如果您希望可以时常见面，欢迎标星收藏哦~ 人工智能的高速发展，给芯片的算力带来了新挑战，伴之而来的就是越来越严峻的"内存 墙"问题。这一切，则需要从统治大半个处理器芯片江湖的冯诺依曼架构说起。作为一个被 广泛应用的架构，冯·诺依曼架构推广了存储程序概念，使计算机更加灵活，更易于重新编 程。这种设计将数据和指令存储在同一内存中，简化了硬件设计，并实现了通用计算。 在一开始，这种创新的架构也的确给处理器产业带来了很大的提升。但在后续的发展中，无论我们 采取何种措施来提升性能，都无法回避一个事实：指令一次只能执行一条，并且只能按顺序执行。 这两个因素都限制了 CPU 的执行能力。与此同时，另一个问题也随之而生。 熟悉的读者应该知道，在当前的人工智能应用，搬运越来越多的数据是一个常态，但由于存储器的 性能发展跟不上 CPU 的性能增长，导致 CPU 需要花费大量的时间等待存储器完成读写操作，从 而降低了系统的整体性能，这就是所谓的"内存墙"问题。 为了解决这些问题，行业尝试了很多解决办法，当中以存算一体作为关注。 存算一体，走向台前 如图所示，当前的主流PIM技术主要分为以下几种，他们都各具优势，面临的挑战也各不相同 ...

（2025 年 6 月 30 日，北京）—— 中国 IC 独角兽联盟今日正式发布 "2024-2025 年度第八届中国 IC 独角兽" 榜 单，评选出 12 家中国 IC 独角兽企业及 2 家新锐企业。本次评选聚焦集成电路全产业链，旨在挖掘具有高成长 性和技术突破的创新企业，通过系统化估值与行业赋能，助力中国半导体产业在全球竞争中实现高质量发展。 一、评选背景与核心目标 本次评选是在连续成功举办七届活动的基础上，由中国 IC 独角兽联盟发起并主办，覆盖集成电路设计、制 造、封装、测试、材料、装备等全产业链环节。评选通过技术创新、市场竞争力、生态协同等多维度评估，最 终从近百份申报企业中遴选出 12 家独角兽企业 和 2 家新锐企业 ，其技术布局与市场表现代表了中国半导体行 业的前沿水平。 二、独角兽企业：全产业链创新标杆 本次入选的 12 家独角兽企业覆盖 存储与计算领域 、 通信与射频领域 、 高端制造与材料领域 、 汽车电子与 智能终端、 核心器件与先进封装、消费电子与物联网等关键领域，其技术突破与市场落地为行业树立了标 杆： 三、新锐企业：未来增长极 本次评选设立 "新锐企业" 奖项，表彰成立时间较短但 ...

Core Insights - The 2025 CCF Youth Elite Conference (YEF2025) is being held in Zhuhai, Guangdong, focusing on diverse development paths in artificial intelligence (AI) [1] - The conference features various sessions including keynote speeches, forums, and specialized discussions, emphasizing the need for foundational breakthroughs and original innovations in AI [1][2] Group 1: Conference Overview - The conference is attended by over 1,000 young experts and scholars from the computer field, discussing the future of AI [1] - The event includes 7 keynote reports, 5 thought showcase reports, 2 main forums, and 22 specialized forums [1] Group 2: Key Themes and Innovations - The conference theme "Intelligent New Paths" reflects the current state of AI, highlighting the transition from lagging behind to achieving parity in certain areas [2] - The need for leading innovations in AI is emphasized, particularly in light of the maturity of large language models [2] Group 3: Research and Development Focus - The Guangdong-Hong Kong-Macao Greater Bay Area is positioned as a high-level open frontier for technology development, with significant policies already implemented [2] - The Guangdong Provincial Institute of Intelligent Science and Technology is focusing on brain science and brain-like intelligence, establishing research centers in cognitive neural networks, brain-like computing systems, and intelligent biomedicine [2] Group 4: Class Brain Intelligence - Academician Zhang Xu discusses class brain intelligence as a crucial future industry direction, aiming to overcome traditional computing limitations by mimicking biological brain mechanisms [3] - Class brain intelligence emphasizes characteristics such as parallelism, integrated storage and computation, and plasticity, which are essential for processing unstructured data efficiently [3] Group 5: Technological Advancements - Professor Ye Le from Peking University introduced a three-dimensional storage-computation integrated AI chip to address challenges like the diminishing effectiveness of Moore's Law and increasing computational demands [6] - The long-term vision includes integrating this chip technology into robots, suggesting a future where the number of robots could exceed that of humans [6]