clm_6943d7f0e27d4157a2c8413995c96ef7_r1

IX2-0036

① SA Source

• Source: 開啟完整 SA 文章
• Section: Key Observations and Results to Highlight
• Line hint: 51

Context Before

Turning to AMD, we find that the biggest issue with inference on their systems and using their software is composability ↗. That is, many of AMDs inference optimization implementations work well in isolation, but when combined with other optimizations, the result is not as competitive as one would expect. Specifically, the composability of disagg prefill, wideEP and FP4 inference optimizations needs significant improvement.

While performance is competitive on AMD when enabling just a subset of the SOTA inference optimizations, enabling all three major optimizations that labs use, AMD’s performance is currently not competitive with Nvidia’s. We strongly recommend to AMD that they focus heavily on composability of different inference optimizations. We have been told that AMD will start focusing on software composability of FP4+distributed inferencing across their whole software stack. This will happen after Chinese New Year as most of their disagg prefill+wideEP 10x inference engineers are based in China

Evidence

It achieves up to 100x on FP8 vs FP4 compared to even a strong H100 disagg+wideEP+MTP baseline and 65x on FP8 vs FP8

Context After

At GTC 2024, Jensen claimed that Blackwell will deliver up to 30x perf on inference compared to H100, Jensen under promised & overdelivered on Blackwell inference performance. This should curtail the instances of analysts cracking “Jensen Math” jokes for some time.

② Atomic Claim

相較強勁的 H100 disagg+wideEP+MTP baseline，在 FP8 對 FP4 的比較中最高可達 100 倍；另一個 FP8 對 FP8 的比較則可達 65 倍。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "comparison_expression": "相較強勁的 H100 disagg+wideEP+MTP baseline，在 FP8 對 FP4 的比較中最高可達 100 倍；另一個 FP8 對 FP8 的比較則可達 65 倍。",
  "entities": [
    {
      "id": "04_knowledge_base/H100",
      "label": "H100"
    },
    {
      "id": "04_knowledge_base/Multi-Token Prediction",
      "label": "MTP"
    },
    {
      "id": "04_knowledge_base/FP8",
      "label": "FP8"
    },
    {
      "id": "04_knowledge_base/FP4",
      "label": "FP4"
    }
  ],
  "frame_type": "COMPARISON",
  "metric": "COUNT",
  "operator": "MULTIPLE_OF",
  "qualifiers": {
    "condition_text": null,
    "numeric_mentions": [
      "100",
      "65"
    ],
    "temporal_mentions": []
  }
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
comparison_entity_0	H100	H100
comparison_entity_1	MTP	04_knowledge_base/Multi-Token Prediction
comparison_entity_2	FP8	FP8
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。