西部某市干部变动，英伟达B200热度大减 西部某省此前大力推进智算业务，高峰时期规划近200个算力项目。近期，该省某市的核心干部变动，智 算业务负责人随之调整，导致已建成投产的算力项目在竣工验收环节遇阻——新任负责人均未批准，大批 项目搁置，直接引发市场行情降温。 对于此前的算力项目，虽仍有批准可能，但验收标准已提高。业内人士分析，在近200个项目中，最终 能"善终"的或仅5-10个。作为算力设备采购的绝对主力，当地项目受阻直接导致B200等算力设备热度大 幅下降。 H100单次维修费用数万元，英伟达GPU维保市场巨大 智算市场火热，国内也已经有上百万张英伟达高性能GPU，海量GPU的后市场业务十分可观。由于数据中 心GPU长时间高温工作，显存故障的概率很高，而维修的成本又很高。以维修一张H100为例，硬件维 修，一次维修的费用高达2-3万元，几乎是售价的10%，而固件的更新，价格更是高达10万甚至20万。 有客户资源、产业链资源的人已经准备做英伟达GPU的维保生意，不过英伟达可能也会自己下场做后市 场。更多英伟达GPU维保市场的交流 添加微信 BENSONEIT 。 浪潮、华勤技术调研服务器回收及维修市场， ...

Core Viewpoint - The introduction of UE8M0 FP8 by Deepseek for the upcoming domestic chips signifies a strategic move to enhance compatibility and efficiency in the Chinese AI ecosystem, addressing the unique requirements of domestic hardware [5][10][12]. Group 1: UE8M0 and FP8 Concept - FP8 is an 8-bit floating-point format that significantly reduces memory usage by 75% compared to 32-bit formats, enhancing computational speed and efficiency for large model training and inference [7][13]. - UE8M0 is a specific encoding format for FP8 tensor data, designed to optimize compatibility with domestic chips, differing from Nvidia's E4M3 and E5M2 formats which focus on precision and dynamic range [9][10]. - The Open Compute Project (OCP) introduced UE8M0 as part of its MXFP8 formats, aiming to standardize FP8 usage across various hardware platforms [8]. Group 2: Strategic Importance of UE8M0 - The development of UE8M0 is crucial for ensuring that domestic chips can effectively utilize FP8 without relying on foreign standards, thus reducing dependency on Nvidia's technology [12]. - Deepseek's integration of UE8M0 into its model development process aims to ensure that models can run stably on upcoming domestic chips, facilitating a smoother transition from development to deployment [11][12]. - The focus of UE8M0 is not to outperform foreign FP8 standards but to provide a viable solution that allows domestic chips to leverage FP8 efficiency [14]. Group 3: Performance and Limitations - UE8M0 can save approximately 75% in memory usage compared to FP32, allowing for larger models or increased request handling during inference [13]. - The inference throughput using UE8M0 can be about twice that of BF16, making it particularly beneficial for large-scale AI applications [13]. - However, UE8M0 is not a one-size-fits-all solution; certain calculations still require higher precision formats like BF16 or FP16, and effective calibration is necessary to avoid errors in extreme value scenarios [15].

Group 1: Foreign Investment Policies - The Ministry of Commerce will further relax foreign investment market access and revise the "Encouraging Foreign Investment Industry Catalog" to provide more opportunities for foreign enterprises in emerging fields and future industries [1] - The implementation of the "2025 Action Plan for Stabilizing Foreign Investment" will accelerate pilot openings in cloud computing, biotechnology, and wholly-owned hospitals [2] Group 2: Housing Fund Policy - Several cities have initiated mutual recognition and lending of housing provident funds, allowing for cross-city loan approvals and fund disbursement [3] - The push for mutual recognition and lending of housing funds is aimed at stabilizing the real estate market and facilitating talent mobility across regions [4] Group 3: Guangzhou Consumption Policy - Guangzhou's draft plan proposes to fully lift restrictions on housing purchases, sales, and prices, and to lower down payment ratios and interest rates to better meet housing consumption needs [5] - The plan emphasizes increasing residents' income and enhancing consumption capacity as prerequisites for stimulating consumption and expanding domestic demand [6] Group 4: AMD AI Chip Launch - AMD launched the new MI350 AI chip series, claiming it outperforms NVIDIA's GB200 in various computational tasks, with a memory capacity 1.6 times greater than that of GB200 [7] - The MI350 series is positioned as a competitive alternative to NVIDIA's offerings, particularly in terms of cost-effectiveness and operational efficiency [8] Group 5: Early Childhood Education and Elderly Care - Many kindergartens are exploring a new model of combined care for the elderly and children, with some transforming into community nursing homes due to declining enrollment [9] - The concept of "old-young co-care" faces practical challenges, as the needs of children and the elderly differ significantly [10] Group 6: Burger King's Strategic Adjustments - RBI, the parent company of Burger King, plans to close underperforming stores while opening 40 to 60 new locations in key urban areas [11] - The company aims to enhance its local management team and adapt its offerings to better meet the preferences of Chinese consumers [12] Group 7: UBS and ICBC Credit Suisse Fund - UBS has become the second-largest shareholder of ICBC Credit Suisse Fund, holding a 20% stake, following its acquisition of Credit Suisse [13] - UBS's investment in the Chinese market has been ongoing since 1989, and it continues to expand its presence in the asset management sector [14] Group 8: Stock Market Performance - The stock market experienced a decline, with the Shanghai Composite Index falling by 0.75% amid geopolitical tensions affecting investor sentiment [15] - The market's performance reflects a lack of confidence in continued upward movement, with expectations for potential positive developments from upcoming financial meetings [16]

Core Viewpoint - AMD's CEO, Lisa Su, stated that the company's latest AI processors are expected to challenge NVIDIA's chips, with the AI market projected to exceed $500 billion in the next three years [1][3]. Group 1: Product Development and Performance - The new MI350 series chips are reported to outperform NVIDIA's competing products, with the MI355 chip showing a performance improvement of 35 times compared to its predecessor [1]. - AMD believes that the MI series will revitalize growth and demonstrate its ability to compete with larger rivals [3]. - The MI355 chip is said to perform better than NVIDIA's B200 and GB200 products when running AI software, and it is priced significantly lower than its competitors [3]. Group 2: Market Dynamics and Competition - Despite AMD's advancements, it still lags behind NVIDIA in the AI accelerator market, which is crucial for developing and running AI tools [1][3]. - The AI accelerator business has helped AMD escape the pressure from Intel in the PC processor market, but NVIDIA has surpassed both companies in revenue, with NVIDIA's annual revenue exceeding $100 billion [3]. - The demand for computing infrastructure is experiencing explosive growth, driven by significant investments from global giants [3][4]. Group 3: Future Outlook and Strategic Moves - Lisa Su announced the upcoming MI400 series, which is expected to establish a clear advantage over NVIDIA's chips released at the same time [3]. - The MI400 will include enhanced memory and high-speed data retrieval components, which are critical for running AI software [3]. - AMD is actively lobbying the U.S. government to ease export restrictions on AI components to other countries, emphasizing the need for more trade freedom to maintain the U.S. technology's core position in global AI development [4].

以下文章来源于IC后摩院 ，作者赵瑜斌 IC后摩院 . 产业为天，学术为地，搬运在这天地间 如果您希望可以时常见面，欢迎标星收藏哦~ | Contents | | | --- | --- | | 引 | | | 01 | 架构设计 | | 02 | 设计实现 | | 03 | 仿真 | | 04 | PV验证/签核 | | 05 | 供电 / 功耗 | | 06 | 标准/底座/生态 | | 07 | 商用工具现状 | 本篇主要分享从设计视角，对于Chiplet tool的真实需求。在开始前，我们略去了Chiplet设计的必要性和 后摩路径的好处（在其他篇中我们再分享），但是我们从一个基本的角度来看必须用Chiplet来构建未来系 统，尤其是算力系统的必要性——晶体管transistor的增长。 上个月底参加了HiPi联盟大会，以及在近期多场和3DIC、先进封装有关的会议中，国内设计界对EDA的呼声 可谓此起彼伏。这几天方得空整理了一下若干专家对此的讨论和分享。由于流程环节繁多，内容庞大，本文仅 挑一些要点做分享，如有更适合的场合再分别做详细介绍。 2024年，成功商用单片（苹果M3 max@3nm）晶体管最 ...

■ 老铺黄金销售逆势高增，使其在高端商业体中的重要性显著提升，2024年强劲业绩提升了租赁谈判 筹码。 ■ 截至2024年底，公司自营门店达 37家 ，2024年净新增 7家 ，覆盖 15个城市 ，包括 北上广深 和7 个新一线城市。 ■ 进驻 杭州万象城、成都IFS、南京德基广场、澳门威尼斯人 等高端购物中心，SKP与万象城体系门 店占比高。 ■ 37家门店中 16家位于一层 ，如 北京SKP、北京国贸商城、天津万象城 等奢侈品核心地段，最大化 品牌曝光度。 ■ 2023-2024年H1 单店收入同比翻倍增长 ，部分核心门店（北京SKP、上海豫园、沈阳万象城）表现 更优。 2、 《 AI芯片功耗提升，高功率电源景气上行》摘要 ■ 英伟达B200 （N4P制程，2080亿晶体管）功耗达 1000W ，远超H100（800亿晶体管，N4制 程）。 ■ Si（硅）、SiC（碳化硅）、GaN（氮化镓） 的组合能提升AI服务器电源系统的 可持续性与可靠性 。 3、 《 2025年政府工作报告解读：更高水平开放》摘要 ■ 会议报告特别强调了 中国-东盟自贸区3.0 、 DEPA（数字经济伙伴关系协定） 、 CPTPP（ ...