VR2R-0232

① SA Source

Context Before

Nvidia’s ICMS inserts a new G3.5 tier between local SSD (G3) and shared storage (G4), optimized specifically for ephemeral, recomputable KV cache. The ICMS requires a dedicated networking layer designed exclusively for KV traffic. Wherever networking is used in this architecture, it is provisioned as a context memory network—isolated from general data movement and optimized for predictable decode latency.

Evidence

The challenge with this is that the volumes of SSDs going to ICMS / CMX are quite overblown by the industry.

Context After

BlueField-4 will be the silicon anchor of this third network. Positioned on the storage array, it terminates NVMe-oF and RDMA traffic at line rate and manages KV movement independently of host CPUs and GPUs. With 2×400G SerDes links providing 800Gb/s of bandwidth, integrated Grace CPU, and LPDDR, BlueField-4 would act as the controller for a distributed context memory fabric.

② Atomic Claim

SemiAnalysis 認為,業界對流向 ICMS/CMX 的 SSD volumes 有相當程度的高估。

  • Epistemic Mode: ASSERTED
  • Mapping Status: COMPLETE

③ Semantic Frame

{
  "attribute": "ANALYST_VIEW_OF_SSD_VOLUME_EXPECTATIONS",
  "context_nodes": [],
  "entity": {
    "id": "04_knowledge_base/Inference Context Memory Storage",
    "label": "ICMS / CMX"
  },
  "frame_type": "ATTRIBUTE",
  "qualifiers": {
    "condition_text": "analyst judgment",
    "numeric_mentions": [],
    "temporal_mentions": []
  },
  "value": {
    "numeric_mentions": [],
    "value_text": "industry SSD volume expectations for ICMS/CMX are quite overblown."
  }
}

④ Canonical Entity Mapping

RoleSurface LabelCanonical Target
entityICMS / CMX04_knowledge_base/Inference Context Memory Storage

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。