SA260324R-0160
① SA Source
- Source: 開啟完整 SA 文章
- Section:
LPU Network - Line hint:
195
Context Before
The LPU network can be divided into the scale-up ‘C2C’ network and
Evidence
scale-out network which interacts with the Nvidia GPUs through Spectrum-X.
Context After
First let’s discuss the scale-up network which can be divided into 3 portions: intra-node, inter-node/intra-rack, inter-rack. For C2C within the rack Nvidia announced a total of 640TB/s of scale up bandwidth per rack which comes from 256 LPUs x 90 lanes x 112Gbps/8 x 2 directions = 645TB/s.
② Atomic Claim
scale-out network 會透過 Spectrum-X 與 Nvidia GPU 互動。
- Epistemic Mode:
ASSERTED - Mapping Status:
PARTIAL
③ Semantic Frame
{
"frame_type": "NARY_RELATION",
"participants": [
{
"node": {
"id": "04_knowledge_base/Scale-out networking",
"label": "scale-out network"
},
"role": "network"
},
{
"node": {
"id": "02_companies/NVDA",
"label": "Nvidia"
},
"role": "gpu_vendor"
},
{
"node": {
"id": "04_knowledge_base/GPU",
"label": "GPU"
},
"role": "compute_device"
}
],
"qualifiers": {
"condition_text": "Spectrum-X 在此 claim 沒有 canonical mapping",
"numeric_mentions": [],
"temporal_mentions": []
},
"relation_type": "INTERACTS_WITH_GPU_THROUGH_SCALEOUT_NETWORK"
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| network | scale-out network | 04_knowledge_base/Scale-out networking |
| gpu_vendor | Nvidia | NVDA |
| compute_device | GPU | GPU |
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。