<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>智算网络 on MessageDaily</title><link>https://inkeast.github.io/MessageDaily/tags/%E6%99%BA%E7%AE%97%E7%BD%91%E7%BB%9C/</link><description>Recent content in 智算网络 on MessageDaily</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Thu, 24 Sep 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://inkeast.github.io/MessageDaily/tags/%E6%99%BA%E7%AE%97%E7%BD%91%E7%BB%9C/index.xml" rel="self" type="application/rss+xml"/><item><title>当Token成本成为AI的商业闭环：一场论坛透出的下一代智算基础设施全景</title><link>https://inkeast.github.io/MessageDaily/posts/2026-09-24-yunqi2026-h3c-token-cost-performance-computing/</link><pubDate>Thu, 24 Sep 2026 00:00:00 +0000</pubDate><guid>https://inkeast.github.io/MessageDaily/posts/2026-09-24-yunqi2026-h3c-token-cost-performance-computing/</guid><description>本文整理自云栖大会2026「新华三：打造Token极致性价比，下一代智算关键技术创新论坛」全场实录。十位来自新华三、阿里云、英特尔、平头哥、博通、清程极智、是石科技的讲者，从超节点、异构协同、智能网卡、跨域网络到KV Cache存储新架构，勾勒出同一条主线：当AI从不计成本地堆算力转向逐个Token地算成本，基础设施的每一层——芯片、网络、存储、软件——都要围绕&amp;quot;每个Token的性价比&amp;quot;重新设计。</description></item><item><title>算力单位不再是芯片，而是Pod——云栖UALink实践专场：开放互连从规范走向落地的第一年</title><link>https://inkeast.github.io/MessageDaily/posts/2026-09-24-yunqi2026-ualink-practice-ecosystem/</link><pubDate>Thu, 24 Sep 2026 00:00:00 +0000</pubDate><guid>https://inkeast.github.io/MessageDaily/posts/2026-09-24-yunqi2026-ualink-practice-ecosystem/</guid><description>云栖2026「超节点Scale-Up开放互连与开源计算生态：UALink实践」专场复盘：带宽年增1.4倍追不上算力年增3倍，互连成为成本与延迟主战场。UALink两年从8家发起长到110多家成员，2.0版发布四份规范，2027年首批交换机与加速器将至；在网计算把集合通信从GPU搬进交换机，内存语义让FPGA与内存池成为域内新公民，协议胜负手是端到端互通性。讲者名按议程校正，规格经外部核验。</description></item></channel></rss>