clm_f65a62dcfdd74fd9ac5e75616a02adf4_r1

VR2-0776

① SA Source

• Source: 開啟完整 SA 文章
• Section: VR NVL72 TCO: BoM and Power Budget Analysis
• Line hint: 981

Context Before

Source: SemiAnalysis AI TCO Model ↗

Evidence

B300 delivering roughly 6.3x the performance at 100 interactivity for Deepseek R1, using 8k input tokens and 1k output tokens

Context After

Given that total cost of ownership is only 1.75x higher, this results in a superior performance-per-TCO profile for B300 despite the more modest marketed figures. Such a wildly different result despite very similar specs underscores that real world performance is not dictated by peak FLOPS or memory bandwidth alone. Software and network capabilities are also major factors that contribute to training and token throughput in real workloads.

Rubin and MI4XX will ship with new microarchitectures, real world performance is especially difficult to predict without and benchmarking like we do with InferenceX.

② Atomic Claim

InferenceX 在 8k input、1k output tokens、100 interactivity 的 Deepseek R1 測試中，測得 B300 performance 約為對手的 6.3 倍。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "comparison_expression": "InferenceX 在 8k input、1k output tokens、100 interactivity 的 Deepseek R1 測試中，測得 B300 performance 約為對手的 6.3 倍。",
  "entities": [
    {
      "id": "04_knowledge_base/InferenceX",
      "label": "InferenceX"
    },
    {
      "id": "02_companies/DeepSeek",
      "label": "Deepseek"
    },
    {
      "id": "04_knowledge_base/NVIDIA B300",
      "label": "B300"
    }
  ],
  "frame_type": "COMPARISON",
  "metric": "INTERACTIVITY",
  "operator": "MULTIPLE_OF",
  "qualifiers": {
    "condition_text": null,
    "numeric_mentions": [
      "8k",
      "1k",
      "100",
      "6.3"
    ],
    "temporal_mentions": []
  }
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
comparison_entity_0	InferenceX	InferenceX
comparison_entity_1	Deepseek	DeepSeek
comparison_entity_2	B300	04_knowledge_base/NVIDIA B300

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。