2025-06-13_amd-advancing-ai-mi350x-and-mi400-ualoe72-mi500-ual256::AMD25-0024
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Introducing the MI400 Helios Rack - Line hint:
359
Context Before
Source: AMD, SemiAnalysis Estimates
Evidence
the scale up bandwidth of MI400s of 1.8TB/s (uni-directional) per GPU achieved using 72 lanes of 200Gbit/s each is the same as we estimated previously.
Context After
As a result of this, at least at the marketing level, most of the flexible I/O lanes coming out of the GPU will be on the UALink protocol: UALink to the Vulcano NICs, xGMI4 (a subset of UALink) to the CPU, and the aforementioned UALoE for the scale up network for communicating with GPUs in different compute trays. In the chart we colored all the “UALink” protocol lanes as purple.
The MI450X GPU will have direct access to three different levels of memory, which indicates that the rack solution will be optimized for inference workloads as the availability of multiple memory tiers enables more efficient KVCaching. It is also interesting to see a direct PCIe channel between GPU and SSD, whereas for the GB200, the GPU still accesses NVMe storage via the Grace CPU. The four tiers of memory are:
② Atomic Claim
MI400 scale-up bandwidth 為每 GPU 1.8TB/s 單向，使用 72 lanes × 200Gbit/s。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"attribute": "scale_up_link_bandwidth",
"context_nodes": [
{
"id": "04_knowledge_base/GPU",
"label": "GPU"
}
],
"entity": {
"id": "04_knowledge_base/MI400",
"label": "MI400"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [
"1.8TB/s",
"72",
"200Gbit/s"
],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [
"1.8TB/s",
"72",
"200Gbit/s"
],
"value_text": "1.8TB/s unidirectional per GPU via 72 × 200Gbit/s lanes"
}
}④ Canonical Entity Mapping
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。