2025-12-04_aws-trainium3-deep-dive-a-potential::TRN3-0012
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Amazon’s Software North Star - Line hint:
81
Context Before
Air Cooled Trainium3 NL32x2 Switched (Codename “Teton3 PDS”)
Evidence
Liquid Cooled Trainium3 NL72x2 Switched (Codename “Teton3 MAX”)
Context After
We will start by briefly reviewing the Trn2 architecture and explaining the changes introduced with Trainium3. The first half of the article will focus on the various Trainium3 rack SKUs’ specifications, silicon design, rack architecture, bill of materials (BoM) and power budget before we turn to the scale-up and scale-out network architecture. In the second half of this article, we will focus on discussing the Trainium3 Microarchitecture and expand further on Amazon’s software strategy. We will conclude with a discussion on Amazon and Anthropic’s AI Datacenters before tying everything together with a Total Cost of Ownership (TCO) and Perf per TCO analysis.
Trainium3 Server Types and Specifications Overview
② Atomic Claim
Trainium3 NL72x2 Switched（Teton3 MAX）為 liquid-cooled rack SKU。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"frame_type": "NARY_RELATION",
"participants": [
{
"node": {
"id": "04_knowledge_base/Trainium",
"label": "Trainium3"
},
"role": "accelerator"
},
{
"node": {
"id": "04_knowledge_base/Teton3 Max",
"label": "Teton3 MAX"
},
"role": "codename"
},
{
"node": {
"id": "04_knowledge_base/液冷散熱",
"label": "Liquid cooling"
},
"role": "cooling"
}
],
"qualifiers": {
"condition_text": "NL72x2 Switched",
"numeric_mentions": [],
"temporal_mentions": []
},
"relation_type": "RACK_SKU"
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| accelerator | Trainium3 | Trainium |
| codename | Teton3 MAX | 04_knowledge_base/Teton3 Max |
| cooling | Liquid cooling | 04_knowledge_base/液冷散熱 |
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。