--- 格式版本: 2 标题: "AMD’s AI Architecture Evolution: From EPYC and Instinct to Helios and Heterogeneous Integration" 原文链接: "https://tspasemiconductor.substack.com/p/amds-ai-architecture-evolution-from" 发布日期: "2026-08-23" 发布时间校准状态: "found" 发布时间需复核: "否" 发布时间来源: "rule:scrape:strict_html_body" 发布时间证据: "div class=pencraft pc-display-flex pc-gap-12 pc-alignItems-center pc-reset byline-wrapper: Aug 23, 2026" 发布时间校准原因: "规则确认唯一严格发布时间，来源 scrape:strict_html_body" 发布时间校准置信度: "high" 发布时间候选数量: 11 发布时间严格候选数量: 2 发布时间原页读取状态: "source template page reused from URL open" 发布时间未找到原因: "" 发布时间校准时间: "2026-08-31T11:37:57+08:00" 发布时间仲裁状态: "skipped" 发布时间仲裁尝试次数: 0 发布时间仲裁耗时毫秒: 0 发现时间: "2026-08-31T11:35:07+08:00" 入库时间: "2026-08-31T03:38:11.575Z" 来源平台: "Substack 数据中心相关博客搜索" 搜索渠道: "source_template" 搜索词: "site:substack.com Rack scale AI" 匹配关键词: - "AI" - "GPU" - "CPO" - "roadmap" - "performance" 相关厂家: - "AMD" 相关专家: [] 内容类型: "网页" 抓取工具: "Free Fetch + Defuddle" 清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取" 原始附件: [] 图片摘要: - "★ ./assets/img-ca5a0f1d.webp | diagram | 展示AMD AI架构从EPYC、Instinct到Helios的演进路线，涵盖芯片、封装与机架级互连技术。" AI优质: "否" AI打分: 25 AI分档: "非优质" AI质检状态: "不通过" AI打分理由: "正文主线是AMD从芯片向异构集成和机架级平台演进的概览，涉及Helios、EPYC、Instinct、先进封装与CPO，但当前页面实际仅有导语、章节目录和图片，未提供规格、机制、客户、量产或部署正文。来源为非官方Substack博客；历史证据已包含Helios正式发布、客户合作及部署信息，本文未给出可核验新增事实。命中正文近似摘要壳/内容不完整及成熟架构综述无新增的否决项，当前页面本身不值得作为独立信息源。" AI质检模型: "gpt-5.6-sol" AI质检时间: "2026-08-31T11:38:20+08:00" AI主题相关性: 12 AI来源权威性: 5 AI新颖性: 2 AI技术细节: 4 AI商业部署信号: 0 AI完整性: 2 AI评分提示词版本: "v17-精简生产版" AI评分提示词SHA256: "48fb9777f386026761b4873eaff30807694fb11e9b352d7c69bf2dfde750cc7d" AI评分知识库版本: "knowledge_base_v1-20260819+runtime.59" AI评分知识库SHA256: "630ed34f914ee61e1f2c29bf5210d738043637a2eaf7d9e8a3ea68614a92fc92" AI评分知识库检索词: "[\"AMD\",\"Rack scale AI\",\"site:substack.com Rack scale AI\",\"CPO\",\"rack-scale\",\"GPU\",\"EPYC\",\"CPU\",\"GPU-centric\",\"GPUs\",\"ASICs\",\"CPUs\"]" AI评分知识库命中: "[{\"id\":\"july-correct-0034\",\"title\":\"AMD Fires Back at Nvidia with Helios AI System, Epyc CPUs\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"AMD\",\"rack-scale\",\"GPU\",\"EPYC\",\"CPU\",\"GPUs\",\"CPUs\"],\"rank\":-24.735214243301883},{\"id\":\"july-correct-0081\",\"title\":\"AAI 2026: 6th Gen AMD EPYC Server CPUs Power the Agentic Data Center\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"AMD\",\"rack-scale\",\"GPU\",\"EPYC\",\"CPU\",\"GPUs\",\"CPUs\"],\"rank\":-20.202984974036767},{\"id\":\"july-correct-0075\",\"title\":\"From EPYC to Helios, AMD and Meta are Scaling the Future of AI Together\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"AMD\",\"rack-scale\",\"GPU\",\"EPYC\",\"CPU\",\"GPUs\",\"CPUs\"],\"rank\":-18.287814335120796},{\"id\":\"july-correct-0068\",\"title\":\"Microsoft Azure Expanding AI Infra Choice with AMD Helios™\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"AMD\",\"rack-scale\",\"GPU\",\"EPYC\",\"CPU\",\"GPUs\",\"CPUs\"],\"rank\":-17.188437104712484},{\"id\":\"july-correct-0079\",\"title\":\"AAI 2026: AMD Launches AMD Helios Rackscale Solution for Frontier AI\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"AMD\",\"rack-scale\",\"GPU\",\"EPYC\",\"CPU\",\"GPUs\",\"CPUs\"],\"rank\":-16.554832359795085}]" AI摘要: "AMD数据中心战略正从CPU/GPU芯片竞争转向AI计算平台，通过EPYC、Instinct GPU、Helios机架架构、定制ASIC及SoIC/CoWoS封装与CPO等技术，实现异构集成以降低数据移动成本，并将竞争范围从芯片扩展到封装。" AI摘要模型: "ali-deepseek-v4-flash" AI摘要时间: "2026-08-31T23:37:36.264Z" 采集批次: "2026年8月31日11点14分25秒" 采集批次ID: "20260831-111425-728" 去重键: "https://tspasemiconductor.substack.com/p/amds-ai-architecture-evolution-from" --- AMD’s data center strategy is evolving from traditional CPU- and GPU-centric chip competition toward a comprehensive AI computing platform. Across next-generation EPYC processors, Instinct GPUs, chiplets, SoIC/CoWoS advanced packaging, the Helios rack-scale architecture, custom ASICs, and CPO, a common technology direction is emerging: **reducing the cost of data movement through heterogeneous integration while expanding the scope of competition from the chip level to the package, node, rack, and ultimately the entire AI cluster.** We will share the following topics **1\. Custom ASICs: Extending AMD Beyond General-Purpose CPUs and GPUs** **2\. EPYC’s Core Advantage Is Shifting from Core Count to Chiplet Architecture** **3\. Memory and I/O Are Becoming the Next EPYC Bottlenecks** **4\. CPU and GPU Architectures Are Moving from External Connectivity Toward Platform-Level Integration** **5\. Helios: AMD’s Transition from a Chip Company to Rack-Scale Computing** **6\. Instinct GPUs: Moving from Process Scaling to Advanced Packaging Scaling** **7\. SoIC and CoWoS: Two Different Dimensions of GPU Scaling** **8\. Advanced Packaging Competition Is Shifting from Capacity Toward Yield** **9\. The Next Generation of Instinct: System Scaling Matters More Than Standalone GPU Performance** **10\. CPO: Addressing the Next Electrical I/O Bottleneck** **11\. AMD’s Technology Roadmap Is Fundamentally About Solving Data Movement** ![图片](./assets/img-ca5a0f1d.webp) ![图片](./assets/img-ca5a0f1d.webp)