clm_ebf6722d1a834e488374831750278881_r1

2025-12-04_aws-trainium3-deep-dive-a-potential::TRN3-0027

① SA Source

• Source: 開啟完整 SA 文章
• Section: Trainium3 NL72x2 Switched Scale Out Networking
• Line hint: 321

Context Before

Option 2: Four Nitro-V6 (2*200G) 400Gbps Nitro NIC modules per JBOG tray of four Trainium3 chips: 400Gbps of EFA bandwidth per Trainium3.

Evidence

Like the Trainium3 NL32x2 Switched, the vast majority of Trainium3 NL72x2 volume will be Option 1: One Nitro-v6 400G NIC per two Trainium3 chips which equals 200Gbps of scale-out bandwidth per Trainium3 chip.

Context After

The difference though is that for the Trainium3 NL72x2 Switched, the host CPUs are now located on the Compute Tray, and the CPU dedicated Nitro-V6 (2*200) 400Gbps NIC module is also located in the tray. Moreover, with PCIe switches, the CPU can also talk to the outside world using the Trainium3 dedicated NICs.

Trainium3 NL72x2 Switched is AWS’s answer to Nvidia’s Oberon rack architecture. The Trainium3 NL72x2 Switched architecture has much higher power density than its predecessor. Due to the high power density and need for liquid cooling ready datacenters, we expect some volume of Trainium3 to be deployed in the Trainium3 NL72x2 Switched SKU, but majority of the volume will still go toward Trainium3 NL32x2 Switched. After understanding the rack layout and compute tray layout/topology, it is time to dive into the real secret sauce that makes Trainium3 NL32x2 Switched and Trainium3 NL72x2 Switched stand out, the switched scale-up network topology.

② Atomic Claim

SemiAnalysis 預期 NL72x2 多數配置為每兩顆 Trainium3 共用一個 Nitro-v6 400G NIC，即每 chip 200Gbps scale-out。

• Epistemic Mode: EXPECTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "attribute": "dominant_scale_out_configuration",
  "context_nodes": [
    {
      "id": "04_knowledge_base/Nitro",
      "label": "Nitro"
    },
    {
      "id": "04_knowledge_base/NIC",
      "label": "NIC"
    }
  ],
  "entity": {
    "id": "04_knowledge_base/Trainium",
    "label": "Trainium3"
  },
  "frame_type": "ATTRIBUTE",
  "qualifiers": {
    "condition_text": "NL72x2",
    "numeric_mentions": [
      "1",
      "400G",
      "2",
      "200Gbps"
    ],
    "temporal_mentions": []
  },
  "value": {
    "numeric_mentions": [
      "1",
      "400G",
      "2",
      "200Gbps"
    ],
    "value_text": "1 Nitro-v6 400G NIC per 2 chips = 200Gbps/chip"
  }
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
entity	Trainium3	Trainium
context_0	Nitro	Nitro
context_1	NIC	NIC

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。