+关注
leisheng526
暂无个人介绍
IP属地:北京
16
关注
0
粉丝
4
主题
0
勋章
主贴
热门
leisheng526
12-04
苹果继续这么整花活,他的APPLE AI会永远停滞不前
苹果采用亚马逊芯片,一个去英伟达化的信号?
去老虎APP查看更多动态
{"i18n":{"language":"zh_CN"},"userPageInfo":{"id":"3492992533487256","uuid":"3492992533487256","gmtCreate":1597457366556,"gmtModify":1597457366556,"name":"leisheng526","pinyin":"leisheng526","introduction":"","introductionEn":"","signature":"","avatar":"https://static.laohu8.com/default-avatar.jpg","hat":null,"hatId":null,"hatName":null,"vip":1,"status":2,"fanSize":0,"headSize":16,"tweetSize":2,"questionSize":0,"limitLevel":999,"accountStatus":4,"level":{"id":0,"name":"","nameTw":"","represent":"","factor":"","iconColor":"","bgColor":""},"themeCounts":4,"badgeCounts":0,"badges":[],"moderator":false,"superModerator":false,"manageSymbols":null,"badgeLevel":null,"boolIsFan":false,"boolIsHead":false,"favoriteSize":0,"symbols":null,"coverImage":null,"realNameVerified":null,"userBadges":[{"badgeId":"228c86a078844d74991fff2b7ab2428d-3","templateUuid":"228c86a078844d74991fff2b7ab2428d","name":"投资合伙人虎","description":"证券账户累计交易金额达到100万美元","bigImgUrl":"https://static.tigerbbs.com/fbeac6bb240db7da8b972e5183d050ba","smallImgUrl":"https://static.tigerbbs.com/436cdf80292b99f0a992e78750ac4e3a","grayImgUrl":"https://static.tigerbbs.com/506a259a7b456f037592c3b23c779599","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.11.07","exceedPercentage":"93.01%","individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1101},{"badgeId":"e50ce593bb40487ebfb542ca54f6a561-3","templateUuid":"e50ce593bb40487ebfb542ca54f6a561","name":"偶像虎友","description":"加入老虎社区1500天","bigImgUrl":"https://static.tigerbbs.com/8b40ae7da5bf081a1c84df14bf9e6367","smallImgUrl":"https://static.tigerbbs.com/f160eceddd7c284a8e1136557615cfad","grayImgUrl":"https://static.tigerbbs.com/11792805c468334a9b31c39f95a41c6a","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2024.09.24","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1001},{"badgeId":"976c19eed35f4cd78f17501c2e99ef37-1","templateUuid":"976c19eed35f4cd78f17501c2e99ef37","name":"博闻投资者","description":"累计交易超过10只正股","bigImgUrl":"https://static.tigerbbs.com/e74cc24115c4fbae6154ec1b1041bf47","smallImgUrl":"https://static.tigerbbs.com/d48265cbfd97c57f9048db29f22227b0","grayImgUrl":"https://static.tigerbbs.com/76c6d6898b073c77e1c537ebe9ac1c57","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2021.12.21","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1102},{"badgeId":"518b5610c3e8410da5cfad115e4b0f5a-1","templateUuid":"518b5610c3e8410da5cfad115e4b0f5a","name":"实盘交易者","description":"完成一笔实盘交易","bigImgUrl":"https://static.tigerbbs.com/2e08a1cc2087a1de93402c2c290fa65b","smallImgUrl":"https://static.tigerbbs.com/4504a6397ce1137932d56e5f4ce27166","grayImgUrl":"https://static.tigerbbs.com/4b22c79415b4cd6e3d8ebc4a0fa32604","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2021.12.21","exceedPercentage":null,"individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1100},{"badgeId":"35ec162348d5460f88c959321e554969-2","templateUuid":"35ec162348d5460f88c959321e554969","name":"宗师交易员","description":"证券或期货账户累计交易次数达到100次","bigImgUrl":"https://static.tigerbbs.com/ad22cfbe2d05aa393b18e9226e4b0307","smallImgUrl":"https://static.tigerbbs.com/36702e6ff3ffe46acafee66cc85273ca","grayImgUrl":"https://static.tigerbbs.com/d52eb88fa385cf5abe2616ed63781765","redirectLinkEnabled":0,"redirectLink":null,"hasAllocated":1,"isWearing":0,"stamp":null,"stampPosition":0,"hasStamp":0,"allocationCount":1,"allocatedDate":"2021.12.21","exceedPercentage":"80.48%","individualDisplayEnabled":0,"backgroundColor":null,"fontColor":null,"individualDisplaySort":0,"categoryType":1100}],"userBadgeCount":5,"currentWearingBadge":null,"individualDisplayBadges":null,"crmLevel":8,"crmLevelSwitch":0,"location":"北京","starInvestorFollowerNum":0,"starInvestorFlag":false,"starInvestorOrderShareNum":0,"subscribeStarInvestorNum":2,"ror":null,"winRationPercentage":null,"showRor":false,"investmentPhilosophy":null,"starInvestorSubscribeFlag":false},"baikeInfo":{},"tab":"post","tweets":[{"id":377986533458280,"gmtCreate":1733324589550,"gmtModify":1733326424521,"author":{"id":"3492992533487256","authorId":"3492992533487256","name":"leisheng526","avatar":"https://static.laohu8.com/default-avatar.jpg","crmLevel":8,"crmLevelSwitch":0,"followedFlag":false,"idStr":"3492992533487256","authorIdStr":"3492992533487256"},"themes":[],"htmlText":"苹果继续这么整花活,他的APPLE AI会永远停滞不前","listText":"苹果继续这么整花活,他的APPLE AI会永远停滞不前","text":"苹果继续这么整花活,他的APPLE AI会永远停滞不前","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/377986533458280","repostId":"2488917961","repostType":2,"repost":{"id":"2488917961","kind":"news","pubTimestamp":1733320355,"share":"https://www.laohu8.com/m/news/2488917961?lang=&edition=full","pubTime":"2024-12-04 21:52","market":"hk","language":"zh","title":"苹果采用亚马逊芯片,一个去英伟达化的信号?","url":"https://stock-news.laohu8.com/highlight/detail?id=2488917961","media":"茶饮消息","summary":"亚马逊宣布推出了未来可能替代英伟达GPU的AI芯片。AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,而且苹果宣布会使用它来训练和部署Apple Intelligence。但随着Trainium2的发布,亚马逊已经做出了重大的调整,正在芯片、系统和软件编译器/框架层面向英伟达产品看齐,提供有竞争力的定制硅芯片解决方案。苹果还一直使用亚马逊的Inferentia和Graviton芯片来支持搜索服务。苹果方面称,亚马逊的这些芯片带来了40%的效率提升。","content":"<html><body><p><strong><a href=\"https://laohu8.com/S/AMZN\">亚马逊</a>宣布推出了未来可能替代<a href=\"https://laohu8.com/S/NVDA\">英伟达</a>GPU的AI芯片。</strong>AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,<strong>而且<a href=\"https://laohu8.com/S/AAPL\">苹果</a>宣布会使用它来训练和部署Apple Intelligence。</strong></p><p>正在拉斯维加斯举行的re:Invent大会上,AWS推出的Trn2服务器(16片Trainium2)提供20.8 Pflops性能,可训练数十亿参数模型,试图媲美英伟达和AMD的GPU。Trn2 UltraServers(64片Trainium2)高峰时提供83.2 Pflops算力,完全可以用来训练和部署最大的模型,包括语言、多模态和视觉模型。</p><p>AWS还宣布了下一代AI芯片Trainium3的计划。预计将比Trainium2提升2倍的性能,改善40%的能效,3纳米制程,将于2025年底推出。</p><p>目前,亚马逊基于Trainium1和Inferentia2的实例,在生成式AI前沿模型训练或推理方面还不太具有竞争力,这是由于硬件规格较弱以及软件集成较弱所致。但随着Trainium2的发布,亚马逊已经做出了重大的调整,<strong>正在芯片、系统和软件编译器/框架层面向英伟达产品看齐,提供有竞争力的定制硅芯片解决方案。</strong></p><p><strong>苹果公司也非同寻常地对外介绍了它与云服务商之间的合作细节,并且表达出积极合作构建AI的意愿。</strong>苹果已经使用AWS服务超过十年,用于Siri、Apple Maps和Apple Music。苹果还一直使用亚马逊的Inferentia和Graviton芯片来支持搜索服务。苹果方面称,亚马逊的这些芯片带来了40%的效率提升。</p><p>苹果最近还将用Trainium2进行其自有模型的预训练。苹果对它的初步评估显示,预训练效率提升了50%。苹果在决定开发Apple Intelligence之后,马上就找到了AWS,寻求AI基础设施的支持。苹果也使用过<a href=\"https://laohu8.com/S/GOOG\">谷歌</a>云的TPU服务器。</p><p>苹果正在引领着个人AI的应用方向,即把AI模型向端侧部署,主要用本地化的计算为用户提供定制化与个人化的AI服务,注重保护用户的隐私。</p><p>所以,对于苹果来说,<strong>最重要的不是用十万张卡去训练大模型,而是用AI更好地服务其20多亿设备用户,</strong>即在iPhone、iPad、Mac等设备上,用自研芯片提供本地算力,只有那些较复杂的计算任务才上云端。苹果还需要云服务商能配合它进行隐私计算。</p><p>Apple Intelligence有自己的步调,它先推出内容提炼、起草邮件、生成表情包等最基本的功能,很快会集成OpenAI的大模型服务,明年会加强Siri功能,因为<a href=\"https://laohu8.com/S/5RE.SI\">智能</a>体技术的加持,它更像个能办事的助理,调动手机App完成用户吩咐的任务。</p><p><strong>而且AWS正在与Anthropic合作,打造40万Trainium2卡级的算力集群,用来训练下一代的Claude大模型。</strong>这个项目名称为Project Rainer,将为Anthropic提供的算力5倍于训练现有模型的Eflops。亚马逊对Anthropic最新的40亿美元投资,实际上将用于这个40万卡集群,目前还没有其他主要客户。</p><p>马斯克的xAI已经建好了10万H100算力集群,而且放也豪言要再买30万张B200;扎克伯格正在用一个超过10万H100的集群加班加点地训练Llama4,更不用说微软/OpenAI等,10万H100已经成为参与军备竞赛的起步价。</p><p><img src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABAQMAAAAl21bKAAAAA1BMVEXy8vJkA4prAAAACklEQVQI12NgAAAAAgAB4iG8MwAAAABJRU5ErkJggg==\"/></p><p>但Trainium算力集群真的要挑战英伟达GPU,还要付出更多努力。据半导体咨询机构semianalysis分析,<strong>40万颗Trainium2的原始浮点运算性能仍少于10万GB200集群。</strong>这意味着由于阿姆达尔定律的限制,Anthropic仍将很难与竞争对手10万卡GB200集群匹敌。在40万颗Trainium2和EFA上执行集合通信将非常困难,因此Anthropic需要在异步训练方面进行一些重大创新。</p><p>*作者注:EFA代表Elastic Fabric Adapter,是AWS提供的一种高性能网络接口技术,主要用于支持高性能计算(HPC)和机器学习工作负载。</p><p>三大云巨头AWS、<a href=\"https://laohu8.com/S/MSFT\">微软</a>Azure和谷歌云,目前数据中心芯片主要来自英伟达、AMD和<a href=\"https://laohu8.com/S/INTC\">英特尔</a>。同时,它们也在积极探索自己研制芯片,会带来成本及定制化服务等方面的好处,既用于通用计算负载,也用于加速计算,如大模型的训练和推理。AWS称,通过Trainium,Anthropic的大模型Claude Haiku 3.5,速度比其他芯片提升了60%。</p><p>随着生成式AI日益进入大规模应用阶段,企业将会寻找更加适合具体应用、为客户定制化、价格更亲民、更具能效的芯片和算力解决方案。</p><p>2025年我们会看到一个趋势,<strong>更多的算力会部署到推理侧进行强化学习,以及AI的大规模应用,这些都对芯片、服务器、工具、架构、服务等提出新的定制化要求,</strong>从而为云服务商的硅技术和初创芯片企业带来新的机会。</p></body></html>","source":"ifeng_tech","collect":0,"html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>苹果采用亚马逊芯片,一个去英伟达化的信号?</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\n苹果采用亚马逊芯片,一个去英伟达化的信号?\n</h2>\n\n<h4 class=\"meta\">\n\n\n2024-12-04 21:52 北京时间 <a href=https://tech.ifeng.com/c/8f3026eRYWB><strong>茶饮消息</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>亚马逊宣布推出了未来可能替代英伟达GPU的AI芯片。AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,而且苹果宣布会使用它来训练和部署Apple Intelligence。正在拉斯维加斯举行的re:Invent大会上,AWS推出的Trn2服务器(16片Trainium2)提供20.8 Pflops性能,可训练数十亿参数模型,试图媲美英伟达和AMD的GPU。...</p>\n\n<a href=\"https://tech.ifeng.com/c/8f3026eRYWB\">Web Link</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","type":0,"thumbnail":"","relate_stocks":{"LU0203202063.USD":"AB SICAV I - ALL MARKET INCOME PORTFOLIO \"A2X\" (USD) ACC","3NVD.UK":"LS 3X NVIDIA","SG9999015986.USD":"LIONGLOBAL DISRUPTIVE INNOVATION \"I\" (USD) ACC","SNVD.UK":"LS -1X NVIDIA","AAPL":"苹果","2NVD.UK":"2X NVIDIA ETP","LU2023250330.USD":"ALLIANZ INCOME AND GROWTH \"AMG\" (USD) INC","LU0215105999.USD":"SCHRODER ISF GLOBAL EQUITY \"A\" ACC","NVDS":"Tradr 1.5X Short NVDA Daily ETF","LU0784385170.HKD":"BGF GLOBAL MULTI ASSET INCOME \"A6\" (HKDHGD) INC","LU2211815571.USD":"ALLIANZ POSITIVE CHANGE \"AT\" (USD) ACC","LU1691799644.USD":"Amundi Funds Polen Capital Global Growth A2 (C) USD","LU0648001328.SGD":"Natixis Harris Associates US Equity RA SGD","SG9999015358.SGD":"United Income Focus Trust Dis SGD-H","NVDU":"Direxion Daily NVDA Bull 2X Shares","NVD3.UK":"LS 3X NVIDIA","LU0985320562.USD":"NORDEA 1 GLOBAL STARS EQUITY \"BP\" (USD) ACC","IE0034235188.USD":"PINEBRIDGE GLOBAL FOCUS EQUITY \"A\" (USD) ACC","LU0719512351.SGD":"JPMorgan Funds - US Technology A (acc) SGD","LU0256863902.USD":"ALLIANZ US EQUITY \"AT\" (USD) ACC","LU0211327993.USD":"TEMPLETON GLOBAL EQUITY INCOME \"A\" (USD) ACC","NVDX":"T-REX 2X LONG NVIDIA DAILY TARGET ETF","LU0882574139.USD":"富达环球消费行业基金A ACC","NVDY":"YIELDMAX NVDA OPTION INCOME STRATEGY ETF","IE00BLSP4239.USD":"Legg Mason ClearBridge - Tactical Dividend Income A Mdis USD Plus","NVD2.UK":"2X NVIDIA ETP","NVDA":"英伟达","NVDS.UK":"LS -1X NVIDIA","SG9999018865.SGD":"United Global Quality Growth Fd Cl Dist SGD-H","LU2433249047.HKD":"THEMATICS META \"R/A\" (HKD) ACC","IE00BHPRN162.USD":"BNY MELLON BLOCKCHAIN INNOVATION \"B\" (USD) ACC","LU2213496289.HKD":"ALLIANZ INCOME AND GROWTH \"AT\" (HKD) ACC","LU2461242641.AUD":"WELLINGTON US QUALITY GROWTH \"A\" (AUDHDG) ACC","LU1066051811.HKD":"HSBC GIF GLOBAL EQUITY VOLATILITY FOCUSED \"AM2\" (HKD) INC","BK4566":"资本集团","LU0056508442.USD":"贝莱德世界科技基金A2","LU2265009873.SGD":"Eastspring Investments - Global Growth Equity AS SGD-H","LU1046421795.USD":"富达环球科技A-ACC","SGXZ51526630.SGD":"大华环球创新基金A Acc SGD","NVD":"GraniteShares 2x Short NVDA Daily ETF","IE00BMPRXR70.SGD":"Neuberger Berman 5G Connectivity A Acc SGD-H","LU2023250843.SGD":"Allianz Thematica Cl AT Acc H2-SGD","LU1935042215.USD":"MANULIFE GF GLOBAL MULTI-ASSET DIVERSIFIED INCOME \"AA\" (USD) INC A","NVDD":"Direxion Daily NVDA Bear 1X Shares","LU1037948541.HKD":"AB LOW VOLATILITY EQUITY PORTFOLIO \"A\" (HKD) ACC","LU2275660780.HKD":"SCHRODER ISF GLOBAL CLIMATE CHANGE EQUITY \"A\" (HKD) ACC","LU1804176565.USD":"EASTSPRING INV GLOBAL GROWTH EQUITY \"A\" (USD) ACC","LU2237957902.USD":"NIKKO AM GLOBAL EQUITY \"F\" (USD) ACC"},"source_url":"https://tech.ifeng.com/c/8f3026eRYWB","is_english":false,"share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","article_id":"2488917961","content_text":"亚马逊宣布推出了未来可能替代英伟达GPU的AI芯片。AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,而且苹果宣布会使用它来训练和部署Apple Intelligence。正在拉斯维加斯举行的re:Invent大会上,AWS推出的Trn2服务器(16片Trainium2)提供20.8 Pflops性能,可训练数十亿参数模型,试图媲美英伟达和AMD的GPU。Trn2 UltraServers(64片Trainium2)高峰时提供83.2 Pflops算力,完全可以用来训练和部署最大的模型,包括语言、多模态和视觉模型。AWS还宣布了下一代AI芯片Trainium3的计划。预计将比Trainium2提升2倍的性能,改善40%的能效,3纳米制程,将于2025年底推出。目前,亚马逊基于Trainium1和Inferentia2的实例,在生成式AI前沿模型训练或推理方面还不太具有竞争力,这是由于硬件规格较弱以及软件集成较弱所致。但随着Trainium2的发布,亚马逊已经做出了重大的调整,正在芯片、系统和软件编译器/框架层面向英伟达产品看齐,提供有竞争力的定制硅芯片解决方案。苹果公司也非同寻常地对外介绍了它与云服务商之间的合作细节,并且表达出积极合作构建AI的意愿。苹果已经使用AWS服务超过十年,用于Siri、Apple Maps和Apple Music。苹果还一直使用亚马逊的Inferentia和Graviton芯片来支持搜索服务。苹果方面称,亚马逊的这些芯片带来了40%的效率提升。苹果最近还将用Trainium2进行其自有模型的预训练。苹果对它的初步评估显示,预训练效率提升了50%。苹果在决定开发Apple Intelligence之后,马上就找到了AWS,寻求AI基础设施的支持。苹果也使用过谷歌云的TPU服务器。苹果正在引领着个人AI的应用方向,即把AI模型向端侧部署,主要用本地化的计算为用户提供定制化与个人化的AI服务,注重保护用户的隐私。所以,对于苹果来说,最重要的不是用十万张卡去训练大模型,而是用AI更好地服务其20多亿设备用户,即在iPhone、iPad、Mac等设备上,用自研芯片提供本地算力,只有那些较复杂的计算任务才上云端。苹果还需要云服务商能配合它进行隐私计算。Apple Intelligence有自己的步调,它先推出内容提炼、起草邮件、生成表情包等最基本的功能,很快会集成OpenAI的大模型服务,明年会加强Siri功能,因为智能体技术的加持,它更像个能办事的助理,调动手机App完成用户吩咐的任务。而且AWS正在与Anthropic合作,打造40万Trainium2卡级的算力集群,用来训练下一代的Claude大模型。这个项目名称为Project Rainer,将为Anthropic提供的算力5倍于训练现有模型的Eflops。亚马逊对Anthropic最新的40亿美元投资,实际上将用于这个40万卡集群,目前还没有其他主要客户。马斯克的xAI已经建好了10万H100算力集群,而且放也豪言要再买30万张B200;扎克伯格正在用一个超过10万H100的集群加班加点地训练Llama4,更不用说微软/OpenAI等,10万H100已经成为参与军备竞赛的起步价。但Trainium算力集群真的要挑战英伟达GPU,还要付出更多努力。据半导体咨询机构semianalysis分析,40万颗Trainium2的原始浮点运算性能仍少于10万GB200集群。这意味着由于阿姆达尔定律的限制,Anthropic仍将很难与竞争对手10万卡GB200集群匹敌。在40万颗Trainium2和EFA上执行集合通信将非常困难,因此Anthropic需要在异步训练方面进行一些重大创新。*作者注:EFA代表Elastic Fabric Adapter,是AWS提供的一种高性能网络接口技术,主要用于支持高性能计算(HPC)和机器学习工作负载。三大云巨头AWS、微软Azure和谷歌云,目前数据中心芯片主要来自英伟达、AMD和英特尔。同时,它们也在积极探索自己研制芯片,会带来成本及定制化服务等方面的好处,既用于通用计算负载,也用于加速计算,如大模型的训练和推理。AWS称,通过Trainium,Anthropic的大模型Claude Haiku 3.5,速度比其他芯片提升了60%。随着生成式AI日益进入大规模应用阶段,企业将会寻找更加适合具体应用、为客户定制化、价格更亲民、更具能效的芯片和算力解决方案。2025年我们会看到一个趋势,更多的算力会部署到推理侧进行强化学习,以及AI的大规模应用,这些都对芯片、服务器、工具、架构、服务等提出新的定制化要求,从而为云服务商的硅技术和初创芯片企业带来新的机会。","news_type":1},"isVote":1,"tweetType":1,"viewCount":37,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0}],"hots":[{"id":377986533458280,"gmtCreate":1733324589550,"gmtModify":1733326424521,"author":{"id":"3492992533487256","authorId":"3492992533487256","name":"leisheng526","avatar":"https://static.laohu8.com/default-avatar.jpg","crmLevel":8,"crmLevelSwitch":0,"followedFlag":false,"idStr":"3492992533487256","authorIdStr":"3492992533487256"},"themes":[],"htmlText":"苹果继续这么整花活,他的APPLE AI会永远停滞不前","listText":"苹果继续这么整花活,他的APPLE AI会永远停滞不前","text":"苹果继续这么整花活,他的APPLE AI会永远停滞不前","images":[],"top":1,"highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"link":"https://laohu8.com/post/377986533458280","repostId":"2488917961","repostType":2,"repost":{"id":"2488917961","kind":"news","pubTimestamp":1733320355,"share":"https://www.laohu8.com/m/news/2488917961?lang=&edition=full","pubTime":"2024-12-04 21:52","market":"hk","language":"zh","title":"苹果采用亚马逊芯片,一个去英伟达化的信号?","url":"https://stock-news.laohu8.com/highlight/detail?id=2488917961","media":"茶饮消息","summary":"亚马逊宣布推出了未来可能替代英伟达GPU的AI芯片。AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,而且苹果宣布会使用它来训练和部署Apple Intelligence。但随着Trainium2的发布,亚马逊已经做出了重大的调整,正在芯片、系统和软件编译器/框架层面向英伟达产品看齐,提供有竞争力的定制硅芯片解决方案。苹果还一直使用亚马逊的Inferentia和Graviton芯片来支持搜索服务。苹果方面称,亚马逊的这些芯片带来了40%的效率提升。","content":"<html><body><p><strong><a href=\"https://laohu8.com/S/AMZN\">亚马逊</a>宣布推出了未来可能替代<a href=\"https://laohu8.com/S/NVDA\">英伟达</a>GPU的AI芯片。</strong>AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,<strong>而且<a href=\"https://laohu8.com/S/AAPL\">苹果</a>宣布会使用它来训练和部署Apple Intelligence。</strong></p><p>正在拉斯维加斯举行的re:Invent大会上,AWS推出的Trn2服务器(16片Trainium2)提供20.8 Pflops性能,可训练数十亿参数模型,试图媲美英伟达和AMD的GPU。Trn2 UltraServers(64片Trainium2)高峰时提供83.2 Pflops算力,完全可以用来训练和部署最大的模型,包括语言、多模态和视觉模型。</p><p>AWS还宣布了下一代AI芯片Trainium3的计划。预计将比Trainium2提升2倍的性能,改善40%的能效,3纳米制程,将于2025年底推出。</p><p>目前,亚马逊基于Trainium1和Inferentia2的实例,在生成式AI前沿模型训练或推理方面还不太具有竞争力,这是由于硬件规格较弱以及软件集成较弱所致。但随着Trainium2的发布,亚马逊已经做出了重大的调整,<strong>正在芯片、系统和软件编译器/框架层面向英伟达产品看齐,提供有竞争力的定制硅芯片解决方案。</strong></p><p><strong>苹果公司也非同寻常地对外介绍了它与云服务商之间的合作细节,并且表达出积极合作构建AI的意愿。</strong>苹果已经使用AWS服务超过十年,用于Siri、Apple Maps和Apple Music。苹果还一直使用亚马逊的Inferentia和Graviton芯片来支持搜索服务。苹果方面称,亚马逊的这些芯片带来了40%的效率提升。</p><p>苹果最近还将用Trainium2进行其自有模型的预训练。苹果对它的初步评估显示,预训练效率提升了50%。苹果在决定开发Apple Intelligence之后,马上就找到了AWS,寻求AI基础设施的支持。苹果也使用过<a href=\"https://laohu8.com/S/GOOG\">谷歌</a>云的TPU服务器。</p><p>苹果正在引领着个人AI的应用方向,即把AI模型向端侧部署,主要用本地化的计算为用户提供定制化与个人化的AI服务,注重保护用户的隐私。</p><p>所以,对于苹果来说,<strong>最重要的不是用十万张卡去训练大模型,而是用AI更好地服务其20多亿设备用户,</strong>即在iPhone、iPad、Mac等设备上,用自研芯片提供本地算力,只有那些较复杂的计算任务才上云端。苹果还需要云服务商能配合它进行隐私计算。</p><p>Apple Intelligence有自己的步调,它先推出内容提炼、起草邮件、生成表情包等最基本的功能,很快会集成OpenAI的大模型服务,明年会加强Siri功能,因为<a href=\"https://laohu8.com/S/5RE.SI\">智能</a>体技术的加持,它更像个能办事的助理,调动手机App完成用户吩咐的任务。</p><p><strong>而且AWS正在与Anthropic合作,打造40万Trainium2卡级的算力集群,用来训练下一代的Claude大模型。</strong>这个项目名称为Project Rainer,将为Anthropic提供的算力5倍于训练现有模型的Eflops。亚马逊对Anthropic最新的40亿美元投资,实际上将用于这个40万卡集群,目前还没有其他主要客户。</p><p>马斯克的xAI已经建好了10万H100算力集群,而且放也豪言要再买30万张B200;扎克伯格正在用一个超过10万H100的集群加班加点地训练Llama4,更不用说微软/OpenAI等,10万H100已经成为参与军备竞赛的起步价。</p><p><img src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABAQMAAAAl21bKAAAAA1BMVEXy8vJkA4prAAAACklEQVQI12NgAAAAAgAB4iG8MwAAAABJRU5ErkJggg==\"/></p><p>但Trainium算力集群真的要挑战英伟达GPU,还要付出更多努力。据半导体咨询机构semianalysis分析,<strong>40万颗Trainium2的原始浮点运算性能仍少于10万GB200集群。</strong>这意味着由于阿姆达尔定律的限制,Anthropic仍将很难与竞争对手10万卡GB200集群匹敌。在40万颗Trainium2和EFA上执行集合通信将非常困难,因此Anthropic需要在异步训练方面进行一些重大创新。</p><p>*作者注:EFA代表Elastic Fabric Adapter,是AWS提供的一种高性能网络接口技术,主要用于支持高性能计算(HPC)和机器学习工作负载。</p><p>三大云巨头AWS、<a href=\"https://laohu8.com/S/MSFT\">微软</a>Azure和谷歌云,目前数据中心芯片主要来自英伟达、AMD和<a href=\"https://laohu8.com/S/INTC\">英特尔</a>。同时,它们也在积极探索自己研制芯片,会带来成本及定制化服务等方面的好处,既用于通用计算负载,也用于加速计算,如大模型的训练和推理。AWS称,通过Trainium,Anthropic的大模型Claude Haiku 3.5,速度比其他芯片提升了60%。</p><p>随着生成式AI日益进入大规模应用阶段,企业将会寻找更加适合具体应用、为客户定制化、价格更亲民、更具能效的芯片和算力解决方案。</p><p>2025年我们会看到一个趋势,<strong>更多的算力会部署到推理侧进行强化学习,以及AI的大规模应用,这些都对芯片、服务器、工具、架构、服务等提出新的定制化要求,</strong>从而为云服务商的硅技术和初创芯片企业带来新的机会。</p></body></html>","source":"ifeng_tech","collect":0,"html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>苹果采用亚马逊芯片,一个去英伟达化的信号?</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\n苹果采用亚马逊芯片,一个去英伟达化的信号?\n</h2>\n\n<h4 class=\"meta\">\n\n\n2024-12-04 21:52 北京时间 <a href=https://tech.ifeng.com/c/8f3026eRYWB><strong>茶饮消息</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>亚马逊宣布推出了未来可能替代英伟达GPU的AI芯片。AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,而且苹果宣布会使用它来训练和部署Apple Intelligence。正在拉斯维加斯举行的re:Invent大会上,AWS推出的Trn2服务器(16片Trainium2)提供20.8 Pflops性能,可训练数十亿参数模型,试图媲美英伟达和AMD的GPU。...</p>\n\n<a href=\"https://tech.ifeng.com/c/8f3026eRYWB\">Web Link</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","type":0,"thumbnail":"","relate_stocks":{"LU0203202063.USD":"AB SICAV I - ALL MARKET INCOME PORTFOLIO \"A2X\" (USD) ACC","3NVD.UK":"LS 3X NVIDIA","SG9999015986.USD":"LIONGLOBAL DISRUPTIVE INNOVATION \"I\" (USD) ACC","SNVD.UK":"LS -1X NVIDIA","AAPL":"苹果","2NVD.UK":"2X NVIDIA ETP","LU2023250330.USD":"ALLIANZ INCOME AND GROWTH \"AMG\" (USD) INC","LU0215105999.USD":"SCHRODER ISF GLOBAL EQUITY \"A\" ACC","NVDS":"Tradr 1.5X Short NVDA Daily ETF","LU0784385170.HKD":"BGF GLOBAL MULTI ASSET INCOME \"A6\" (HKDHGD) INC","LU2211815571.USD":"ALLIANZ POSITIVE CHANGE \"AT\" (USD) ACC","LU1691799644.USD":"Amundi Funds Polen Capital Global Growth A2 (C) USD","LU0648001328.SGD":"Natixis Harris Associates US Equity RA SGD","SG9999015358.SGD":"United Income Focus Trust Dis SGD-H","NVDU":"Direxion Daily NVDA Bull 2X Shares","NVD3.UK":"LS 3X NVIDIA","LU0985320562.USD":"NORDEA 1 GLOBAL STARS EQUITY \"BP\" (USD) ACC","IE0034235188.USD":"PINEBRIDGE GLOBAL FOCUS EQUITY \"A\" (USD) ACC","LU0719512351.SGD":"JPMorgan Funds - US Technology A (acc) SGD","LU0256863902.USD":"ALLIANZ US EQUITY \"AT\" (USD) ACC","LU0211327993.USD":"TEMPLETON GLOBAL EQUITY INCOME \"A\" (USD) ACC","NVDX":"T-REX 2X LONG NVIDIA DAILY TARGET ETF","LU0882574139.USD":"富达环球消费行业基金A ACC","NVDY":"YIELDMAX NVDA OPTION INCOME STRATEGY ETF","IE00BLSP4239.USD":"Legg Mason ClearBridge - Tactical Dividend Income A Mdis USD Plus","NVD2.UK":"2X NVIDIA ETP","NVDA":"英伟达","NVDS.UK":"LS -1X NVIDIA","SG9999018865.SGD":"United Global Quality Growth Fd Cl Dist SGD-H","LU2433249047.HKD":"THEMATICS META \"R/A\" (HKD) ACC","IE00BHPRN162.USD":"BNY MELLON BLOCKCHAIN INNOVATION \"B\" (USD) ACC","LU2213496289.HKD":"ALLIANZ INCOME AND GROWTH \"AT\" (HKD) ACC","LU2461242641.AUD":"WELLINGTON US QUALITY GROWTH \"A\" (AUDHDG) ACC","LU1066051811.HKD":"HSBC GIF GLOBAL EQUITY VOLATILITY FOCUSED \"AM2\" (HKD) INC","BK4566":"资本集团","LU0056508442.USD":"贝莱德世界科技基金A2","LU2265009873.SGD":"Eastspring Investments - Global Growth Equity AS SGD-H","LU1046421795.USD":"富达环球科技A-ACC","SGXZ51526630.SGD":"大华环球创新基金A Acc SGD","NVD":"GraniteShares 2x Short NVDA Daily ETF","IE00BMPRXR70.SGD":"Neuberger Berman 5G Connectivity A Acc SGD-H","LU2023250843.SGD":"Allianz Thematica Cl AT Acc H2-SGD","LU1935042215.USD":"MANULIFE GF GLOBAL MULTI-ASSET DIVERSIFIED INCOME \"AA\" (USD) INC A","NVDD":"Direxion Daily NVDA Bear 1X Shares","LU1037948541.HKD":"AB LOW VOLATILITY EQUITY PORTFOLIO \"A\" (HKD) ACC","LU2275660780.HKD":"SCHRODER ISF GLOBAL CLIMATE CHANGE EQUITY \"A\" (HKD) ACC","LU1804176565.USD":"EASTSPRING INV GLOBAL GROWTH EQUITY \"A\" (USD) ACC","LU2237957902.USD":"NIKKO AM GLOBAL EQUITY \"F\" (USD) ACC"},"source_url":"https://tech.ifeng.com/c/8f3026eRYWB","is_english":false,"share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","article_id":"2488917961","content_text":"亚马逊宣布推出了未来可能替代英伟达GPU的AI芯片。AWS的Trainium2芯片,将用来构建一个40万卡的集群,训练下一代Claude模型,而且苹果宣布会使用它来训练和部署Apple Intelligence。正在拉斯维加斯举行的re:Invent大会上,AWS推出的Trn2服务器(16片Trainium2)提供20.8 Pflops性能,可训练数十亿参数模型,试图媲美英伟达和AMD的GPU。Trn2 UltraServers(64片Trainium2)高峰时提供83.2 Pflops算力,完全可以用来训练和部署最大的模型,包括语言、多模态和视觉模型。AWS还宣布了下一代AI芯片Trainium3的计划。预计将比Trainium2提升2倍的性能,改善40%的能效,3纳米制程,将于2025年底推出。目前,亚马逊基于Trainium1和Inferentia2的实例,在生成式AI前沿模型训练或推理方面还不太具有竞争力,这是由于硬件规格较弱以及软件集成较弱所致。但随着Trainium2的发布,亚马逊已经做出了重大的调整,正在芯片、系统和软件编译器/框架层面向英伟达产品看齐,提供有竞争力的定制硅芯片解决方案。苹果公司也非同寻常地对外介绍了它与云服务商之间的合作细节,并且表达出积极合作构建AI的意愿。苹果已经使用AWS服务超过十年,用于Siri、Apple Maps和Apple Music。苹果还一直使用亚马逊的Inferentia和Graviton芯片来支持搜索服务。苹果方面称,亚马逊的这些芯片带来了40%的效率提升。苹果最近还将用Trainium2进行其自有模型的预训练。苹果对它的初步评估显示,预训练效率提升了50%。苹果在决定开发Apple Intelligence之后,马上就找到了AWS,寻求AI基础设施的支持。苹果也使用过谷歌云的TPU服务器。苹果正在引领着个人AI的应用方向,即把AI模型向端侧部署,主要用本地化的计算为用户提供定制化与个人化的AI服务,注重保护用户的隐私。所以,对于苹果来说,最重要的不是用十万张卡去训练大模型,而是用AI更好地服务其20多亿设备用户,即在iPhone、iPad、Mac等设备上,用自研芯片提供本地算力,只有那些较复杂的计算任务才上云端。苹果还需要云服务商能配合它进行隐私计算。Apple Intelligence有自己的步调,它先推出内容提炼、起草邮件、生成表情包等最基本的功能,很快会集成OpenAI的大模型服务,明年会加强Siri功能,因为智能体技术的加持,它更像个能办事的助理,调动手机App完成用户吩咐的任务。而且AWS正在与Anthropic合作,打造40万Trainium2卡级的算力集群,用来训练下一代的Claude大模型。这个项目名称为Project Rainer,将为Anthropic提供的算力5倍于训练现有模型的Eflops。亚马逊对Anthropic最新的40亿美元投资,实际上将用于这个40万卡集群,目前还没有其他主要客户。马斯克的xAI已经建好了10万H100算力集群,而且放也豪言要再买30万张B200;扎克伯格正在用一个超过10万H100的集群加班加点地训练Llama4,更不用说微软/OpenAI等,10万H100已经成为参与军备竞赛的起步价。但Trainium算力集群真的要挑战英伟达GPU,还要付出更多努力。据半导体咨询机构semianalysis分析,40万颗Trainium2的原始浮点运算性能仍少于10万GB200集群。这意味着由于阿姆达尔定律的限制,Anthropic仍将很难与竞争对手10万卡GB200集群匹敌。在40万颗Trainium2和EFA上执行集合通信将非常困难,因此Anthropic需要在异步训练方面进行一些重大创新。*作者注:EFA代表Elastic Fabric Adapter,是AWS提供的一种高性能网络接口技术,主要用于支持高性能计算(HPC)和机器学习工作负载。三大云巨头AWS、微软Azure和谷歌云,目前数据中心芯片主要来自英伟达、AMD和英特尔。同时,它们也在积极探索自己研制芯片,会带来成本及定制化服务等方面的好处,既用于通用计算负载,也用于加速计算,如大模型的训练和推理。AWS称,通过Trainium,Anthropic的大模型Claude Haiku 3.5,速度比其他芯片提升了60%。随着生成式AI日益进入大规模应用阶段,企业将会寻找更加适合具体应用、为客户定制化、价格更亲民、更具能效的芯片和算力解决方案。2025年我们会看到一个趋势,更多的算力会部署到推理侧进行强化学习,以及AI的大规模应用,这些都对芯片、服务器、工具、架构、服务等提出新的定制化要求,从而为云服务商的硅技术和初创芯片企业带来新的机会。","news_type":1},"isVote":1,"tweetType":1,"viewCount":37,"authorTweetTopStatus":1,"verified":2,"comments":[],"imageCount":0,"langContent":"CN","totalScore":0}],"lives":[]}