the NVLink 6 Switch is double the SerDes speed of NVLink 5 Switch but with the same number of DPs.

Context After

As such, in order to deliver double the aggregate scale-up bandwidth required for NVLink 6, Vera Rubin NVL72 racks will contain double the number of NVLink Switches as compared to GB200 racks.

② Atomic Claim

VR NVL72 使用的 NVLink 6 Switch，每顆 switch 的 aggregate bandwidth 與 NVLink 5 Switch 相同，都是 28.8T；但 NVLink 6 Switch 的 SerDes speed 是 NVLink 5 的 2 倍，同時 DP 數量相同。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "comparison_expression": "NVLink 6 and NVLink 5 switches each provide 28.8T aggregate bandwidth, while NVLink 6 SerDes speed is 2x with the same number of DPs.",
  "entities": [
    {
      "id": "04_knowledge_base/NVLink",
      "label": "NVLink 6"
    },
    {
      "id": "04_knowledge_base/NVLink5",
      "label": "NVLink 5"
    },
    {
      "id": "04_knowledge_base/SerDes",
      "label": "SerDes"
    },
    {
      "id": "04_knowledge_base/VR NVL72",
      "label": "VR NVL72"
    }
  ],
  "frame_type": "COMPARISON",
  "metric": "SWITCH_BANDWIDTH_AND_SERDES_SPEED",
  "operator": "MIXED_EQUAL_AND_MULTIPLE",
  "qualifiers": {
    "condition_text": "per-switch comparison",
    "numeric_mentions": [
      "28.8T",
      "2x"
    ],
    "temporal_mentions": []
  }
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
entity	NVLink 6	NVLink
entity	NVLink 5	NVLink5
entity	SerDes	SerDes
entity	VR NVL72	04_knowledge_base/VR NVL72

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。