IX2-0246
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Jensen Under Promising and Overdelivering - Hopper vs Blackwell vs Rack Scale NVL72 - Line hint:
400
Context Before
Source: SemiAnalysis InferenceX ↗
Jensen Under Promising and Overdelivering - Hopper vs Blackwell vs Rack Scale NVL72
Evidence
At GTC 2024, Jensen was on stage promising up to 30x performance gains from H100 to GB200 NVL72, everyone thought it was classic marketing lookmaxxing and would not be achievable in real world. ↗ Many looked to come up with labels for this perceived use of a reality distortion field so they could crack more Jensen Math jokes. Indeed – we did point to the comparison of 30x performance difference between the worst case ↗ for H200 on FP8 to a reasonable case of the GB200 on FP4.
Context After
Nvidia Blackwell Perf TCO Analysis - B100 vs B200 vs GB200NVL72
② Atomic Claim
SemiAnalysis 當時也曾指出，30 倍比較是拿 H200 FP8 的較差情境，對比 GB200 FP4 的合理情境。
- Epistemic Mode:
ASSERTED - Mapping Status:
PARTIAL
③ Semantic Frame
{
"comparison_expression": "SemiAnalysis 當時也曾指出，30 倍比較是拿 H200 FP8 的較差情境，對比 GB200 FP4 的合理情境。",
"entities": [
{
"id": "04_knowledge_base/H200",
"label": "H200"
},
{
"id": "04_knowledge_base/FP8",
"label": "FP8"
},
{
"id": "04_knowledge_base/GB200",
"label": "GB200"
},
{
"id": "04_knowledge_base/FP4",
"label": "FP4"
}
],
"frame_type": "COMPARISON",
"metric": "UNSPECIFIED_METRIC",
"operator": "MULTIPLE_OF",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [
"30"
],
"temporal_mentions": []
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| comparison_entity_0 | H200 | H200 |
| comparison_entity_1 | FP8 | FP8 |
| comparison_entity_2 | GB200 | GB200 |
| comparison_entity_3 | FP4 | FP4 |
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。