慣性聚合 高效追讀感興趣之博客、新聞、科技資訊
閱原文 以慣性聚合開啟

推薦訂閱源

小众软件
小众软件
博客园 - 叶小钗
有赞技术团队
有赞技术团队
大猫的无限游戏
大猫的无限游戏
博客园_首页
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
L
LangChain Blog
Hugging Face - Blog
Hugging Face - Blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
aimingoo的专栏
aimingoo的专栏
Blog — PlanetScale
Blog — PlanetScale
爱范儿
爱范儿
T
Tailwind CSS Blog
Jina AI
Jina AI
量子位
Stack Overflow Blog
Stack Overflow Blog
人人都是产品经理
人人都是产品经理
J
Java Code Geeks
V
Visual Studio Blog
月光博客
月光博客

WIRED

‘Avatar: Aang, The Last Airbender’ Leaked Online. Some Fans Say Paramount Deserves the Fallout NASA Wants to Put Nuclear Reactors on the Moon AI Could Democratize One of Tech's Most Valuable Resources Microsoft Surface PCs Are Getting Big Price Hikes, and the Cheaper Models Are Going Away Why Amazon Is Buying Globalstar—and What It Means for Your iPhone The US Government Will Ask Data Centers How Much Power They Use MAGA Is Starting to Look Beyond Trump Allbirds Is Pivoting to AI Compute. Sure, Why Not Best Smart Smoke Detector (and Why You Still Need a Dumb One) 12 Best Standing Desks of 2026, Tested and Reviewed Best Wi-Fi Routers of 2026 for Working, Gaming, and Streaming Best GoPro Camera (2026): Compact, Budget, Accessories The Caves That Could Help Us Find, or Become, Aliens AI Slop Is Making the Internet Fake-Happy The Deepfake Nudes Crisis in Schools Is Much Worse Than You Thought In the Wake of Anthropic’s Mythos, OpenAI Has a New Cybersecurity Model—and Strategy Telegram Is Still Hosting a Sanctioned $21 Billion Crypto Scammer Black Market The FCC Has a Fast Lane for Complaints About Trump’s Media Critics Top iRestore Deals for Hair Growth and LED Therapy Devices Meta Is Warned That Facial Recognition Glasses Will Arm Sexual Predators You Should Be More Freaked Out by Shingles BYD’s Fastest-Charging Car in the World Is Astonishing—in Good and Bad Ways The 4 Best Water Filter Pitchers (2026): PFAS, Microplastics The Internet's Most Powerful Archiving Tool Is in Peril The Dumbest Hack of the Year Exposed a Very Real Problem AI Agents Are Coming for Your Dating Life ‘The Audacity’ Is the Broligarchy Takedown You Were Waiting For Why Is It So Hard to Fix an Electric Bike? (2026) Best 2-in-1 Laptops (2026): Microsoft, Lenovo, and the iPad There’s a Secret Ingredient to Making Luxury Ice at Home The Screen Time Maximalists Who Spend an Ungodly Amount of Time on Their Phones Mammotion Spino E1 Review: A Budget Pool Bot That Comes Up Short The 3 Best Mug Warmers of 2026: Better Than the Ember Mug? Zuvi ColorBox Review: Please Just Go to a Professional MacBook Neo vs. MacBook Air: Which One Should You Buy? Best Electric Cargo Bikes (2026): Urban Arrow, Lectric, Tern, and More ‘Crimson Desert’ Is a Cat Dad Simulator Your Push Notifications Aren’t Safe From the FBI Flight Path Data Shows How Mosquitoes Target Humans How the Internet Broke Everyone’s Bullshit Detectors The All-Clad Factory Seconds Sale Is Back—for Now (2026) Artemis II Astronauts Safely Return to Earth After Historic Flight Around the Moon Home Depot Spring Black Friday (2026): Best Tool and Grill Deals Motorola’s Souped-Up Folding Phone Is Almost Half Off Anthropic’s Mythos Will Force a Cybersecurity Reckoning—Just Not the One You Think The Future of the Artemis Program Is Riding on Reentry Suspect Arrested for Allegedly Throwing Molotov Cocktail at Sam Altman’s Home "Uncanny Valley": OpenAI and Musk Fight Again; DOJ Mishandles Voter Data; Artemis II Comes Home This Clever Bike Bell Can Even Be Heard by People Wearing Noise-Canceling Headphones This Startup Wants You to Pay Up to Talk With AI Versions of Human Experts
华为之“芯片女王”掷下战书
Will Knight · 2026-05-28 · via WIRED

情势渐趋复杂于人工智能芯片之竞逐。

丁波何,總統也华为(Huawei)“的”芯片设计子曰:HiSilicon(海思半导体)之属,谓其匠者,创新法以调半导体,且信数载之内,可弥中西方芯片之能差。

华为之法,简言之,重在加速芯片、电路乃至整个计算系统之运算,而非将更多组件压缩于单一硅片之上。

“吾等觅得新途,”贺言。 去年周末于上海之IEEE国际电路与系统会议。彼,中国称之为华为“芯片之后”,许公司将以新芯片,于未来数月,证此新法之可行。

“二零二六冬节之前,吾将献惊喜。”贺氏言。“非饱和,非延续,乃大进也。”

芯片之主谓新法曰 Tau 之增律,言其已代 摩尔之律 为海思之圭臬。摩尔之律,以英特尔合创者戈登·摩尔名之,其旨曰:计算之进,在于每两载将芯片所纳之晶体管,或逻辑门,倍增其数。

铸造尖端芯片,今需以价值连城之蚀刻设备,于硅晶之上镌刻组件,此乃供应链之精妙绝伦,亦需广博之工程学识。

美利坚之出口管制,禁华为与台积电(TSMC,全球首屈一指之晶圆代工厂)合作。华为乃转赖中国之中芯国际(SMIC),其使用之光刻机乃较旧之世代。尤关紧要者,此限制损及中国发展尖端人工智能之能力,不得自用其硅晶。据某些估算,其落后于前沿者,已逾五载。

然芯片之业已初遇摩尔定律之限。当晶体管仅数纳米之宽,量子之效扰其常行。众芯片已多采变通之法:如苹果之最强大处理器,即由二片相缀欲成一力更巨之器。

华为之告,似谓已得越此限之法。亦似谓欲挫中国芯片之禁令,反激新创,历时而使国得建更精之芯片业,与西竞。终,华为等之新创,或可蚀美之技优。

“六载之前,吾等之几何缩放已至平台,”彼于周末言,所指乃光刻微缩之术。“吾等旋觉,半导体之演进非止几何缩放。”

彼举数端,言公司以新法增进芯片之能。其内含所谓LogicFolding者,能减电路中施以要务逻辑之所需时。

海思云谷言,亦以考量纳米尺度电现象、设计芯片协同作业、研发加速芯片间通讯之互连,此乃训练大型人工智能模型之要诀也.

“无论[人工智能]训练抑或推演,其胜非惟缩短计算之时,亦在缩短数据于芯片间流转及芯片内运行之期。”彼言如是。

华为云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云云

其宣告非谓华为有明径以胜美之禁令,亦非众皆信其必可行。独立之半导体与人工智能政策分析师海尔曼·莱纳特曰,华为之策示其于单凭缩小与密集成芯片以榨取更多性能,已遇极限。彼曰,华为益赖混合键合与三维芯片堆叠之技以增性能。

然华为之芯片女王,似信此公司能易局。彼言于演辞曰:“此创新将入量产。”或非今年,然自二二七年后矣。


此乃 威尔·奈特之 人工智能实验室简报之一期。欲读旧简报 于此。