VR2R-0736_r1

VR2R-0736

① SA Source

• Source: 開啟完整 SA 文章
• Section: VR NVL72 TCO: BoM and Power Budget Analysis
• Line hint: 651

Context Before

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.

Evidence

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

Context After

Notably, both operating modes share identical memory bandwidth specifications at 8TB/s. Yet, despite the parity in memory bandwidth inference performance still diverges materially.

② Atomic Claim

Rubin 與 MI4XX 都會採用新的 microarchitectures，因此若沒有像 InferenceX 這樣的 benchmark，真實世界 performance 特別難以預測。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "frame_type": "NARY_RELATION",
  "participants": [
    {
      "node": {
        "id": "04_knowledge_base/Vera Rubin",
        "label": "Rubin"
      },
      "role": "platform"
    },
    {
      "node": {
        "id": "04_knowledge_base/MI400",
        "label": "MI4XX"
      },
      "role": "comparison_platform"
    },
    {
      "node": {
        "id": "04_knowledge_base/InferenceX",
        "label": "InferenceX"
      },
      "role": "benchmark"
    }
  ],
  "qualifiers": {
    "condition_text": "new microarchitectures make unbenchmarked real-world performance difficult to predict",
    "numeric_mentions": [],
    "temporal_mentions": [
      "will ship"
    ]
  },
  "relation_type": "NEW_MICROARCHITECTURES_REQUIRE_REAL_WORLD_BENCHMARKING"
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
platform	Rubin	04_knowledge_base/Vera Rubin
comparison_platform	MI4XX	MI400
benchmark	InferenceX	InferenceX

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。