clm_6a021a59fa71485ca74ea24f1614b10f_r1

IX2-0026

① SA Source

• Source: 開啟完整 SA 文章
• Section: Key Observations and Results to Highlight
• Line hint: 45

Context Before

SemiAnalysis InferenceX is free open source software and reader-supported. To receive new posts and support our work consider becoming a free or paid subscriber.

Subscribed

Evidence

When it comes to the latest inference techniques that are used by the most prominent frontier large-scale inference services (such as disagg prefill+wideEP+FP4), Nvidia absolutely frame mogs with the B200, B300 and ASU frat leader, rack scale GB200/GB300 NVL72 across both SGLang and TRTLLM

Context After

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

② Atomic Claim

在 frontier 大規模推論服務採用的最新技術（例如 disagg prefill+wideEP+FP4）上，Nvidia 的 B200、B300、rack-scale GB200／GB300 NVL72，無論搭配 SGLang 或 TRTLLM 都明顯領先。

• Epistemic Mode: ASSERTED
• Mapping Status: PARTIAL

③ Semantic Frame

{
  "frame_type": "NARY_RELATION",
  "participants": [
    {
      "node": {
        "id": "04_knowledge_base/Prefill",
        "label": "prefill"
      },
      "role": "user_or_subject"
    },
    {
      "node": {
        "id": "04_knowledge_base/FP4",
        "label": "FP4"
      },
      "role": "used_entity"
    },
    {
      "node": {
        "id": "02_companies/NVDA",
        "label": "Nvidia"
      },
      "role": "used_entity"
    },
    {
      "node": {
        "id": "04_knowledge_base/NVIDIA B200",
        "label": "B200"
      },
      "role": "used_entity"
    },
    {
      "node": {
        "id": "04_knowledge_base/NVIDIA B300",
        "label": "B300"
      },
      "role": "used_entity"
    },
    {
      "node": {
        "id": "04_knowledge_base/GB200",
        "label": "GB200"
      },
      "role": "used_entity"
    },
    {
      "node": {
        "id": "04_knowledge_base/GB300 NVL72",
        "label": "GB300 NVL72"
      },
      "role": "used_entity"
    },
    {
      "node": {
        "id": "04_knowledge_base/SGLang",
        "label": "SGLang"
      },
      "role": "used_entity"
    },
    {
      "node": {
        "id": "04_knowledge_base/TensorRT-LLM",
        "label": "TRTLLM"
      },
      "role": "used_entity"
    }
  ],
  "qualifiers": {
    "condition_text": null,
    "numeric_mentions": [],
    "temporal_mentions": []
  },
  "relation_type": "USES"
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
user_or_subject	prefill	Prefill
used_entity	FP4	FP4
used_entity	Nvidia	NVDA
used_entity	B200	04_knowledge_base/NVIDIA B200
used_entity	B300	04_knowledge_base/NVIDIA B300
used_entity	GB200	GB200
used_entity	GB300 NVL72	04_knowledge_base/GB300 NVL72
used_entity	SGLang	SGLang
used_entity	TRTLLM	TensorRT-LLM

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。