[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"portal-settings:figure:zh":3,"public-menus:all":45,"post:the-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system:zh":531,"related:post:the-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system:zh:1":1921},{"statusCode":4,"data":5,"message":44},200,{"tenantId":6,"lang":7,"defaultLang":8,"siteUrl":9,"contactEmail":10,"brandName":11,"logoUrl":12,"siteName":11,"siteDescription":13,"ogImage":14,"robotsIndex":15,"socialLinks":14,"reservedSlugs":14,"seoPolicy":16},"figure","zh","en","https:\u002F\u002Ffigure.rocks","info@stajic.de","Figure Rocks","\u002Ffavicon-32x32.png","",null,true,{"branding":17,"relatedContent":18,"crossDomainLinks":19},{"logoUrl":12},{"enabled":15},[20,23,26,29,32,35,38,41],{"url":21,"label":22,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002Floving.rocks","loving.rocks",{"url":24,"label":25,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002Fstajic.de","stajic.de",{"url":27,"label":28,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002Fbazify.com","bazify.com",{"url":30,"label":31,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002Fbazify.de","bazify.de",{"url":33,"label":34,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002F2mesta.com","2mesta.com",{"url":36,"label":37,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002F2mesta.de","2mesta.de",{"url":39,"label":40,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002Fbazify.at","bazify.at",{"url":42,"label":43,"isActive":15,"showInFooter":15,"includeInSameAs":15},"https:\u002F\u002Fweb-hoch3.de","web-hoch3.de","Portal settings resolved",[46],{"id":47,"name":48,"location":49,"isActive":15,"isDefault":15,"items":50},2,"Main Menu","sidebar",[51,65,189,282,358,424],{"id":52,"title":53,"url":61,"target":62,"icon":63,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":64},"1",{"de":54,"en":55,"es":56,"fr":57,"it":55,"ru":58,"sr":59,"zh":60},"Startseite","Home","Inicio","Accueil","Главная","Početna","首页","\u002F","_self","i-lucide-home",[],{"id":66,"title":67,"url":76,"target":62,"icon":77,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":78},"3",{"de":68,"en":69,"es":70,"fr":71,"it":72,"ru":73,"sr":74,"zh":75},"Spiele","Games","Juegos","Jeux","Giochi","Игры","Igre","游戏","\u002Fgames","i-lucide-gamepad-2",[79,92,107,121,135,148,162,176],{"id":80,"title":81,"url":76,"target":62,"icon":90,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":91},"3-1",{"de":82,"en":83,"es":84,"fr":85,"it":86,"ru":87,"sr":88,"zh":89},"Spiele-Hub","Games Hub","Centro de juegos","Hub de jeux","Hub Giochi","Игровой центр","Centar za igre","游戏中心","i-lucide-layout-grid",[],{"id":93,"title":94,"url":103,"target":62,"icon":104,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":106},"1-1",{"de":95,"en":96,"es":97,"fr":98,"it":99,"ru":100,"sr":101,"zh":102},"Neu & Angesagt","New & Trending","Novedades & Tendencias","Nouveautés & Tendances","Novità & Tendenze","Новинки & тренды","Novo & Popularno","新品 & 热门","\u002Fnews\u002Fgaming-news","i-lucide-sparkles","custom",[],{"id":108,"title":109,"url":118,"target":62,"icon":119,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":120},"1-2",{"de":110,"en":111,"es":112,"fr":113,"it":114,"ru":115,"sr":116,"zh":117},"Angebote","Deals","Ofertas","Offres","Offerte","Акции","Ponude","优惠","\u002Fdeals","i-lucide-badge-percent",[],{"id":122,"title":123,"url":132,"target":62,"icon":133,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":134},"1-3",{"de":124,"en":125,"es":126,"fr":127,"it":128,"ru":129,"sr":130,"zh":131},"Release-Kalender","Release Calendar","Calendario de lanzamientos","Calendrier des sorties","Calendario delle uscite","Календарь релизов","Kalendar izdanja","发布日历","\u002Freleases","i-lucide-calendar-days",[],{"id":136,"title":137,"url":145,"target":62,"icon":146,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":147},"3-2",{"de":138,"en":139,"es":140,"fr":141,"it":142,"sr":143,"zh":144},"amiibo-kompatible Spiele","amiibo-Compatible Games","Juegos compatibles con Amiibo","Jeux compatibles Amiibo","Giochi compatibili con amiibo","Amiibo kompatibilne igre","Amiibo 兼容游戏","\u002Freference\u002Fcompatibility","i-lucide-check-circle-2",[],{"id":149,"title":150,"url":159,"target":62,"icon":160,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":161},"3-3",{"de":151,"en":152,"es":153,"fr":154,"it":155,"ru":156,"sr":157,"zh":158},"Wie amiibo funktionieren","How amiibo work","Cómo funcionan los Amiibo","Comment fonctionnent les Amiibo","Come funzionano gli Amiibo","Как работают Amiibo","Kako Amiibo rade","Amiibo 如何运作","\u002Fgames\u002Fhow-amiibo-unlocks-work","i-lucide-puzzle",[],{"id":163,"title":164,"url":173,"target":62,"icon":174,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":175},"3-4",{"de":165,"en":166,"es":167,"fr":168,"it":169,"ru":170,"sr":171,"zh":172},"Beste Amiibo pro Spiel","Best amiibo per Game","Mejores Amiibo por juego","Meilleurs Amiibo par jeu","Migliori Amiibo per gioco","Лучшие Amiibo по играм","Najbolji Amiibo po igri","各游戏最佳 Amiibo","\u002Fgames\u002Fbest-amiibo-per-game","i-lucide-award",[],{"id":177,"title":178,"url":186,"target":62,"icon":187,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":188},"3-5",{"de":179,"en":180,"es":181,"fr":182,"it":183,"ru":184,"sr":185},"Freischaltbares & Belohnungen","Unlockables & Rewards","Desbloqueables & Recompensas","Déblocables & Récompenses","Sbloccabili & Ricompense","Разблокировки & награды","Otključavanja & nagrade","\u002Fgames\u002Funlocks-and-benefits","i-lucide-gift",[],{"id":190,"title":191,"url":199,"target":62,"icon":200,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":201},"4",{"de":192,"en":193,"es":194,"it":195,"ru":196,"sr":197,"zh":198},"Sammeln","Collecting","Coleccionismo","Collezionismo","Коллекционирование","Kolekcionarstvo","收藏","\u002Fcollecting","i-lucide-gem",[202,213,227,240,254,268],{"id":203,"title":204,"url":211,"target":62,"icon":90,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":212},"4-1",{"de":205,"en":206,"fr":207,"it":208,"ru":209,"zh":210},"Sammel-Hub","Collecting Hub","Hub de collecte","Hub di raccolta","Центр сбора","收藏中心","\u002Famiibo\u002Fcollecting",[],{"id":214,"title":215,"url":224,"target":62,"icon":225,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":226},"4-2",{"de":216,"en":217,"es":218,"fr":219,"it":220,"ru":221,"sr":222,"zh":223},"Kaufberatung","Buying Guide","Guía de compra","Guide d'achat","Guida all'acquisto","Гид покупателя","Vodič za kupovinu","购买指南","\u002Famiibo-library\u002Famiibo-buying-smart","i-lucide-shopping-bag",[],{"id":228,"title":229,"url":237,"target":62,"icon":238,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":239},"4-3",{"de":230,"en":230,"es":231,"fr":232,"it":233,"ru":234,"sr":235,"zh":236},"Fake vs Real","Falso vs. Real","Faux vs Vrai","Falso vs Vero","Фейк vs Оригинал","Lažno vs Pravo","真假对比","\u002Famiibo-library\u002Famiibo-buying-smart\u002Favoid-fake-listings","i-lucide-scan",[],{"id":241,"title":242,"url":251,"target":62,"icon":252,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":253},"4-4",{"de":243,"en":244,"es":245,"fr":246,"it":247,"ru":248,"sr":249,"zh":250},"Aufbewahrung & Präsentation","Storage & Display","Almacenamiento & exhibición","Rangement & Présentation","Contenitori & Esposizione","Хранение & витрины","Odlaganje & izlaganje","存储 & 展示","\u002Famiibo-library\u002Famiibo-care-and-storage","i-lucide-box",[],{"id":255,"title":256,"url":265,"target":62,"icon":266,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":267},"4-5",{"de":257,"en":258,"es":259,"fr":260,"it":261,"ru":262,"sr":263,"zh":264},"Preisübersicht","Price Guide","Guía de precios","Guide des prix","Guida ai prezzi","Гид по ценам","Cenovnik","价格指南","\u002Fcollections\u002Fprice-guide","i-lucide-bar-chart-3",[],{"id":269,"title":270,"url":279,"target":62,"icon":280,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":281},"4-6",{"de":271,"en":272,"es":273,"fr":274,"it":275,"ru":276,"sr":277,"zh":278},"Raritäten","Rare Finds","Hallazgos únicos","Trouvailles rares","Pezzi rari","Редкие находки","Retki nalazi","稀有发现","\u002Fcollections\u002Frare-and-notable","i-lucide-trophy",[],{"id":283,"title":284,"url":292,"target":62,"icon":293,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":294},"5",{"de":285,"en":286,"es":287,"fr":286,"it":288,"ru":289,"sr":290,"zh":291},"Technik","Tech","Tecnología","Tecnologia","Технологии","Tehnologija","科技","\u002Ftech","i-lucide-cpu",[295,306,318,331,344],{"id":296,"title":297,"url":292,"target":62,"icon":90,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":305},"5-1",{"de":298,"en":298,"es":299,"fr":300,"it":301,"ru":302,"sr":303,"zh":304},"Tech Hub","Centro tecnológico","Hub Tech","Hub Tecnologico","Технохаб","Tehnološki centar","科技中心",[],{"id":307,"title":308,"url":315,"target":62,"icon":316,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":317},"5-2",{"de":309,"en":310,"fr":311,"it":312,"ru":313,"zh":314},"Audio- & Mikrofonqualität","Audio & Mic Quality","Qualité audio & micro","Qualità audio & mic","Качество звука & микрофона","音频 & 麦克风质量","\u002Fgear\u002Faudio","i-lucide-mic",[],{"id":319,"title":320,"url":328,"target":62,"icon":329,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":330},"5-3",{"de":321,"en":322,"es":323,"fr":324,"it":325,"sr":326,"zh":327},"Controller & Zubehör","Controllers & Accessories","Mandos & accesorios","Manettes & accessoires","Controller & Accessori","Kontroleri & dodatna oprema","控制器 & 配件","\u002Fgear\u002Fcontrols","i-lucide-joystick",[],{"id":332,"title":333,"url":341,"target":62,"icon":342,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":343},"5-4",{"de":334,"en":335,"es":336,"fr":337,"it":338,"ru":339,"sr":340},"Displays & Aufnahme","Displays & Capture","Pantallas & Captura","Écrans & Capture","Display & Acquisizione","Дисплеи & Захват","Ekrani & snimanje","\u002Fgear\u002Fdisplays","i-lucide-monitor",[],{"id":345,"title":346,"url":355,"target":62,"icon":356,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":357},"5-5",{"de":347,"en":348,"es":349,"fr":350,"it":351,"ru":352,"sr":353,"zh":354},"Netzwerkstabilität","Network Stability","Estabilidad de la red","Stabilité du réseau","Stabilità della rete","Стабильность сети","Stabilnost mreže","网络稳定性","\u002Fplaybooks\u002Fnetwork-stability","i-lucide-wifi",[],{"id":359,"title":360,"url":368,"target":62,"icon":369,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":370},"6",{"de":361,"en":361,"es":362,"fr":363,"it":364,"ru":365,"sr":366,"zh":367},"Shop","Tienda","Boutique","Negozio","Магазин","Prodavnica","商店","\u002Fshop","i-lucide-shopping-cart",[371,383,391,399,412],{"id":372,"title":373,"url":368,"target":62,"icon":381,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":382},"6-1",{"de":374,"en":375,"es":376,"fr":377,"ru":378,"sr":379,"zh":380},"Shop-Hub","Shop Hub","Centro de compras","Espace Boutique","Центр магазина","Centar za kupovinu","购物中心","i-lucide-store",[],{"id":384,"title":385,"url":388,"target":62,"icon":389,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":390},"6-2",{"de":386,"en":386,"es":387,"fr":387,"it":387,"sr":387,"zh":387},"amiibo","Amiibo","\u002Famiibo-shop","i-lucide-scan-line",[],{"id":392,"title":393,"url":396,"target":62,"icon":397,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":398},"6-3",{"de":394,"en":394,"es":394,"fr":394,"it":394,"ru":394,"sr":394,"zh":395},"LEGO","乐高","\u002Flego-shop","i-lucide-blocks",[],{"id":400,"title":401,"url":409,"target":62,"icon":410,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":411},"6-4",{"de":402,"en":403,"es":404,"fr":405,"it":406,"sr":407,"zh":408},"Figuren & Sammlerstücke","Figures & Collectibles","Figuras & Coleccionables","Figurines & Objets de collection","Figure & Collezionismo","Figure & kolekcionarstvo","手办 & 收藏品","\u002Fshop\u002Ffigures","i-lucide-package",[],{"id":413,"title":414,"url":421,"target":62,"icon":422,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":423},"6-5",{"de":415,"en":415,"fr":416,"it":417,"ru":418,"sr":419,"zh":420},"Gaming Gear","Équipement gaming","Accessori gaming","Игровое снаряжение","Gejming oprema","游戏装备","\u002Fgaming-gear-shop","i-lucide-headphones",[],{"id":425,"title":426,"url":427,"target":62,"icon":389,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":428},"2",{"de":386,"en":386,"es":386,"fr":386,"it":386,"ru":386,"sr":386,"zh":386},"\u002Famiibo",[429,440,447,461,475,489,503,517],{"id":430,"title":431,"url":427,"target":62,"icon":90,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":439},"2-1",{"de":432,"en":433,"es":434,"fr":435,"it":435,"ru":436,"sr":437,"zh":438},"amiibo-Hub","amiibo Hub","Centro amiibo","Hub amiibo","Хаб amiibo","amiibo centar","amiibo 中心",[],{"id":441,"title":442,"url":444,"target":62,"icon":445,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":446},"1772724791061",{"de":443,"en":443},"Franchise","\u002Famiibo\u002Ffranchise","i-lucide-star",[],{"id":448,"title":449,"url":458,"target":62,"icon":459,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":460},"2-3",{"de":450,"en":451,"es":452,"fr":453,"it":454,"ru":455,"sr":456,"zh":457},"Region & Verpackung (EU\u002FUS\u002FJP)","Region & Packaging (EU\u002FUS\u002FJP)","Región y embalaje (UE\u002FEE. UU.\u002FJP)","Région & Emballage (UE\u002FUS\u002FJP)","Regione & Confezione (EU\u002FUS\u002FJP)","Регион & упаковка (EU\u002FUS\u002FJP)","Region & pakovanje (EU\u002FUS\u002FJP)","地区 & 包装 (EU\u002FUS\u002FJP)","\u002Famiibo-library\u002Famiibo-editions-and-regions","i-lucide-globe",[],{"id":462,"title":463,"url":472,"target":62,"icon":473,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":474},"2-4",{"de":464,"en":465,"es":466,"fr":467,"it":468,"ru":469,"sr":470,"zh":471},"Schnellidentifikations-Checkliste","Fast Identification Checklist","Lista de identificación rápida","Liste d'identification rapide","Checklist identificazione rapida","Чек-лист быстрой идентификации","Kontrolna lista za brzu identifikaciju","快速识别清单","\u002Famiibo\u002Fidentify-checklist","i-lucide-list-checks",[],{"id":476,"title":477,"url":486,"target":62,"icon":487,"isActive":15,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":488},"2-6",{"de":478,"en":479,"es":480,"fr":481,"it":482,"ru":483,"sr":484,"zh":485},"Tipps für Sealed-Sammlungen","Sealed Collection Tips","Consejos de colección sellada","Conseils collection scellée","Consigli collezione sigillata","Советы по коллекции Sealed","Saveti za zapečaćene kolekcije","密封收藏贴士","\u002Famiibo\u002Fsealed-tips","i-lucide-shield-check",[],{"id":490,"title":491,"url":500,"target":62,"icon":501,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":502},"2-5",{"de":492,"en":493,"es":494,"fr":495,"it":496,"ru":497,"sr":498,"zh":499},"Zustand & Bewertung","Condition & Grading","Estado & Graduación","État & Évaluation","Condizioni & Valutazione","Состояние & Оценка","Stanje & ocenjivanje","品相与分级","\u002Famiibo-library\u002Famiibo-condition-and-grading","i-lucide-badge-check",[],{"id":504,"title":505,"url":514,"target":62,"icon":515,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":516},"2-2",{"de":506,"en":507,"es":508,"fr":509,"it":510,"ru":511,"sr":512,"zh":513},"Ausgaben & Nachdrucke","Editions & Reprints","Ediciones & Reimpresiones","Éditions & Réimpressions","Edizioni & ristampe","Издания & переиздания","Izdanja & reizdanja","版本 & 重印","\u002Famiibo\u002Fcollecting\u002F","i-lucide-layers",[],{"id":518,"title":519,"url":528,"target":62,"icon":529,"isActive":15,"type":105,"productId":14,"categoryId":14,"shopCategoryId":14,"articleId":14,"pageId":14,"portfolioId":14,"children":530},"2-7",{"de":520,"en":521,"es":522,"fr":523,"it":524,"ru":525,"sr":526,"zh":527},"Nachdruck-Timeline","Reprint Timeline","Cronología de reimpresión","Chronologie des réimpressions","Cronologia ristampe","Хронология переизданий","Hronologija reprinta","重印时间轴","\u002Freference\u002Ftimelines\u002Famiibo-reprint-timeline","i-lucide-timer",[],{"statusCode":4,"data":532,"message":1920},{"id":533,"title":534,"slug":535,"content":536,"contentJson":537,"excerpt":1128,"featuredImage":1129,"featuredImageAlt":1130,"featuredImageCaption":14,"featuredImageTitle":14,"featuredImageCopyright":14,"featuredImageAuthor":14,"featuredImageSourceUrl":14,"featuredImageLicense":14,"featuredImageIsAiGenerated":729,"status":1131,"publishedAt":1132,"createdAt":1133,"updatedAt":1134,"seoLocalePaths":1135,"categories":1144,"author":1160,"translations":1163},"458","游戏引擎正逐渐演变为部分渲染器、部分调度器、部分AI系统","the-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u003Cp>现代游戏引擎已经在帧截止时间内协调模拟、动画、物理、音频、网络、资产流送和渲染。神经渲染和游戏AI并不取代这些职责。它们增加了一类新的工作，与原有工作争夺相同的时间、内存和硬件资源。这使引擎更像一个面向传统计算和学习型计算的实时资源调度器。\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--info my-6 rounded-xl border p-5 border-blue-300 bg-blue-50 dark:border-blue-900 dark:bg-blue-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">直接回答\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">\u003Cstrong>引擎正在成为渲染、数据移动和推理协同协商的场所。\u003C\u002Fstrong>DirectX正将机器学习原语引入图形栈，DLSS 5展示了生成式渲染阶段，本地游戏助手消耗GPU资源，学习型预取表明内存移动本身可以被预测。技术挑战不再仅仅是“在游戏中运行AI”；而是决定哪些AI工作属于何处，同时不破坏帧时间、延迟或游戏状态正确性。\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Caside class=\"editorjs-callout editorjs-callout--note my-6 rounded-xl border p-5 border-gray-300 bg-gray-50 dark:border-gray-700 dark:bg-gray-900\u002F40\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">范围\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">本文描述的方向由当前API、研究和已发布\u002F已宣布的系统综合而成。最终集成的引擎架构是Figure Rocks的合成，并非声称Unreal、Unity或其他引擎已实现整个技术栈。\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Cnav class=\"editorjs-toc\" data-editorjs-toc=\"true\" aria-label=\"目录\">\u003Cstrong class=\"editorjs-toc__title\">目录\u003C\u002Fstrong>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-0\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-5\" class=\"editorjs-toc__link\">旧的心智模型本就不完整\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-8\" class=\"editorjs-toc__link\">神经工作正分裂为多个时间尺度\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-10\" class=\"editorjs-toc__link\">DirectX正使机器学习成为图形基础设施的一部分\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-13\" class=\"editorjs-toc__link\">DLSS 5将生成更深入地融入帧叙事\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-16\" class=\"editorjs-toc__link\">引擎可能同时优化五个预算\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-19\" class=\"editorjs-toc__link\">内存调度成为图形调度的一部分\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-23\" class=\"editorjs-toc__link\">预测性内存是引擎问题，因为引擎拥有上下文\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-26\" class=\"editorjs-toc__link\">游戏智能体创造了另一种引擎问题\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-30\" class=\"editorjs-toc__link\">调度器成为真正的产品\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-33\" class=\"editorjs-toc__link\">这就是stajic.de和Figure Rocks应该轻触的地方\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-36\" class=\"editorjs-toc__link\">神经游戏引擎必须执行的六个边界\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-38\" class=\"editorjs-toc__link\">架构可能的样子\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-41\" class=\"editorjs-toc__link\">什么会改变这个答案？\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-44\" class=\"editorjs-toc__link\">主要来源\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fnav>\n\u003Ch2 id=\"section-5\">旧的心智模型本就不完整\u003C\u002Fh2>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">经典简化引擎循环\u003C\u002Fh3>\u003Cdiv class=\"grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4\">\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">输入\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">读取玩家\u002F控制器\u002F网络输入。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">模拟\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">推进游戏逻辑、物理、AI和动画状态。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">准备渲染工作\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">构建可见性、绘制\u002F调度列表、资源和管线。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">渲染\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">光栅化、光线追踪和计算生成帧。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">5\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">呈现\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">帧到达显示器，同时下一帧开始。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Cp>即使这种简化也隐藏了异步任务、多线程和流送。但它有用，因为神经工作负载现在可以在多个不同点进入，而不是存在于一个孤立的“AI”框中。\u003C\u002Fp>\n\u003Ch2 id=\"section-8\">神经工作正分裂为多个时间尺度\u003C\u002Fh2>\n\u003Csection class=\"editorjs-comparison my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">一个引擎，多个截止时间\u003C\u002Fh3>\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left dark:border-gray-700 dark:bg-gray-900\">\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">典型截止时间\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">如果调度不当\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">主要压力\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">输入\u002F模拟响应\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">毫秒\u002F游戏刻\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">延迟或模拟不稳定\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">CPU\u002F游戏状态调度\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">光栅化\u002F光线追踪\u002F着色\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">当前帧预算\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">帧时间尖峰或未达目标帧率\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">GPU计算+内存带宽\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">资产\u002F内存预取\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">未来需求之前\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">延迟则停顿；过早则浪费带宽\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">存储、互连、缓存和驻留\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">帧关键神经图形\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">帧生成内部或相邻\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">伪影、延迟或未达帧预算\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">GPU矩阵\u002FTensor计算+带宽\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">智能体\u002FNPC推理\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">通常数十\u002F数百毫秒或异步\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">仓促则规划差；放入反射循环则游戏滞后\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">模型内存+CPU\u002FGPU\u002FNPU推理时间\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-10\">DirectX正使机器学习成为图形基础设施的一部分\u003C\u002Fh2>\n\u003Cp>微软2026年的DirectX工作很重要，因为它将机器学习数学视为图形平台关注的问题。协作向量和DX线性代数让较小的神经工作负载靠近着色器执行，而宣布的计算图编译器则针对更大的模型图。这为引擎创建了一条API级路径，将机器学习视为可调度的GPU工作，而非外部黑盒。\u003C\u002Fp>\n\u003Caside class=\"editorjs-referral my-6\">\u003Ca href=\"https:\u002F\u002Ffigure.rocks\u002Fzh\u002Fblog\u002Fdirectx-is-becoming-an-ml-platform-what-linear-algebra-and-neural-shaders-mean-for-future-games\" class=\"flex flex-col sm:flex-row gap-4 rounded-xl border border-gray-200 dark:border-gray-700 p-4 transition hover:border-primary-500\">\u003Cdiv class=\"min-w-0 flex-1\">\u003Cstrong class=\"block text-lg text-gray-900 dark:text-gray-100\">DirectX正在成为机器学习平台：线性代数和神经着色器对未来游戏意味着什么\u003C\u002Fstrong>\u003Cp class=\"mt-2 text-sm text-gray-600 dark:text-gray-300\">深入探讨着色器级机器学习、模型级图以及为什么通用图形API对采用至关重要。\u003C\u002Fp>\u003Cspan class=\"mt-3 inline-flex text-sm font-medium text-primary-600 dark:text-primary-400\">阅读DirectX机器学习文章 →\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fa>\u003C\u002Faside>\n\u003Ch2 id=\"section-13\">DLSS 5将生成更深入地融入帧叙事\u003C\u002Fh2>\n\u003Cp>NVIDIA将DLSS 5描述为一个实时生成式渲染阶段，通过学习到的外观先验补充传统渲染。架构含义比单个产品版本更重要：最终帧可以越来越多地包含并非仅由传统场景模拟和着色产生的信息。\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--success my-6 rounded-xl border p-5 border-emerald-300 bg-emerald-50 dark:border-emerald-900 dark:bg-emerald-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">新的调度问题\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">如果一个神经阶段可以节省传统渲染工作，但自身消耗计算、内存和时间，引擎需要一个\u003Cstrong>价值测试\u003C\u002Fstrong>：什么成本消失了，增加了什么推理成本，玩家是否在相同的延迟包络内真正获得了更好的帧？\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-16\">引擎可能同时优化五个预算\u003C\u002Fh2>\n\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">预算\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">传统问题\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">神经时代问题\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">存储\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">资产是否已安装？\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">紧凑表示能否替代部分存储细节？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">内存容量\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">工作集是否适合？\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">哪些资产、模型状态和神经表示值得驻留？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">带宽\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">所需字节能否及时到达？\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">压缩\u002F预取能否减少或隐藏传输？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">渲染计算\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">着色器\u002F光线追踪能否在帧内完成？\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">哪些细节推理比传统渲染更便宜？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">推理计算\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">主要在图形循环之外\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">哪些神经工作在本帧值得其计算和延迟成本？\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Caside class=\"editorjs-callout editorjs-callout--success my-6 rounded-xl border p-5 border-emerald-300 bg-emerald-50 dark:border-emerald-900 dark:bg-emerald-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">架构简写\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">\u003Cstrong>少存储 → 少移动 → 多预测 → 少渲染 → 多推理\u003C\u002Fstrong>只有在每个替代都比其取代的工作更便宜时才有用。引擎调度器是必须衡量这种权衡的地方。\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-19\">内存调度成为图形调度的一部分\u003C\u002Fh2>\n\u003Cp>神经图形在纹理、几何体和渲染目标之上增加了权重、潜在表示和历史缓冲区。资源流送已经必须预测玩家将看到什么。未来的引擎可以通过对未来的内存需求进行排序，并决定在帧到达之前哪些表示应该驻留，从而使这种预测更加明确。\u003C\u002Fp>\n\u003Caside class=\"editorjs-referral my-6\">\u003Ca href=\"https:\u002F\u002Ffigure.rocks\u002Fzh\u002Fblog\u002Fwhat-if-a-game-could-predict-its-next-memory-access-ai-prefetching-and-the-future-of-stutter\" class=\"flex flex-col sm:flex-row gap-4 rounded-xl border border-gray-200 dark:border-gray-700 p-4 transition hover:border-primary-500\">\u003Cdiv class=\"min-w-0 flex-1\">\u003Cstrong class=\"block text-lg text-gray-900 dark:text-gray-100\">如果游戏能预测下一次内存访问会怎样？AI预取与卡顿的未来\u003C\u002Fstrong>\u003Cp class=\"mt-2 text-sm text-gray-600 dark:text-gray-300\">Figure Rocks预测性游戏内存假说：确定性游戏状态加上学习型预取。\u003C\u002Fp>\u003Cspan class=\"mt-3 inline-flex text-sm font-medium text-primary-600 dark:text-primary-400\">阅读预测性内存分析 →\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fa>\u003C\u002Faside>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">Figure Rocks神经引擎循环\u003C\u002Fh3>\u003Cdiv class=\"grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4\">\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">1. 解析游戏状态\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">当前模拟中什么是真实的？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">2. 预测近期需求\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">接下来可能涉及哪个场景、资源、着色器和神经数据？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">3. 分配帧预算\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">多少时间用于模拟、传统图形、光线追踪和神经阶段？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">4. 分配内存\u002F带宽预算\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">哪些资源保持驻留，哪些可以压缩或预取？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">5\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">5. 执行快速神经工作\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">在图形路径内进行重建、降噪、纹理\u002F材质推断或生成。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">6\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">6. 异步执行慢速AI\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">助手\u002FNPC推理尽可能在关键渲染截止时间之外运行。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">7\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">7. 验证输出\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">检查帧节奏、延迟、图像稳定性和游戏状态正确性。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">8\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">8. 反馈测量结果\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">调度器根据未命中、停顿和质量结果调整策略。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-23\">预测性内存是引擎问题，因为引擎拥有上下文\u003C\u002Fh2>\n\u003Cp>驱动程序可以观察内存行为，但游戏引擎知道语义状态：下一个关卡单元、摄像机方向、资源依赖关系、动画图、附近角色和移动约束。这使得引擎可以将确定性知识与学习型排序相结合，然后仅在当前渲染和模拟留有足够带宽余量时调度推测性移动。\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--note my-6 rounded-xl border p-5 border-gray-300 bg-gray-50 dark:border-gray-700 dark:bg-gray-900\u002F40\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">研究证据支持该机制，而非游戏结果\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">在已发表的GPU基准测试中，学习型UVM预取提高了页面命中率并减少了CPU–GPU流量，而2026年的异步KV缓存研究展示了主动移动如何隐藏HBM延迟。Figure Rocks的步骤是\u003Cstrong>引擎语义扩展\u003C\u002Fstrong>：利用对游戏世界的知识来改进预测内容。\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-26\">游戏智能体创造了另一种引擎问题\u003C\u002Fh2>\n\u003Cp>AI队友或助手不仅仅是另一个着色器。它对游戏知识和实时状态进行推理，可能调用工具，并可能产生影响世界的动作。Figure Rocks已经在PUBG Ally架构中将快速反应式游戏控制与较慢的语言推理分开。同样的分离应属于引擎层面：并非每个智能操作都属于同一个循环或应获得相同的权限。\u003C\u002Fp>\n\u003Caside class=\"editorjs-referral my-6\">\u003Ca href=\"https:\u002F\u002Ffigure.rocks\u002Fzh\u002Fblog\u002Fpubg-ally-shows-why-ai-teammates-need-two-brains-fast-reflexes-and-slow-reasoning\" class=\"flex flex-col sm:flex-row gap-4 rounded-xl border border-gray-200 dark:border-gray-700 p-4 transition hover:border-primary-500\">\u003Cdiv class=\"min-w-0 flex-1\">\u003Cstrong class=\"block text-lg text-gray-900 dark:text-gray-100\">PUBG Ally展示了为什么AI队友需要两个大脑：快速反射与慢速推理\u003C\u002Fstrong>\u003Cp class=\"mt-2 text-sm text-gray-600 dark:text-gray-300\">为什么语言推理和tick级反应式控制具有不同的时序和权限要求。\u003C\u002Fp>\u003Cspan class=\"mt-3 inline-flex text-sm font-medium text-primary-600 dark:text-primary-400\">阅读双速AI架构 →\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fa>\u003C\u002Faside>\n\u003Csection class=\"editorjs-comparison my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">图形推理 vs 智能体推理\u003C\u002Fh3>\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left dark:border-gray-700 dark:bg-gray-900\">\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">图形推理\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">游戏智能体推理\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">主要输入\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">帧缓冲区、运动\u002F历史、几何\u002F材质数据\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">游戏状态、玩家请求、检索到的知识\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">主要输出\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">像素、重建特征或渲染数据\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">计划、解释、工具\u002F动作请求\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">时序\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">通常帧关键\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">通常异步或较慢\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">验证\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">图像质量、稳定性、延迟\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">状态有效性、权限、正确性、安全性\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">故障可见性\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">伪影或卡顿\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">错误建议或错误动作\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-30\">调度器成为真正的产品\u003C\u002Fh2>\n\u003Cp>随着更多专用硬件的出现，引擎的价值从仅仅暴露功能转变为正确放置工作。CPU可能空闲而GPU因数据而停顿。张量硬件可能可用而VRAM已饱和。NPU可能存在但与引擎的帧路径连接太弱。如果调度器无法为系统提供数据并安排顺序，最快的组件也无法决定体验。\u003C\u002Fp>\n\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">引擎决策\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">糟糕的简化\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">更好的问题\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">运行神经渲染\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">\"张量核心很快\"\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">推理节省的帧成本是否超过其增加的成本，包括数据移动？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">运行本地助手\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">\"模型能放进VRAM\"\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">在实际游戏过程中还剩多少空闲内存\u002F计算？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">预取资源\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">\"加载下一个区域\"\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">哪个子集最可能需要，安全的带宽预算是多少？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">使用统一内存\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">\"一切都能访问一切\"\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">在帧截止时间哪个工作负载获得带宽优先权？\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">生成帧\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">\"显示的FPS更高\"\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">基础渲染速率、延迟和时间质量如何？\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Ch2 id=\"section-33\">这就是stajic.de和Figure Rocks应该轻触的地方\u003C\u002Fh2>\n\u003Cp>通用AI架构问题涉及模型、检索、工具、协议和权限。Figure Rocks应保持应用边界：当这些组件必须与实时游戏引擎共存时会发生什么。一个工具协议在软件层面可能是正确的，但如果它引入帧卡顿或作用于过时的游戏状态，仍然不可用。\u003C\u002Fp>\n\u003Caside class=\"editorjs-referral my-6\">\u003Ca href=\"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"flex flex-col sm:flex-row gap-4 rounded-xl border border-gray-200 dark:border-gray-700 p-4 transition hover:border-primary-500\">\u003Cdiv class=\"min-w-0 flex-1\">\u003Cstrong class=\"block text-lg text-gray-900 dark:text-gray-100\">MCP vs A2A vs UCP vs AP2 vs A2UI：智能体协议栈解释\u003C\u002Fstrong>\u003Cp class=\"mt-2 text-sm text-gray-600 dark:text-gray-300\">stajic.de上的通用协议架构。Figure Rocks将同样的边界思维应用于游戏、工具和实时状态。\u003C\u002Fp>\u003Cspan class=\"mt-3 inline-flex text-sm font-medium text-primary-600 dark:text-primary-400\">阅读协议栈解释 →\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fa>\u003C\u002Faside>\n\u003Ch2 id=\"section-36\">神经游戏引擎必须执行的六个边界\u003C\u002Fh2>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">神经引擎边界测试\u003C\u002Fh3>\u003Cdiv class=\"grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4\">\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">1. 帧边界\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">哪些工作必须在呈现之前完成？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">2. 状态边界\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">哪些数据描述的是当前游戏状态而非通用知识？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">3. 内存边界\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">哪些资源必须现在驻留，哪些是为后续预测的？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">4. 权限边界\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">哪个模型可以建议操作，哪个子系统可以执行操作？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">5\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">5. 质量边界\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">对于此功能，可接受的产物、错误或幻觉水平是什么？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">6\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">6. 回退边界\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">当神经路径太慢、不可用或不确定时，哪条确定性路径接管？\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-38\">架构可能的样子\u003C\u002Fh2>\n\u003Csection class=\"editorjs-comparison my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">一种可能的未来引擎栈\u003C\u002Fh3>\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left dark:border-gray-700 dark:bg-gray-900\">\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">角色\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">职责\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">游戏状态核心\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">确定性权威\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">模拟、规则、权威状态\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">实时调度器\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">截止时间管理器\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">CPU\u002FGPU\u002FRT\u002FML 时序与预算\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">内存智能\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">驻留\u002F预取管理器\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">压缩、分层、预测与传输\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">混合渲染器\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">视觉生产者\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">传统 + 神经图形阶段\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">代理层\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">审慎智能\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">在权限限制下的推理、对话、规划与工具\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">验证层\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">护栏\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">帧质量、状态有效性、延迟与回退\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Caside class=\"editorjs-callout editorjs-callout--note my-6 rounded-xl border p-5 border-gray-300 bg-gray-50 dark:border-gray-700 dark:bg-gray-900\u002F40\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Figure Rocks 综合\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">重要的转变不是“游戏将使用更多 AI”。而是\u003Cstrong>引擎可能成为确定性模拟与概率计算之间的仲裁层\u003C\u002Fstrong>。引擎决定学习系统何时有价值、可以消耗多少资源、可以看到什么状态，以及当它们出错时会发生什么。\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-41\">什么会改变这个答案？\u003C\u002Fh2>\n\u003Cp>如果神经图形和本地游戏代理仍然是孤立可选功能，引擎可以继续将它们视为围绕传统核心的插件。如果它们变得普遍，引擎架构将需要像现代引擎已经为 CPU、GPU 和流式工作做预算那样，为推理、内存移动和概率输出提供一等预算。\u003C\u002Fp>\n\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">观察\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">对假设的影响\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">神经图形仍限于可选后处理\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">削弱推理成为核心引擎调度领域的说法\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">内存经济正常化，消费级带宽\u002F容量再次廉价扩展\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">减少用预测\u002F压缩替代原始资源的压力\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">跨厂商 ML API 未能成熟\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">使神经技术保持碎片化且集成成本高昂\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">预测开销\u002F浪费超过隐藏延迟\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">削弱预测内存的价值\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">开发者将 AI 代理完全置于延迟关键的引擎路径之外\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">将调度器收敛限制在图形\u002F数据而非所有 AI\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Ch2 id=\"section-44\">主要来源\u003C\u002Fh2>\n\u003Ca href=\"https:\u002F\u002Fdevblogs.microsoft.com\u002Fdirectx\u002Fevolving-directx-for-the-ml-era-on-windows\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Microsoft DirectX — 为 ML 时代演进 DirectX\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">DirectX 着色器级和模型级 ML 方向的官方来源。\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fabs\u002Fpii\u002FS0743731522002490\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">CPU-GPU UVM 中基于深度学习的数据预取\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">学习预测可以改善 GPU 内存页面预取的研究证据。\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F39224\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">AAAI 2026 — 异步 KV 缓存预取\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">主动内存移动和延迟隐藏的研究证据。\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fresearch.nvidia.com\u002Flabs\u002Fadlr\u002FDLSS5\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">NVIDIA Research — DLSS 5\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">交互式图形管线中生成式神经渲染的主要研究来源。\u003C\u002Fp>\u003C\u002Fa>\n\u003Csection class=\"editorjs-faq my-6 rounded-xl border border-gray-200 p-5 dark:border-gray-700\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">未来游戏引擎架构\u003C\u002Fh3>\u003Cdiv id=\"faq1\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">AI 会取代游戏引擎吗？\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">不会。引擎仍然是拥有模拟、状态、调度和集成的系统。AI 成为该架构内的另一类工作负载。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq2\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">为什么不异步运行每个模型？\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">一些图形 ML 必须影响当前帧。其他 AI 可以异步，但仍需要当前状态和资源限制。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq3\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">最大的技术挑战是什么？\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">协调截止时间、内存和权限，使概率系统改善体验而不破坏帧时间或游戏状态的稳定性。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq4\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">统一内存能解决调度问题吗？\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">不能。它可以简化容量和共享，但也可能增加争用。调度和局部性仍然重要。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq5\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">Figure Rocks 的核心观点是什么？\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">未来引擎可能在确定性模拟和概率计算之间进行仲裁，而不是将 AI 视为一个单独的功能。\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Csection class=\"editorjs-glossary my-6 rounded-xl border border-gray-200 dark:border-gray-700 p-5\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">神经引擎术语\u003C\u002Fh3>\u003Cdl>\u003Cdiv id=\"frame-critical\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">帧关键推理\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">必须在渲染截止时间内或附近完成的 ML 工作。\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"deliberative\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">审慎推理\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">较慢的推理或规划，通常可以在即时帧截止时间之外运行。\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"authority-boundary\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">权限边界\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">定义模型可以推荐什么与确定性工具或游戏系统实际可以改变什么的规则。\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"neural-engine-loop\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">神经引擎循环\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Figure Rocks 综合，用于在实时预算下协调状态、预测、内存、渲染、代理和验证。\u003C\u002Fdd>\u003C\u002Fdiv>\u003C\u002Fdl>\u003C\u002Fsection>",{"time":538,"blocks":539,"version":1127},1792340796707,[540,546,554,561,568,574,597,602,607,660,665,670,679,684,689,696,701,731,737,742,747,755,785,790,795,801,806,811,819,858,863,868,897,902,907,915,920,944,949,996,1002,1007,1012,1035,1040,1050,1059,1068,1077,1104],{"id":541,"data":542,"type":544,"tunes":545},"p-0173",{"text":543},"现代游戏引擎已经在帧截止时间内协调模拟、动画、物理、音频、网络、资产流送和渲染。神经渲染和游戏AI并不取代这些职责。它们增加了一类新的工作，与原有工作争夺相同的时间、内存和硬件资源。这使引擎更像一个面向传统计算和学习型计算的实时资源调度器。","paragraph",{},{"id":547,"data":548,"type":552,"tunes":553},"c-0174",{"body":549,"title":550,"variant":551},"\u003Cstrong>引擎正在成为渲染、数据移动和推理协同协商的场所。\u003C\u002Fstrong>DirectX正将机器学习原语引入图形栈，DLSS 5展示了生成式渲染阶段，本地游戏助手消耗GPU资源，学习型预取表明内存移动本身可以被预测。技术挑战不再仅仅是“在游戏中运行AI”；而是决定哪些AI工作属于何处，同时不破坏帧时间、延迟或游戏状态正确性。","直接回答","info","callout",{},{"id":555,"data":556,"type":552,"tunes":560},"c-0175",{"body":557,"title":558,"variant":559},"本文描述的方向由当前API、研究和已发布\u002F已宣布的系统综合而成。最终集成的引擎架构是Figure Rocks的合成，并非声称Unreal、Unity或其他引擎已实现整个技术栈。","范围","note",{},{"id":562,"data":563,"type":566,"tunes":567},"toc-0176",{"title":564,"maxLevel":565,"minLevel":47},"目录",3,"tableOfContents",{},{"id":569,"data":570,"type":572,"tunes":573},"h-0177",{"text":571,"level":47},"旧的心智模型本就不完整","header",{},{"id":575,"data":576,"type":595,"tunes":596},"flow-0178",{"steps":577,"title":593,"orientation":594},[578,581,584,587,590],{"label":579,"description":580},"输入","读取玩家\u002F控制器\u002F网络输入。",{"label":582,"description":583},"模拟","推进游戏逻辑、物理、AI和动画状态。",{"label":585,"description":586},"准备渲染工作","构建可见性、绘制\u002F调度列表、资源和管线。",{"label":588,"description":589},"渲染","光栅化、光线追踪和计算生成帧。",{"label":591,"description":592},"呈现","帧到达显示器，同时下一帧开始。","经典简化引擎循环","auto","processFlow",{},{"id":598,"data":599,"type":544,"tunes":601},"p-0179",{"text":600},"即使这种简化也隐藏了异步任务、多线程和流送。但它有用，因为神经工作负载现在可以在多个不同点进入，而不是存在于一个孤立的“AI”框中。",{},{"id":603,"data":604,"type":572,"tunes":606},"h-0180",{"text":605,"level":47},"神经工作正分裂为多个时间尺度",{},{"id":608,"data":609,"type":658,"tunes":659},"cmp-0181",{"rows":610,"title":646,"layout":647,"columns":648},[611,618,625,632,639],{"id":612,"label":613,"values":614},"input","输入\u002F模拟响应",{"failure":615,"deadline":616,"resource":617},"延迟或模拟不稳定","毫秒\u002F游戏刻","CPU\u002F游戏状态调度",{"id":619,"label":620,"values":621},"render","光栅化\u002F光线追踪\u002F着色",{"failure":622,"deadline":623,"resource":624},"帧时间尖峰或未达目标帧率","当前帧预算","GPU计算+内存带宽",{"id":626,"label":627,"values":628},"prefetch","资产\u002F内存预取",{"failure":629,"deadline":630,"resource":631},"延迟则停顿；过早则浪费带宽","未来需求之前","存储、互连、缓存和驻留",{"id":633,"label":634,"values":635},"neuralgfx","帧关键神经图形",{"failure":636,"deadline":637,"resource":638},"伪影、延迟或未达帧预算","帧生成内部或相邻","GPU矩阵\u002FTensor计算+带宽",{"id":640,"label":641,"values":642},"agent","智能体\u002FNPC推理",{"failure":643,"deadline":644,"resource":645},"仓促则规划差；放入反射循环则游戏滞后","通常数十\u002F数百毫秒或异步","模型内存+CPU\u002FGPU\u002FNPU推理时间","一个引擎，多个截止时间","table",[649,652,655],{"id":650,"label":651},"deadline","典型截止时间",{"id":653,"label":654},"failure","如果调度不当",{"id":656,"label":657},"resource","主要压力","comparison",{},{"id":661,"data":662,"type":572,"tunes":664},"h-0182",{"text":663,"level":47},"DirectX正使机器学习成为图形基础设施的一部分",{},{"id":666,"data":667,"type":544,"tunes":669},"p-0183",{"text":668},"微软2026年的DirectX工作很重要，因为它将机器学习数学视为图形平台关注的问题。协作向量和DX线性代数让较小的神经工作负载靠近着色器执行，而宣布的计算图编译器则针对更大的模型图。这为引擎创建了一条API级路径，将机器学习视为可调度的GPU工作，而非外部黑盒。",{},{"id":671,"data":672,"type":677,"tunes":678},"ref-0184",{"url":673,"title":674,"excerpt":675,"ctaLabel":676},"https:\u002F\u002Ffigure.rocks\u002Fzh\u002Fblog\u002Fdirectx-is-becoming-an-ml-platform-what-linear-algebra-and-neural-shaders-mean-for-future-games","DirectX正在成为机器学习平台：线性代数和神经着色器对未来游戏意味着什么","深入探讨着色器级机器学习、模型级图以及为什么通用图形API对采用至关重要。","阅读DirectX机器学习文章","referralArticle",{},{"id":680,"data":681,"type":572,"tunes":683},"h-0185",{"text":682,"level":47},"DLSS 5将生成更深入地融入帧叙事",{},{"id":685,"data":686,"type":544,"tunes":688},"p-0186",{"text":687},"NVIDIA将DLSS 5描述为一个实时生成式渲染阶段，通过学习到的外观先验补充传统渲染。架构含义比单个产品版本更重要：最终帧可以越来越多地包含并非仅由传统场景模拟和着色产生的信息。",{},{"id":690,"data":691,"type":552,"tunes":695},"c-0187",{"body":692,"title":693,"variant":694},"如果一个神经阶段可以节省传统渲染工作，但自身消耗计算、内存和时间，引擎需要一个\u003Cstrong>价值测试\u003C\u002Fstrong>：什么成本消失了，增加了什么推理成本，玩家是否在相同的延迟包络内真正获得了更好的帧？","新的调度问题","success",{},{"id":697,"data":698,"type":572,"tunes":700},"v3-store-move-render",{"text":699,"level":47},"引擎可能同时优化五个预算",{},{"id":702,"data":703,"type":647,"tunes":730},"v3-five-budgets",{"content":704,"stretched":729,"withHeadings":15},[705,709,713,717,721,725],[706,707,708],"预算","传统问题","神经时代问题",[710,711,712],"存储","资产是否已安装？","紧凑表示能否替代部分存储细节？",[714,715,716],"内存容量","工作集是否适合？","哪些资产、模型状态和神经表示值得驻留？",[718,719,720],"带宽","所需字节能否及时到达？","压缩\u002F预取能否减少或隐藏传输？",[722,723,724],"渲染计算","着色器\u002F光线追踪能否在帧内完成？","哪些细节推理比传统渲染更便宜？",[726,727,728],"推理计算","主要在图形循环之外","哪些神经工作在本帧值得其计算和延迟成本？",false,{},{"id":732,"data":733,"type":552,"tunes":736},"v3-five-mantra",{"body":734,"title":735,"variant":694},"\u003Cstrong>少存储 → 少移动 → 多预测 → 少渲染 → 多推理\u003C\u002Fstrong>只有在每个替代都比其取代的工作更便宜时才有用。引擎调度器是必须衡量这种权衡的地方。","架构简写",{},{"id":738,"data":739,"type":572,"tunes":741},"h-0188",{"text":740,"level":47},"内存调度成为图形调度的一部分",{},{"id":743,"data":744,"type":544,"tunes":746},"p-0189",{"text":745},"神经图形在纹理、几何体和渲染目标之上增加了权重、潜在表示和历史缓冲区。资源流送已经必须预测玩家将看到什么。未来的引擎可以通过对未来的内存需求进行排序，并决定在帧到达之前哪些表示应该驻留，从而使这种预测更加明确。",{},{"id":748,"data":749,"type":677,"tunes":754},"ref-0190",{"url":750,"title":751,"excerpt":752,"ctaLabel":753},"https:\u002F\u002Ffigure.rocks\u002Fzh\u002Fblog\u002Fwhat-if-a-game-could-predict-its-next-memory-access-ai-prefetching-and-the-future-of-stutter","如果游戏能预测下一次内存访问会怎样？AI预取与卡顿的未来","Figure Rocks预测性游戏内存假说：确定性游戏状态加上学习型预取。","阅读预测性内存分析",{},{"id":756,"data":757,"type":595,"tunes":784},"flow-0191",{"steps":758,"title":783,"orientation":594},[759,762,765,768,771,774,777,780],{"label":760,"description":761},"1. 解析游戏状态","当前模拟中什么是真实的？",{"label":763,"description":764},"2. 预测近期需求","接下来可能涉及哪个场景、资源、着色器和神经数据？",{"label":766,"description":767},"3. 分配帧预算","多少时间用于模拟、传统图形、光线追踪和神经阶段？",{"label":769,"description":770},"4. 分配内存\u002F带宽预算","哪些资源保持驻留，哪些可以压缩或预取？",{"label":772,"description":773},"5. 执行快速神经工作","在图形路径内进行重建、降噪、纹理\u002F材质推断或生成。",{"label":775,"description":776},"6. 异步执行慢速AI","助手\u002FNPC推理尽可能在关键渲染截止时间之外运行。",{"label":778,"description":779},"7. 验证输出","检查帧节奏、延迟、图像稳定性和游戏状态正确性。",{"label":781,"description":782},"8. 反馈测量结果","调度器根据未命中、停顿和质量结果调整策略。","Figure Rocks神经引擎循环",{},{"id":786,"data":787,"type":572,"tunes":789},"v3-predictive-engine",{"text":788,"level":47},"预测性内存是引擎问题，因为引擎拥有上下文",{},{"id":791,"data":792,"type":544,"tunes":794},"v3-predictive-engine-p",{"text":793},"驱动程序可以观察内存行为，但游戏引擎知道语义状态：下一个关卡单元、摄像机方向、资源依赖关系、动画图、附近角色和移动约束。这使得引擎可以将确定性知识与学习型排序相结合，然后仅在当前渲染和模拟留有足够带宽余量时调度推测性移动。",{},{"id":796,"data":797,"type":552,"tunes":800},"v3-engine-research",{"body":798,"title":799,"variant":559},"在已发表的GPU基准测试中，学习型UVM预取提高了页面命中率并减少了CPU–GPU流量，而2026年的异步KV缓存研究展示了主动移动如何隐藏HBM延迟。Figure Rocks的步骤是\u003Cstrong>引擎语义扩展\u003C\u002Fstrong>：利用对游戏世界的知识来改进预测内容。","研究证据支持该机制，而非游戏结果",{},{"id":802,"data":803,"type":572,"tunes":805},"h-0192",{"text":804,"level":47},"游戏智能体创造了另一种引擎问题",{},{"id":807,"data":808,"type":544,"tunes":810},"p-0193",{"text":809},"AI队友或助手不仅仅是另一个着色器。它对游戏知识和实时状态进行推理，可能调用工具，并可能产生影响世界的动作。Figure Rocks已经在PUBG Ally架构中将快速反应式游戏控制与较慢的语言推理分开。同样的分离应属于引擎层面：并非每个智能操作都属于同一个循环或应获得相同的权限。",{},{"id":812,"data":813,"type":677,"tunes":818},"ref-0194",{"url":814,"title":815,"excerpt":816,"ctaLabel":817},"https:\u002F\u002Ffigure.rocks\u002Fzh\u002Fblog\u002Fpubg-ally-shows-why-ai-teammates-need-two-brains-fast-reflexes-and-slow-reasoning","PUBG Ally展示了为什么AI队友需要两个大脑：快速反射与慢速推理","为什么语言推理和tick级反应式控制具有不同的时序和权限要求。","阅读双速AI架构",{},{"id":820,"data":821,"type":658,"tunes":857},"cmp-0195",{"rows":822,"title":850,"layout":647,"columns":851},[823,828,834,839,845],{"id":612,"label":824,"values":825},"主要输入",{"gfx":826,"agent":827},"帧缓冲区、运动\u002F历史、几何\u002F材质数据","游戏状态、玩家请求、检索到的知识",{"id":829,"label":830,"values":831},"output","主要输出",{"gfx":832,"agent":833},"像素、重建特征或渲染数据","计划、解释、工具\u002F动作请求",{"id":650,"label":835,"values":836},"时序",{"gfx":837,"agent":838},"通常帧关键","通常异步或较慢",{"id":840,"label":841,"values":842},"validation","验证",{"gfx":843,"agent":844},"图像质量、稳定性、延迟","状态有效性、权限、正确性、安全性",{"id":653,"label":846,"values":847},"故障可见性",{"gfx":848,"agent":849},"伪影或卡顿","错误建议或错误动作","图形推理 vs 智能体推理",[852,855],{"id":853,"label":854},"gfx","图形推理",{"id":640,"label":856},"游戏智能体推理",{},{"id":859,"data":860,"type":572,"tunes":862},"h-0196",{"text":861,"level":47},"调度器成为真正的产品",{},{"id":864,"data":865,"type":544,"tunes":867},"p-0197",{"text":866},"随着更多专用硬件的出现，引擎的价值从仅仅暴露功能转变为正确放置工作。CPU可能空闲而GPU因数据而停顿。张量硬件可能可用而VRAM已饱和。NPU可能存在但与引擎的帧路径连接太弱。如果调度器无法为系统提供数据并安排顺序，最快的组件也无法决定体验。",{},{"id":869,"data":870,"type":647,"tunes":896},"t-0198",{"content":871,"stretched":729,"withHeadings":15},[872,876,880,884,888,892],[873,874,875],"引擎决策","糟糕的简化","更好的问题",[877,878,879],"运行神经渲染","\"张量核心很快\"","推理节省的帧成本是否超过其增加的成本，包括数据移动？",[881,882,883],"运行本地助手","\"模型能放进VRAM\"","在实际游戏过程中还剩多少空闲内存\u002F计算？",[885,886,887],"预取资源","\"加载下一个区域\"","哪个子集最可能需要，安全的带宽预算是多少？",[889,890,891],"使用统一内存","\"一切都能访问一切\"","在帧截止时间哪个工作负载获得带宽优先权？",[893,894,895],"生成帧","\"显示的FPS更高\"","基础渲染速率、延迟和时间质量如何？",{},{"id":898,"data":899,"type":572,"tunes":901},"h-0199",{"text":900,"level":47},"这就是stajic.de和Figure Rocks应该轻触的地方",{},{"id":903,"data":904,"type":544,"tunes":906},"p-0200",{"text":905},"通用AI架构问题涉及模型、检索、工具、协议和权限。Figure Rocks应保持应用边界：当这些组件必须与实时游戏引擎共存时会发生什么。一个工具协议在软件层面可能是正确的，但如果它引入帧卡顿或作用于过时的游戏状态，仍然不可用。",{},{"id":908,"data":909,"type":677,"tunes":914},"ref-0201",{"url":910,"title":911,"excerpt":912,"ctaLabel":913},"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","MCP vs A2A vs UCP vs AP2 vs A2UI：智能体协议栈解释","stajic.de上的通用协议架构。Figure Rocks将同样的边界思维应用于游戏、工具和实时状态。","阅读协议栈解释",{},{"id":916,"data":917,"type":572,"tunes":919},"h-0202",{"text":918,"level":47},"神经游戏引擎必须执行的六个边界",{},{"id":921,"data":922,"type":595,"tunes":943},"flow-0203",{"steps":923,"title":942,"orientation":594},[924,927,930,933,936,939],{"label":925,"description":926},"1. 帧边界","哪些工作必须在呈现之前完成？",{"label":928,"description":929},"2. 状态边界","哪些数据描述的是当前游戏状态而非通用知识？",{"label":931,"description":932},"3. 内存边界","哪些资源必须现在驻留，哪些是为后续预测的？",{"label":934,"description":935},"4. 权限边界","哪个模型可以建议操作，哪个子系统可以执行操作？",{"label":937,"description":938},"5. 质量边界","对于此功能，可接受的产物、错误或幻觉水平是什么？",{"label":940,"description":941},"6. 回退边界","当神经路径太慢、不可用或不确定时，哪条确定性路径接管？","神经引擎边界测试",{},{"id":945,"data":946,"type":572,"tunes":948},"h-0204",{"text":947,"level":47},"架构可能的样子",{},{"id":950,"data":951,"type":658,"tunes":995},"cmp-0205",{"rows":952,"title":987,"layout":647,"columns":988},[953,959,965,971,976,981],{"id":954,"label":955,"values":956},"state","游戏状态核心",{"role":957,"responsibility":958},"确定性权威","模拟、规则、权威状态",{"id":960,"label":961,"values":962},"sched","实时调度器",{"role":963,"responsibility":964},"截止时间管理器","CPU\u002FGPU\u002FRT\u002FML 时序与预算",{"id":966,"label":967,"values":968},"memory","内存智能",{"role":969,"responsibility":970},"驻留\u002F预取管理器","压缩、分层、预测与传输",{"id":619,"label":972,"values":973},"混合渲染器",{"role":974,"responsibility":975},"视觉生产者","传统 + 神经图形阶段",{"id":640,"label":977,"values":978},"代理层",{"role":979,"responsibility":980},"审慎智能","在权限限制下的推理、对话、规划与工具",{"id":982,"label":983,"values":984},"verify","验证层",{"role":985,"responsibility":986},"护栏","帧质量、状态有效性、延迟与回退","一种可能的未来引擎栈",[989,992],{"id":990,"label":991},"role","角色",{"id":993,"label":994},"responsibility","职责",{},{"id":997,"data":998,"type":552,"tunes":1001},"c-0206",{"body":999,"title":1000,"variant":559},"重要的转变不是“游戏将使用更多 AI”。而是\u003Cstrong>引擎可能成为确定性模拟与概率计算之间的仲裁层\u003C\u002Fstrong>。引擎决定学习系统何时有价值、可以消耗多少资源、可以看到什么状态，以及当它们出错时会发生什么。","Figure Rocks 综合",{},{"id":1003,"data":1004,"type":572,"tunes":1006},"h-0207",{"text":1005,"level":47},"什么会改变这个答案？",{},{"id":1008,"data":1009,"type":544,"tunes":1011},"p-0208",{"text":1010},"如果神经图形和本地游戏代理仍然是孤立可选功能，引擎可以继续将它们视为围绕传统核心的插件。如果它们变得普遍，引擎架构将需要像现代引擎已经为 CPU、GPU 和流式工作做预算那样，为推理、内存移动和概率输出提供一等预算。",{},{"id":1013,"data":1014,"type":647,"tunes":1034},"v3-falsification-table",{"content":1015,"stretched":729,"withHeadings":15},[1016,1019,1022,1025,1028,1031],[1017,1018],"观察","对假设的影响",[1020,1021],"神经图形仍限于可选后处理","削弱推理成为核心引擎调度领域的说法",[1023,1024],"内存经济正常化，消费级带宽\u002F容量再次廉价扩展","减少用预测\u002F压缩替代原始资源的压力",[1026,1027],"跨厂商 ML API 未能成熟","使神经技术保持碎片化且集成成本高昂",[1029,1030],"预测开销\u002F浪费超过隐藏延迟","削弱预测内存的价值",[1032,1033],"开发者将 AI 代理完全置于延迟关键的引擎路径之外","将调度器收敛限制在图形\u002F数据而非所有 AI",{},{"id":1036,"data":1037,"type":572,"tunes":1039},"h-0209",{"text":1038,"level":47},"主要来源",{},{"id":1041,"data":1042,"type":1048,"tunes":1049},"src-0210",{"link":1043,"meta":1044},"https:\u002F\u002Fdevblogs.microsoft.com\u002Fdirectx\u002Fevolving-directx-for-the-ml-era-on-windows\u002F",{"image":1045,"title":1046,"description":1047},{"url":13},"Microsoft DirectX — 为 ML 时代演进 DirectX","DirectX 着色器级和模型级 ML 方向的官方来源。","linkTool",{},{"id":1051,"data":1052,"type":1048,"tunes":1058},"src-0212",{"link":1053,"meta":1054},"https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fabs\u002Fpii\u002FS0743731522002490",{"image":1055,"title":1056,"description":1057},{"url":13},"CPU-GPU UVM 中基于深度学习的数据预取","学习预测可以改善 GPU 内存页面预取的研究证据。",{},{"id":1060,"data":1061,"type":1048,"tunes":1067},"v3-src-aaai6",{"link":1062,"meta":1063},"https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F39224",{"image":1064,"title":1065,"description":1066},{"url":13},"AAAI 2026 — 异步 KV 缓存预取","主动内存移动和延迟隐藏的研究证据。",{},{"id":1069,"data":1070,"type":1048,"tunes":1076},"src-0211",{"link":1071,"meta":1072},"https:\u002F\u002Fresearch.nvidia.com\u002Flabs\u002Fadlr\u002FDLSS5\u002F",{"image":1073,"title":1074,"description":1075},{"url":13},"NVIDIA Research — DLSS 5","交互式图形管线中生成式神经渲染的主要研究来源。",{},{"id":1078,"data":1079,"type":1102,"tunes":1103},"faq-0213",{"items":1080,"title":1101},[1081,1085,1089,1093,1097],{"id":1082,"answer":1083,"question":1084},"faq1","不会。引擎仍然是拥有模拟、状态、调度和集成的系统。AI 成为该架构内的另一类工作负载。","AI 会取代游戏引擎吗？",{"id":1086,"answer":1087,"question":1088},"faq2","一些图形 ML 必须影响当前帧。其他 AI 可以异步，但仍需要当前状态和资源限制。","为什么不异步运行每个模型？",{"id":1090,"answer":1091,"question":1092},"faq3","协调截止时间、内存和权限，使概率系统改善体验而不破坏帧时间或游戏状态的稳定性。","最大的技术挑战是什么？",{"id":1094,"answer":1095,"question":1096},"faq4","不能。它可以简化容量和共享，但也可能增加争用。调度和局部性仍然重要。","统一内存能解决调度问题吗？",{"id":1098,"answer":1099,"question":1100},"faq5","未来引擎可能在确定性模拟和概率计算之间进行仲裁，而不是将 AI 视为一个单独的功能。","Figure Rocks 的核心观点是什么？","未来游戏引擎架构","faq",{},{"id":1105,"data":1106,"type":1125,"tunes":1126},"gloss-0214",{"title":1107,"entries":1108},"神经引擎术语",[1109,1113,1117,1121],{"term":1110,"anchor":1111,"definition":1112},"帧关键推理","frame-critical","必须在渲染截止时间内或附近完成的 ML 工作。",{"term":1114,"anchor":1115,"definition":1116},"审慎推理","deliberative","较慢的推理或规划，通常可以在即时帧截止时间之外运行。",{"term":1118,"anchor":1119,"definition":1120},"权限边界","authority-boundary","定义模型可以推荐什么与确定性工具或游戏系统实际可以改变什么的规则。",{"term":1122,"anchor":1123,"definition":1124},"神经引擎循环","neural-engine-loop","Figure Rocks 综合，用于在实时预算下协调状态、预测、内存、渲染、代理和验证。","glossary",{},"2.31","游戏引擎过去用于编排模拟与渲染。神经图形、预测性内存和游戏智能体正在增加第三项职责：决定推理在帧和游戏状态中的位置。","\u002Fuploads\u002F2026\u002F10\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system-1792340569647-63jv23.webp","the-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system-1792340569647-63jv23","PUBLISHED","2026-10-18T12:21:00.000Z","2026-10-18T16:21:24.252Z","2026-10-18T16:26:41.169Z",{"en":1136,"de":1137,"sr":1138,"es":1139,"fr":1140,"it":1141,"ru":1142,"zh":1143},"\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u002Fde\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u002Fsr\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u002Fes\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u002Ffr\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u002Fit\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u002Fru\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system","\u002Fzh\u002Fblog\u002Fthe-game-engine-is-becoming-part-renderer-part-scheduler-part-ai-system",[1145,1148,1152,1156],{"id":1146,"name":291,"slug":1147},144,"tech",{"id":1149,"name":1150,"slug":1151},170,"科技新闻","tech-news",{"id":1153,"name":1154,"slug":1155},165,"新闻","news",{"id":1157,"name":1158,"slug":1159},427,"新品","new",{"id":283,"login":1161,"email":13,"displayName":1162},"aleksandar","Aleksandar Stajic",[1164,1660],{"lang":8,"title":1165,"content":1166,"contentJson":1167,"excerpt":1659},"The Game Engine Is Becoming Part Renderer, Part Scheduler, Part AI System","{\"time\":1792340552128,\"blocks\":[{\"id\":\"p-0173\",\"data\":{\"text\":\"A modern game engine already coordinates simulation, animation, physics, audio, networking, asset streaming and rendering under a frame deadline. Neural rendering and game AI do not replace those responsibilities. They add another class of work that competes for the same time, memory and hardware. That turns the engine into something closer to a real-time resource scheduler for conventional and learned computation.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"c-0174\",\"data\":{\"body\":\"\u003Cstrong>The engine is becoming the place where rendering, data movement and inference are negotiated together.\u003C\u002Fstrong> DirectX is moving ML primitives into the graphics stack, DLSS 5 demonstrates a generative rendering stage, local game assistants consume GPU resources, and learned prefetching shows that memory movement itself can be predicted. The technical challenge is no longer merely 'run AI in a game'; it is deciding which AI work belongs where without breaking frame time, latency or game-state correctness.\",\"title\":\"Direct answer\",\"variant\":\"info\"},\"type\":\"callout\",\"tunes\":{}},{\"id\":\"c-0175\",\"data\":{\"body\":\"This article describes a direction assembled from current APIs, research and shipping\u002Fannounced systems. The final integrated engine architecture is a Figure Rocks synthesis, not a claim that Unreal, Unity or another engine already implements the whole stack.\",\"title\":\"Scope\",\"variant\":\"note\"},\"type\":\"callout\",\"tunes\":{}},{\"id\":\"toc-0176\",\"data\":{\"title\":\"Contents\",\"maxLevel\":3,\"minLevel\":2},\"type\":\"tableOfContents\",\"tunes\":{}},{\"id\":\"h-0177\",\"data\":{\"text\":\"The old mental model was already incomplete\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"flow-0178\",\"data\":{\"steps\":[{\"label\":\"Input\",\"description\":\"Read player\u002Fcontroller\u002Fnetwork input.\"},{\"label\":\"Simulation\",\"description\":\"Advance game logic, physics, AI and animation state.\"},{\"label\":\"Prepare render work\",\"description\":\"Build visibility, draw\u002Fdispatch lists, resources and pipelines.\"},{\"label\":\"Render\",\"description\":\"Raster, ray tracing and compute produce the frame.\"},{\"label\":\"Present\",\"description\":\"Frame reaches the display while the next frame begins.\"}],\"title\":\"Classic simplified engine loop\",\"orientation\":\"auto\"},\"type\":\"processFlow\",\"tunes\":{}},{\"id\":\"p-0179\",\"data\":{\"text\":\"Even this simplification hides asynchronous jobs, multithreading and streaming. But it is useful because neural workloads can now enter at several different points instead of living in one isolated 'AI' box.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"h-0180\",\"data\":{\"text\":\"Neural work is splitting into multiple timescales\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"cmp-0181\",\"data\":{\"rows\":[{\"id\":\"input\",\"label\":\"Input \u002F simulation response\",\"values\":{\"failure\":\"Latency or simulation instability\",\"deadline\":\"Milliseconds \u002F game tick\",\"resource\":\"CPU\u002Fgame-state scheduling\"}},{\"id\":\"render\",\"label\":\"Raster \u002F RT \u002F shading\",\"values\":{\"failure\":\"Frame-time spike or missed target FPS\",\"deadline\":\"Current frame budget\",\"resource\":\"GPU compute + memory bandwidth\"}},{\"id\":\"prefetch\",\"label\":\"Asset \u002F memory prefetch\",\"values\":{\"failure\":\"Stall if late; bandwidth waste if too early\",\"deadline\":\"Before future demand\",\"resource\":\"Storage, interconnect, cache and residency\"}},{\"id\":\"neuralgfx\",\"label\":\"Frame-critical neural graphics\",\"values\":{\"failure\":\"Artifact, latency or missed frame budget\",\"deadline\":\"Inside or adjacent to frame production\",\"resource\":\"GPU matrix\u002FTensor compute + bandwidth\"}},{\"id\":\"agent\",\"label\":\"Agent \u002F NPC reasoning\",\"values\":{\"failure\":\"Bad planning if rushed; gameplay lag if placed in reflex loop\",\"deadline\":\"Often tens\u002Fhundreds of milliseconds or asynchronous\",\"resource\":\"Model memory + CPU\u002FGPU\u002FNPU inference time\"}}],\"title\":\"One engine, multiple deadlines\",\"layout\":\"table\",\"columns\":[{\"id\":\"deadline\",\"label\":\"Typical deadline\"},{\"id\":\"failure\",\"label\":\"If scheduled badly\"},{\"id\":\"resource\",\"label\":\"Primary pressure\"}]},\"type\":\"comparison\",\"tunes\":{}},{\"id\":\"h-0182\",\"data\":{\"text\":\"DirectX is making ML part of graphics infrastructure\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"p-0183\",\"data\":{\"text\":\"Microsoft's 2026 DirectX work is important because it treats machine-learning math as a graphics-platform concern. Cooperative Vector and DX Linear Algebra let smaller neural workloads sit close to shader execution, while the announced Compute Graph Compiler targets larger model graphs. That creates an API-level path for engines to reason about ML as scheduled GPU work rather than an external black box.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"ref-0184\",\"data\":{\"url\":\"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fdirectx-is-becoming-an-ml-platform-what-linear-algebra-and-neural-shaders-mean-for-future-games\",\"title\":\"DirectX Is Becoming an ML Platform: What Linear Algebra and Neural Shaders Mean for Future Games\",\"excerpt\":\"A deeper look at shader-level ML, model-level graphs and why common graphics APIs matter for adoption.\",\"ctaLabel\":\"Read the DirectX ML article\"},\"type\":\"referralArticle\",\"tunes\":{}},{\"id\":\"h-0185\",\"data\":{\"text\":\"DLSS 5 moves generation deeper into the frame story\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"p-0186\",\"data\":{\"text\":\"NVIDIA describes DLSS 5 as a real-time generative rendering stage that complements conventional rendering with learned appearance priors. The architectural implication is more important than one product version: the final frame can increasingly contain information that was not produced only by conventional scene simulation and shading.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"c-0187\",\"data\":{\"body\":\"If a neural stage can save conventional rendering work but itself consumes compute, memory and time, the engine needs a \u003Cstrong>value test\u003C\u002Fstrong>: what cost disappeared, what inference cost was added, and did the player actually get a better frame within the same latency envelope?\",\"title\":\"The new scheduling question\",\"variant\":\"success\"},\"type\":\"callout\",\"tunes\":{}},{\"id\":\"v3-store-move-render\",\"data\":{\"text\":\"The engine may optimize five budgets at once\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"v3-five-budgets\",\"data\":{\"content\":[[\"Budget\",\"Traditional question\",\"Neural-era question\"],[\"Storage\",\"Is the asset installed?\",\"Can a compact representation replace part of the stored detail?\"],[\"Memory capacity\",\"Does the working set fit?\",\"Which assets, model states and neural representations deserve residency?\"],[\"Bandwidth\",\"Can required bytes arrive in time?\",\"Can compression\u002Fprefetch reduce or hide transfers?\"],[\"Render compute\",\"Can shaders\u002FRT finish inside the frame?\",\"Which detail is cheaper to infer than conventionally render?\"],[\"Inference compute\",\"Mostly outside the graphics loop\",\"Which neural work earns its compute and latency cost this frame?\"]],\"stretched\":false,\"withHeadings\":true},\"type\":\"table\",\"tunes\":{}},{\"id\":\"v3-five-mantra\",\"data\":{\"body\":\"\u003Cstrong>Store less → move less → predict more → render less → infer more\u003C\u002Fstrong> is useful only if each substitution is cheaper than the work it replaces. The engine scheduler is where that trade has to be measured.\",\"title\":\"The architectural shorthand\",\"variant\":\"success\"},\"type\":\"callout\",\"tunes\":{}},{\"id\":\"h-0188\",\"data\":{\"text\":\"Memory scheduling becomes part of graphics scheduling\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"p-0189\",\"data\":{\"text\":\"Neural graphics adds weights, latent representations and history buffers on top of textures, geometry and render targets. Asset streaming already has to anticipate what the player will see. A future engine can make that anticipation more explicit by ranking future memory demand and deciding which representation should be resident before the frame reaches it.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"ref-0190\",\"data\":{\"url\":\"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fwhat-if-a-game-could-predict-its-next-memory-access-ai-prefetching-and-the-future-of-stutter\",\"title\":\"What If a Game Could Predict Its Next Memory Access? AI Prefetching and the Future of Stutter\",\"excerpt\":\"The Figure Rocks Predictive Gaming Memory hypothesis: deterministic game state plus learned prefetching.\",\"ctaLabel\":\"Read the predictive memory analysis\"},\"type\":\"referralArticle\",\"tunes\":{}},{\"id\":\"flow-0191\",\"data\":{\"steps\":[{\"label\":\"1. Resolve game state\",\"description\":\"What is true in simulation now?\"},{\"label\":\"2. Predict near-future demand\",\"description\":\"Which scene, assets, shaders and neural data are likely next?\"},{\"label\":\"3. Allocate frame budget\",\"description\":\"How much time goes to simulation, conventional graphics, RT and neural stages?\"},{\"label\":\"4. Allocate memory\u002Fbandwidth budget\",\"description\":\"Which resources stay resident and which can be compressed or prefetched?\"},{\"label\":\"5. Execute fast neural work\",\"description\":\"Reconstruction, denoising, texture\u002Fmaterial inference or generation inside the graphics path.\"},{\"label\":\"6. Execute slow AI asynchronously\",\"description\":\"Assistants\u002FNPC reasoning operate outside the critical render deadline where possible.\"},{\"label\":\"7. Validate output\",\"description\":\"Check frame pacing, latency, image stability and game-state correctness.\"},{\"label\":\"8. Feed measurements back\",\"description\":\"The scheduler adapts policies based on misses, stalls and quality results.\"}],\"title\":\"The Figure Rocks Neural Engine Loop\",\"orientation\":\"auto\"},\"type\":\"processFlow\",\"tunes\":{}},{\"id\":\"v3-predictive-engine\",\"data\":{\"text\":\"Predictive memory is an engine problem because the engine owns context\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"v3-predictive-engine-p\",\"data\":{\"text\":\"A driver can observe memory behavior, but a game engine knows semantic state: the next level cell, camera direction, asset dependencies, animation graph, nearby actors and traversal constraints. That lets an engine combine deterministic knowledge with learned ranking, then schedule speculative movement only when current rendering and simulation leave enough bandwidth headroom.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"v3-engine-research\",\"data\":{\"body\":\"Learned UVM prefetching has improved page hit rate and reduced CPU–GPU traffic in published GPU benchmarks, while 2026 asynchronous KV-cache research shows how proactive movement can hide HBM latency. The Figure Rocks step is the \u003Cstrong>engine-semantic extension\u003C\u002Fstrong>: using knowledge of the game world to improve what gets predicted.\",\"title\":\"Research evidence supports the mechanism, not the gaming result\",\"variant\":\"note\"},\"type\":\"callout\",\"tunes\":{}},{\"id\":\"h-0192\",\"data\":{\"text\":\"Game agents create a different kind of engine problem\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"p-0193\",\"data\":{\"text\":\"An AI teammate or assistant is not just another shader. It reasons over game knowledge and live state, may call tools, and can produce actions that affect the world. Figure Rocks has already separated fast reactive gameplay control from slower language reasoning in the PUBG Ally architecture. That same separation belongs at the engine level: not every intelligent operation belongs in the same loop or deserves the same authority.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"ref-0194\",\"data\":{\"url\":\"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fpubg-ally-shows-why-ai-teammates-need-two-brains-fast-reflexes-and-slow-reasoning\",\"title\":\"PUBG Ally Shows Why AI Teammates Need Two Brains: Fast Reflexes and Slow Reasoning\",\"excerpt\":\"Why language reasoning and tick-level reactive control have different timing and authority requirements.\",\"ctaLabel\":\"Read the two-speed AI architecture\"},\"type\":\"referralArticle\",\"tunes\":{}},{\"id\":\"cmp-0195\",\"data\":{\"rows\":[{\"id\":\"input\",\"label\":\"Primary input\",\"values\":{\"gfx\":\"Frame buffers, motion\u002Fhistory, geometry\u002Fmaterial data\",\"agent\":\"Game state, player request, retrieved knowledge\"}},{\"id\":\"output\",\"label\":\"Primary output\",\"values\":{\"gfx\":\"Pixels, reconstructed features or render data\",\"agent\":\"Plan, explanation, tool\u002Faction request\"}},{\"id\":\"deadline\",\"label\":\"Timing\",\"values\":{\"gfx\":\"Usually frame-critical\",\"agent\":\"Often asynchronous or slower\"}},{\"id\":\"validation\",\"label\":\"Validation\",\"values\":{\"gfx\":\"Image quality, stability, latency\",\"agent\":\"State validity, authority, correctness, safety\"}},{\"id\":\"failure\",\"label\":\"Failure visibility\",\"values\":{\"gfx\":\"Artifact or hitch\",\"agent\":\"Bad advice or wrong action\"}}],\"title\":\"Graphics inference vs agent reasoning\",\"layout\":\"table\",\"columns\":[{\"id\":\"gfx\",\"label\":\"Graphics inference\"},{\"id\":\"agent\",\"label\":\"Game-agent reasoning\"}]},\"type\":\"comparison\",\"tunes\":{}},{\"id\":\"h-0196\",\"data\":{\"text\":\"The scheduler becomes the real product\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"p-0197\",\"data\":{\"text\":\"As more specialized hardware appears, the engine's value shifts from merely exposing features to placing work correctly. A CPU can be idle while the GPU stalls on data. Tensor hardware can be available while VRAM is saturated. An NPU can exist but be too disconnected from the engine's frame path. The fastest component does not determine the experience if the scheduler cannot feed and sequence the system.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"t-0198\",\"data\":{\"content\":[[\"Engine decision\",\"Bad simplification\",\"Better question\"],[\"Run neural rendering\",\"\\\"Tensor cores are fast\\\"\",\"Does inference save more frame cost than it adds, including data movement?\"],[\"Run local assistant\",\"\\\"The model fits in VRAM\\\"\",\"How much free memory\u002Fcompute remains during the actual game?\"],[\"Prefetch assets\",\"\\\"Load the next zone\\\"\",\"Which subset is most likely needed and what bandwidth budget is safe?\"],[\"Use unified memory\",\"\\\"Everything can access everything\\\"\",\"Which workload gets bandwidth priority at the frame deadline?\"],[\"Generate frames\",\"\\\"Displayed FPS is higher\\\"\",\"What is the base render rate, latency and temporal quality?\"]],\"stretched\":false,\"withHeadings\":true},\"type\":\"table\",\"tunes\":{}},{\"id\":\"h-0199\",\"data\":{\"text\":\"This is where stajic.de and Figure Rocks should lightly meet\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"p-0200\",\"data\":{\"text\":\"The general AI architecture question is about models, retrieval, tools, protocols and authority. Figure Rocks should keep the applied boundary: what happens when those components must coexist with a real-time game engine. A tool protocol can be correct in software terms and still be unusable if it introduces a frame hitch or acts on stale game state.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"ref-0201\",\"data\":{\"url\":\"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained\",\"title\":\"MCP vs A2A vs UCP vs AP2 vs A2UI: The Agent Protocol Stack Explained\",\"excerpt\":\"General protocol architecture on stajic.de. Figure Rocks applies the same boundary thinking to games, tools and real-time state.\",\"ctaLabel\":\"Read the protocol-stack explainer\"},\"type\":\"referralArticle\",\"tunes\":{}},{\"id\":\"h-0202\",\"data\":{\"text\":\"The six boundaries a neural game engine must enforce\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"flow-0203\",\"data\":{\"steps\":[{\"label\":\"1. Frame boundary\",\"description\":\"Which work must finish before present?\"},{\"label\":\"2. State boundary\",\"description\":\"Which data describes the current game rather than general knowledge?\"},{\"label\":\"3. Memory boundary\",\"description\":\"Which resources must be resident now versus predicted for later?\"},{\"label\":\"4. Authority boundary\",\"description\":\"Which model may suggest an action and which subsystem may execute it?\"},{\"label\":\"5. Quality boundary\",\"description\":\"What artifact, error or hallucination level is acceptable for this feature?\"},{\"label\":\"6. Fallback boundary\",\"description\":\"What deterministic path takes over when the neural path is too slow, unavailable or uncertain?\"}],\"title\":\"Neural Engine Boundary Test\",\"orientation\":\"auto\"},\"type\":\"processFlow\",\"tunes\":{}},{\"id\":\"h-0204\",\"data\":{\"text\":\"What the architecture could look like\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"cmp-0205\",\"data\":{\"rows\":[{\"id\":\"state\",\"label\":\"Game-state core\",\"values\":{\"role\":\"Deterministic authority\",\"responsibility\":\"Simulation, rules, authoritative state\"}},{\"id\":\"sched\",\"label\":\"Real-time scheduler\",\"values\":{\"role\":\"Deadline manager\",\"responsibility\":\"CPU\u002FGPU\u002FRT\u002FML timing and budgets\"}},{\"id\":\"memory\",\"label\":\"Memory intelligence\",\"values\":{\"role\":\"Residency\u002Fprefetch manager\",\"responsibility\":\"Compression, tiering, prediction and transfers\"}},{\"id\":\"render\",\"label\":\"Hybrid renderer\",\"values\":{\"role\":\"Visual producer\",\"responsibility\":\"Conventional + neural graphics stages\"}},{\"id\":\"agent\",\"label\":\"Agent layer\",\"values\":{\"role\":\"Deliberative intelligence\",\"responsibility\":\"Reasoning, dialogue, planning and tools under authority limits\"}},{\"id\":\"verify\",\"label\":\"Validation layer\",\"values\":{\"role\":\"Guardrail\",\"responsibility\":\"Frame quality, state validity, latency and fallback\"}}],\"title\":\"A possible future engine stack\",\"layout\":\"table\",\"columns\":[{\"id\":\"role\",\"label\":\"Role\"},{\"id\":\"responsibility\",\"label\":\"Responsibility\"}]},\"type\":\"comparison\",\"tunes\":{}},{\"id\":\"c-0206\",\"data\":{\"body\":\"The important transition is not 'games will use more AI.' It is that \u003Cstrong>the engine may become the arbitration layer between deterministic simulation and probabilistic computation\u003C\u002Fstrong>. The engine decides when learned systems are valuable, how much resource they may consume, what state they may see, and what happens when they are wrong.\",\"title\":\"Figure Rocks synthesis\",\"variant\":\"note\"},\"type\":\"callout\",\"tunes\":{}},{\"id\":\"h-0207\",\"data\":{\"text\":\"What would change this answer?\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"p-0208\",\"data\":{\"text\":\"If neural graphics and local game agents remain isolated optional features, the engine can keep treating them as plug-ins around a mostly conventional core. If they become common, engine architecture will need first-class budgeting for inference, memory movement and probabilistic outputs in the same way modern engines already budget CPU, GPU and streaming work.\"},\"type\":\"paragraph\",\"tunes\":{}},{\"id\":\"v3-falsification-table\",\"data\":{\"content\":[[\"Observation\",\"Effect on hypothesis\"],[\"Neural graphics remains limited to optional post-processing\",\"Weakens the claim that inference becomes a core engine scheduling domain\"],[\"Memory economics normalize and consumer bandwidth\u002Fcapacity scale cheaply again\",\"Reduces pressure to substitute prediction\u002Fcompression for raw resources\"],[\"Cross-vendor ML APIs fail to mature\",\"Keeps neural techniques fragmented and expensive to integrate\"],[\"Prediction overhead\u002Fwaste exceeds hidden latency\",\"Weakens predictive-memory value\"],[\"Developers keep AI agents entirely outside latency-critical engine paths\",\"Limits the scheduler convergence to graphics\u002Fdata rather than all AI\"]],\"stretched\":false,\"withHeadings\":true},\"type\":\"table\",\"tunes\":{}},{\"id\":\"h-0209\",\"data\":{\"text\":\"Primary sources\",\"level\":2},\"type\":\"header\",\"tunes\":{}},{\"id\":\"src-0210\",\"data\":{\"link\":\"https:\u002F\u002Fdevblogs.microsoft.com\u002Fdirectx\u002Fevolving-directx-for-the-ml-era-on-windows\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Microsoft DirectX — Evolving DirectX for the ML Era\",\"description\":\"Official source for DirectX shader-level and model-level ML direction.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"src-0212\",\"data\":{\"link\":\"https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fabs\u002Fpii\u002FS0743731522002490\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Deep learning based data prefetching in CPU-GPU UVM\",\"description\":\"Research evidence that learned prediction can improve GPU-memory page prefetching.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"v3-src-aaai6\",\"data\":{\"link\":\"https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F39224\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"AAAI 2026 — Asynchronous KV Cache Prefetching\",\"description\":\"Research evidence for proactive memory movement and latency hiding.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"src-0211\",\"data\":{\"link\":\"https:\u002F\u002Fresearch.nvidia.com\u002Flabs\u002Fadlr\u002FDLSS5\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"NVIDIA Research — DLSS 5\",\"description\":\"Primary research source for generative neural rendering in an interactive graphics pipeline.\"}},\"type\":\"linkTool\",\"tunes\":{}},{\"id\":\"faq-0213\",\"data\":{\"items\":[{\"id\":\"faq1\",\"answer\":\"No. The engine remains the system that owns simulation, state, scheduling and integration. AI becomes another class of workload inside that architecture.\",\"question\":\"Will AI replace the game engine?\"},{\"id\":\"faq2\",\"answer\":\"Some graphics ML must affect the current frame. Other AI can be asynchronous, but it still needs current state and resource limits.\",\"question\":\"Why not run every model asynchronously?\"},{\"id\":\"faq3\",\"answer\":\"Coordinating deadlines, memory and authority so probabilistic systems improve the experience without destabilizing frame time or game state.\",\"question\":\"What is the biggest technical challenge?\"},{\"id\":\"faq4\",\"answer\":\"No. It can simplify capacity and sharing, but it can also increase contention. Scheduling and locality still matter.\",\"question\":\"Does unified memory solve the scheduling problem?\"},{\"id\":\"faq5\",\"answer\":\"Future engines may arbitrate between deterministic simulation and probabilistic compute rather than treating AI as one separate feature.\",\"question\":\"What is the key Figure Rocks idea?\"}],\"title\":\"Future game-engine architecture\"},\"type\":\"faq\",\"tunes\":{}},{\"id\":\"gloss-0214\",\"data\":{\"title\":\"Neural engine terms\",\"entries\":[{\"term\":\"Frame-critical inference\",\"anchor\":\"frame-critical\",\"definition\":\"ML work that must complete within or adjacent to the render deadline.\"},{\"term\":\"Deliberative inference\",\"anchor\":\"deliberative\",\"definition\":\"Slower reasoning or planning that can usually run outside the immediate frame deadline.\"},{\"term\":\"Authority boundary\",\"anchor\":\"authority-boundary\",\"definition\":\"The rule defining what a model may recommend versus what a deterministic tool or game system may actually change.\"},{\"term\":\"Neural Engine Loop\",\"anchor\":\"neural-engine-loop\",\"definition\":\"Figure Rocks synthesis for coordinating state, prediction, memory, rendering, agents and validation under real-time budgets.\"}]},\"type\":\"glossary\",\"tunes\":{}}],\"version\":\"2.31.6\"}",{"time":1168,"blocks":1169,"version":1658},1792340552128,[1170,1174,1179,1184,1188,1192,1212,1216,1220,1262,1266,1270,1277,1281,1285,1290,1294,1322,1327,1331,1335,1342,1371,1375,1379,1384,1388,1392,1399,1434,1438,1442,1470,1474,1478,1484,1488,1511,1515,1555,1560,1564,1568,1590,1594,1601,1608,1615,1621,1641],{"id":541,"data":1171,"type":544,"tunes":1173},{"text":1172},"A modern game engine already coordinates simulation, animation, physics, audio, networking, asset streaming and rendering under a frame deadline. Neural rendering and game AI do not replace those responsibilities. They add another class of work that competes for the same time, memory and hardware. That turns the engine into something closer to a real-time resource scheduler for conventional and learned computation.",{},{"id":547,"data":1175,"type":552,"tunes":1178},{"body":1176,"title":1177,"variant":551},"\u003Cstrong>The engine is becoming the place where rendering, data movement and inference are negotiated together.\u003C\u002Fstrong> DirectX is moving ML primitives into the graphics stack, DLSS 5 demonstrates a generative rendering stage, local game assistants consume GPU resources, and learned prefetching shows that memory movement itself can be predicted. The technical challenge is no longer merely 'run AI in a game'; it is deciding which AI work belongs where without breaking frame time, latency or game-state correctness.","Direct answer",{},{"id":555,"data":1180,"type":552,"tunes":1183},{"body":1181,"title":1182,"variant":559},"This article describes a direction assembled from current APIs, research and shipping\u002Fannounced systems. The final integrated engine architecture is a Figure Rocks synthesis, not a claim that Unreal, Unity or another engine already implements the whole stack.","Scope",{},{"id":562,"data":1185,"type":566,"tunes":1187},{"title":1186,"maxLevel":565,"minLevel":47},"Contents",{},{"id":569,"data":1189,"type":572,"tunes":1191},{"text":1190,"level":47},"The old mental model was already incomplete",{},{"id":575,"data":1193,"type":595,"tunes":1211},{"steps":1194,"title":1210,"orientation":594},[1195,1198,1201,1204,1207],{"label":1196,"description":1197},"Input","Read player\u002Fcontroller\u002Fnetwork input.",{"label":1199,"description":1200},"Simulation","Advance game logic, physics, AI and animation state.",{"label":1202,"description":1203},"Prepare render work","Build visibility, draw\u002Fdispatch lists, resources and pipelines.",{"label":1205,"description":1206},"Render","Raster, ray tracing and compute produce the frame.",{"label":1208,"description":1209},"Present","Frame reaches the display while the next frame begins.","Classic simplified engine loop",{},{"id":598,"data":1213,"type":544,"tunes":1215},{"text":1214},"Even this simplification hides asynchronous jobs, multithreading and streaming. But it is useful because neural workloads can now enter at several different points instead of living in one isolated 'AI' box.",{},{"id":603,"data":1217,"type":572,"tunes":1219},{"text":1218,"level":47},"Neural work is splitting into multiple timescales",{},{"id":608,"data":1221,"type":658,"tunes":1261},{"rows":1222,"title":1253,"layout":647,"columns":1254},[1223,1229,1235,1241,1247],{"id":612,"label":1224,"values":1225},"Input \u002F simulation response",{"failure":1226,"deadline":1227,"resource":1228},"Latency or simulation instability","Milliseconds \u002F game tick","CPU\u002Fgame-state scheduling",{"id":619,"label":1230,"values":1231},"Raster \u002F RT \u002F shading",{"failure":1232,"deadline":1233,"resource":1234},"Frame-time spike or missed target FPS","Current frame budget","GPU compute + memory bandwidth",{"id":626,"label":1236,"values":1237},"Asset \u002F memory prefetch",{"failure":1238,"deadline":1239,"resource":1240},"Stall if late; bandwidth waste if too early","Before future demand","Storage, interconnect, cache and residency",{"id":633,"label":1242,"values":1243},"Frame-critical neural graphics",{"failure":1244,"deadline":1245,"resource":1246},"Artifact, latency or missed frame budget","Inside or adjacent to frame production","GPU matrix\u002FTensor compute + bandwidth",{"id":640,"label":1248,"values":1249},"Agent \u002F NPC reasoning",{"failure":1250,"deadline":1251,"resource":1252},"Bad planning if rushed; gameplay lag if placed in reflex loop","Often tens\u002Fhundreds of milliseconds or asynchronous","Model memory + CPU\u002FGPU\u002FNPU inference time","One engine, multiple deadlines",[1255,1257,1259],{"id":650,"label":1256},"Typical deadline",{"id":653,"label":1258},"If scheduled badly",{"id":656,"label":1260},"Primary pressure",{},{"id":661,"data":1263,"type":572,"tunes":1265},{"text":1264,"level":47},"DirectX is making ML part of graphics infrastructure",{},{"id":666,"data":1267,"type":544,"tunes":1269},{"text":1268},"Microsoft's 2026 DirectX work is important because it treats machine-learning math as a graphics-platform concern. Cooperative Vector and DX Linear Algebra let smaller neural workloads sit close to shader execution, while the announced Compute Graph Compiler targets larger model graphs. That creates an API-level path for engines to reason about ML as scheduled GPU work rather than an external black box.",{},{"id":671,"data":1271,"type":677,"tunes":1276},{"url":1272,"title":1273,"excerpt":1274,"ctaLabel":1275},"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fdirectx-is-becoming-an-ml-platform-what-linear-algebra-and-neural-shaders-mean-for-future-games","DirectX Is Becoming an ML Platform: What Linear Algebra and Neural Shaders Mean for Future Games","A deeper look at shader-level ML, model-level graphs and why common graphics APIs matter for adoption.","Read the DirectX ML article",{},{"id":680,"data":1278,"type":572,"tunes":1280},{"text":1279,"level":47},"DLSS 5 moves generation deeper into the frame story",{},{"id":685,"data":1282,"type":544,"tunes":1284},{"text":1283},"NVIDIA describes DLSS 5 as a real-time generative rendering stage that complements conventional rendering with learned appearance priors. The architectural implication is more important than one product version: the final frame can increasingly contain information that was not produced only by conventional scene simulation and shading.",{},{"id":690,"data":1286,"type":552,"tunes":1289},{"body":1287,"title":1288,"variant":694},"If a neural stage can save conventional rendering work but itself consumes compute, memory and time, the engine needs a \u003Cstrong>value test\u003C\u002Fstrong>: what cost disappeared, what inference cost was added, and did the player actually get a better frame within the same latency envelope?","The new scheduling question",{},{"id":697,"data":1291,"type":572,"tunes":1293},{"text":1292,"level":47},"The engine may optimize five budgets at once",{},{"id":702,"data":1295,"type":647,"tunes":1321},{"content":1296,"stretched":729,"withHeadings":15},[1297,1301,1305,1309,1313,1317],[1298,1299,1300],"Budget","Traditional question","Neural-era question",[1302,1303,1304],"Storage","Is the asset installed?","Can a compact representation replace part of the stored detail?",[1306,1307,1308],"Memory capacity","Does the working set fit?","Which assets, model states and neural representations deserve residency?",[1310,1311,1312],"Bandwidth","Can required bytes arrive in time?","Can compression\u002Fprefetch reduce or hide transfers?",[1314,1315,1316],"Render compute","Can shaders\u002FRT finish inside the frame?","Which detail is cheaper to infer than conventionally render?",[1318,1319,1320],"Inference compute","Mostly outside the graphics loop","Which neural work earns its compute and latency cost this frame?",{},{"id":732,"data":1323,"type":552,"tunes":1326},{"body":1324,"title":1325,"variant":694},"\u003Cstrong>Store less → move less → predict more → render less → infer more\u003C\u002Fstrong> is useful only if each substitution is cheaper than the work it replaces. The engine scheduler is where that trade has to be measured.","The architectural shorthand",{},{"id":738,"data":1328,"type":572,"tunes":1330},{"text":1329,"level":47},"Memory scheduling becomes part of graphics scheduling",{},{"id":743,"data":1332,"type":544,"tunes":1334},{"text":1333},"Neural graphics adds weights, latent representations and history buffers on top of textures, geometry and render targets. Asset streaming already has to anticipate what the player will see. A future engine can make that anticipation more explicit by ranking future memory demand and deciding which representation should be resident before the frame reaches it.",{},{"id":748,"data":1336,"type":677,"tunes":1341},{"url":1337,"title":1338,"excerpt":1339,"ctaLabel":1340},"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fwhat-if-a-game-could-predict-its-next-memory-access-ai-prefetching-and-the-future-of-stutter","What If a Game Could Predict Its Next Memory Access? AI Prefetching and the Future of Stutter","The Figure Rocks Predictive Gaming Memory hypothesis: deterministic game state plus learned prefetching.","Read the predictive memory analysis",{},{"id":756,"data":1343,"type":595,"tunes":1370},{"steps":1344,"title":1369,"orientation":594},[1345,1348,1351,1354,1357,1360,1363,1366],{"label":1346,"description":1347},"1. Resolve game state","What is true in simulation now?",{"label":1349,"description":1350},"2. Predict near-future demand","Which scene, assets, shaders and neural data are likely next?",{"label":1352,"description":1353},"3. Allocate frame budget","How much time goes to simulation, conventional graphics, RT and neural stages?",{"label":1355,"description":1356},"4. Allocate memory\u002Fbandwidth budget","Which resources stay resident and which can be compressed or prefetched?",{"label":1358,"description":1359},"5. Execute fast neural work","Reconstruction, denoising, texture\u002Fmaterial inference or generation inside the graphics path.",{"label":1361,"description":1362},"6. Execute slow AI asynchronously","Assistants\u002FNPC reasoning operate outside the critical render deadline where possible.",{"label":1364,"description":1365},"7. Validate output","Check frame pacing, latency, image stability and game-state correctness.",{"label":1367,"description":1368},"8. Feed measurements back","The scheduler adapts policies based on misses, stalls and quality results.","The Figure Rocks Neural Engine Loop",{},{"id":786,"data":1372,"type":572,"tunes":1374},{"text":1373,"level":47},"Predictive memory is an engine problem because the engine owns context",{},{"id":791,"data":1376,"type":544,"tunes":1378},{"text":1377},"A driver can observe memory behavior, but a game engine knows semantic state: the next level cell, camera direction, asset dependencies, animation graph, nearby actors and traversal constraints. That lets an engine combine deterministic knowledge with learned ranking, then schedule speculative movement only when current rendering and simulation leave enough bandwidth headroom.",{},{"id":796,"data":1380,"type":552,"tunes":1383},{"body":1381,"title":1382,"variant":559},"Learned UVM prefetching has improved page hit rate and reduced CPU–GPU traffic in published GPU benchmarks, while 2026 asynchronous KV-cache research shows how proactive movement can hide HBM latency. The Figure Rocks step is the \u003Cstrong>engine-semantic extension\u003C\u002Fstrong>: using knowledge of the game world to improve what gets predicted.","Research evidence supports the mechanism, not the gaming result",{},{"id":802,"data":1385,"type":572,"tunes":1387},{"text":1386,"level":47},"Game agents create a different kind of engine problem",{},{"id":807,"data":1389,"type":544,"tunes":1391},{"text":1390},"An AI teammate or assistant is not just another shader. It reasons over game knowledge and live state, may call tools, and can produce actions that affect the world. Figure Rocks has already separated fast reactive gameplay control from slower language reasoning in the PUBG Ally architecture. That same separation belongs at the engine level: not every intelligent operation belongs in the same loop or deserves the same authority.",{},{"id":812,"data":1393,"type":677,"tunes":1398},{"url":1394,"title":1395,"excerpt":1396,"ctaLabel":1397},"https:\u002F\u002Ffigure.rocks\u002Fblog\u002Fpubg-ally-shows-why-ai-teammates-need-two-brains-fast-reflexes-and-slow-reasoning","PUBG Ally Shows Why AI Teammates Need Two Brains: Fast Reflexes and Slow Reasoning","Why language reasoning and tick-level reactive control have different timing and authority requirements.","Read the two-speed AI architecture",{},{"id":820,"data":1400,"type":658,"tunes":1433},{"rows":1401,"title":1427,"layout":647,"columns":1428},[1402,1407,1412,1417,1422],{"id":612,"label":1403,"values":1404},"Primary input",{"gfx":1405,"agent":1406},"Frame buffers, motion\u002Fhistory, geometry\u002Fmaterial data","Game state, player request, retrieved knowledge",{"id":829,"label":1408,"values":1409},"Primary output",{"gfx":1410,"agent":1411},"Pixels, reconstructed features or render data","Plan, explanation, tool\u002Faction request",{"id":650,"label":1413,"values":1414},"Timing",{"gfx":1415,"agent":1416},"Usually frame-critical","Often asynchronous or slower",{"id":840,"label":1418,"values":1419},"Validation",{"gfx":1420,"agent":1421},"Image quality, stability, latency","State validity, authority, correctness, safety",{"id":653,"label":1423,"values":1424},"Failure visibility",{"gfx":1425,"agent":1426},"Artifact or hitch","Bad advice or wrong action","Graphics inference vs agent reasoning",[1429,1431],{"id":853,"label":1430},"Graphics inference",{"id":640,"label":1432},"Game-agent reasoning",{},{"id":859,"data":1435,"type":572,"tunes":1437},{"text":1436,"level":47},"The scheduler becomes the real product",{},{"id":864,"data":1439,"type":544,"tunes":1441},{"text":1440},"As more specialized hardware appears, the engine's value shifts from merely exposing features to placing work correctly. A CPU can be idle while the GPU stalls on data. Tensor hardware can be available while VRAM is saturated. An NPU can exist but be too disconnected from the engine's frame path. The fastest component does not determine the experience if the scheduler cannot feed and sequence the system.",{},{"id":869,"data":1443,"type":647,"tunes":1469},{"content":1444,"stretched":729,"withHeadings":15},[1445,1449,1453,1457,1461,1465],[1446,1447,1448],"Engine decision","Bad simplification","Better question",[1450,1451,1452],"Run neural rendering","\"Tensor cores are fast\"","Does inference save more frame cost than it adds, including data movement?",[1454,1455,1456],"Run local assistant","\"The model fits in VRAM\"","How much free memory\u002Fcompute remains during the actual game?",[1458,1459,1460],"Prefetch assets","\"Load the next zone\"","Which subset is most likely needed and what bandwidth budget is safe?",[1462,1463,1464],"Use unified memory","\"Everything can access everything\"","Which workload gets bandwidth priority at the frame deadline?",[1466,1467,1468],"Generate frames","\"Displayed FPS is higher\"","What is the base render rate, latency and temporal quality?",{},{"id":898,"data":1471,"type":572,"tunes":1473},{"text":1472,"level":47},"This is where stajic.de and Figure Rocks should lightly meet",{},{"id":903,"data":1475,"type":544,"tunes":1477},{"text":1476},"The general AI architecture question is about models, retrieval, tools, protocols and authority. Figure Rocks should keep the applied boundary: what happens when those components must coexist with a real-time game engine. A tool protocol can be correct in software terms and still be unusable if it introduces a frame hitch or acts on stale game state.",{},{"id":908,"data":1479,"type":677,"tunes":1483},{"url":910,"title":1480,"excerpt":1481,"ctaLabel":1482},"MCP vs A2A vs UCP vs AP2 vs A2UI: The Agent Protocol Stack Explained","General protocol architecture on stajic.de. Figure Rocks applies the same boundary thinking to games, tools and real-time state.","Read the protocol-stack explainer",{},{"id":916,"data":1485,"type":572,"tunes":1487},{"text":1486,"level":47},"The six boundaries a neural game engine must enforce",{},{"id":921,"data":1489,"type":595,"tunes":1510},{"steps":1490,"title":1509,"orientation":594},[1491,1494,1497,1500,1503,1506],{"label":1492,"description":1493},"1. Frame boundary","Which work must finish before present?",{"label":1495,"description":1496},"2. State boundary","Which data describes the current game rather than general knowledge?",{"label":1498,"description":1499},"3. Memory boundary","Which resources must be resident now versus predicted for later?",{"label":1501,"description":1502},"4. Authority boundary","Which model may suggest an action and which subsystem may execute it?",{"label":1504,"description":1505},"5. Quality boundary","What artifact, error or hallucination level is acceptable for this feature?",{"label":1507,"description":1508},"6. Fallback boundary","What deterministic path takes over when the neural path is too slow, unavailable or uncertain?","Neural Engine Boundary Test",{},{"id":945,"data":1512,"type":572,"tunes":1514},{"text":1513,"level":47},"What the architecture could look like",{},{"id":950,"data":1516,"type":658,"tunes":1554},{"rows":1517,"title":1548,"layout":647,"columns":1549},[1518,1523,1528,1533,1538,1543],{"id":954,"label":1519,"values":1520},"Game-state core",{"role":1521,"responsibility":1522},"Deterministic authority","Simulation, rules, authoritative state",{"id":960,"label":1524,"values":1525},"Real-time scheduler",{"role":1526,"responsibility":1527},"Deadline manager","CPU\u002FGPU\u002FRT\u002FML timing and budgets",{"id":966,"label":1529,"values":1530},"Memory intelligence",{"role":1531,"responsibility":1532},"Residency\u002Fprefetch manager","Compression, tiering, prediction and transfers",{"id":619,"label":1534,"values":1535},"Hybrid renderer",{"role":1536,"responsibility":1537},"Visual producer","Conventional + neural graphics stages",{"id":640,"label":1539,"values":1540},"Agent layer",{"role":1541,"responsibility":1542},"Deliberative intelligence","Reasoning, dialogue, planning and tools under authority limits",{"id":982,"label":1544,"values":1545},"Validation layer",{"role":1546,"responsibility":1547},"Guardrail","Frame quality, state validity, latency and fallback","A possible future engine stack",[1550,1552],{"id":990,"label":1551},"Role",{"id":993,"label":1553},"Responsibility",{},{"id":997,"data":1556,"type":552,"tunes":1559},{"body":1557,"title":1558,"variant":559},"The important transition is not 'games will use more AI.' It is that \u003Cstrong>the engine may become the arbitration layer between deterministic simulation and probabilistic computation\u003C\u002Fstrong>. The engine decides when learned systems are valuable, how much resource they may consume, what state they may see, and what happens when they are wrong.","Figure Rocks synthesis",{},{"id":1003,"data":1561,"type":572,"tunes":1563},{"text":1562,"level":47},"What would change this answer?",{},{"id":1008,"data":1565,"type":544,"tunes":1567},{"text":1566},"If neural graphics and local game agents remain isolated optional features, the engine can keep treating them as plug-ins around a mostly conventional core. If they become common, engine architecture will need first-class budgeting for inference, memory movement and probabilistic outputs in the same way modern engines already budget CPU, GPU and streaming work.",{},{"id":1013,"data":1569,"type":647,"tunes":1589},{"content":1570,"stretched":729,"withHeadings":15},[1571,1574,1577,1580,1583,1586],[1572,1573],"Observation","Effect on hypothesis",[1575,1576],"Neural graphics remains limited to optional post-processing","Weakens the claim that inference becomes a core engine scheduling domain",[1578,1579],"Memory economics normalize and consumer bandwidth\u002Fcapacity scale cheaply again","Reduces pressure to substitute prediction\u002Fcompression for raw resources",[1581,1582],"Cross-vendor ML APIs fail to mature","Keeps neural techniques fragmented and expensive to integrate",[1584,1585],"Prediction overhead\u002Fwaste exceeds hidden latency","Weakens predictive-memory value",[1587,1588],"Developers keep AI agents entirely outside latency-critical engine paths","Limits the scheduler convergence to graphics\u002Fdata rather than all AI",{},{"id":1036,"data":1591,"type":572,"tunes":1593},{"text":1592,"level":47},"Primary sources",{},{"id":1041,"data":1595,"type":1048,"tunes":1600},{"link":1043,"meta":1596},{"image":1597,"title":1598,"description":1599},{"url":13},"Microsoft DirectX — Evolving DirectX for the ML Era","Official source for DirectX shader-level and model-level ML direction.",{},{"id":1051,"data":1602,"type":1048,"tunes":1607},{"link":1053,"meta":1603},{"image":1604,"title":1605,"description":1606},{"url":13},"Deep learning based data prefetching in CPU-GPU UVM","Research evidence that learned prediction can improve GPU-memory page prefetching.",{},{"id":1060,"data":1609,"type":1048,"tunes":1614},{"link":1062,"meta":1610},{"image":1611,"title":1612,"description":1613},{"url":13},"AAAI 2026 — Asynchronous KV Cache Prefetching","Research evidence for proactive memory movement and latency hiding.",{},{"id":1069,"data":1616,"type":1048,"tunes":1620},{"link":1071,"meta":1617},{"image":1618,"title":1074,"description":1619},{"url":13},"Primary research source for generative neural rendering in an interactive graphics pipeline.",{},{"id":1078,"data":1622,"type":1102,"tunes":1640},{"items":1623,"title":1639},[1624,1627,1630,1633,1636],{"id":1082,"answer":1625,"question":1626},"No. The engine remains the system that owns simulation, state, scheduling and integration. AI becomes another class of workload inside that architecture.","Will AI replace the game engine?",{"id":1086,"answer":1628,"question":1629},"Some graphics ML must affect the current frame. Other AI can be asynchronous, but it still needs current state and resource limits.","Why not run every model asynchronously?",{"id":1090,"answer":1631,"question":1632},"Coordinating deadlines, memory and authority so probabilistic systems improve the experience without destabilizing frame time or game state.","What is the biggest technical challenge?",{"id":1094,"answer":1634,"question":1635},"No. It can simplify capacity and sharing, but it can also increase contention. Scheduling and locality still matter.","Does unified memory solve the scheduling problem?",{"id":1098,"answer":1637,"question":1638},"Future engines may arbitrate between deterministic simulation and probabilistic compute rather than treating AI as one separate feature.","What is the key Figure Rocks idea?","Future game-engine architecture",{},{"id":1105,"data":1642,"type":1125,"tunes":1657},{"title":1643,"entries":1644},"Neural engine terms",[1645,1648,1651,1654],{"term":1646,"anchor":1111,"definition":1647},"Frame-critical inference","ML work that must complete within or adjacent to the render deadline.",{"term":1649,"anchor":1115,"definition":1650},"Deliberative inference","Slower reasoning or planning that can usually run outside the immediate frame deadline.",{"term":1652,"anchor":1119,"definition":1653},"Authority boundary","The rule defining what a model may recommend versus what a deterministic tool or game system may actually change.",{"term":1655,"anchor":1123,"definition":1656},"Neural Engine Loop","Figure Rocks synthesis for coordinating state, prediction, memory, rendering, agents and validation under real-time budgets.",{},"2.31.6","Game engines used to orchestrate simulation and rendering. Neural graphics, predictive memory and game agents are adding a third responsibility: deciding where inference belongs inside the frame and the game state.",{"lang":7,"title":534,"content":536,"contentJson":1661,"excerpt":1128},{"time":538,"blocks":1662,"version":1127},[1663,1666,1669,1672,1675,1678,1687,1690,1693,1711,1714,1717,1720,1723,1726,1729,1732,1742,1745,1748,1751,1754,1766,1769,1772,1775,1778,1781,1784,1801,1804,1807,1817,1820,1823,1826,1829,1839,1842,1861,1864,1867,1870,1880,1883,1888,1893,1898,1903,1912],{"id":541,"data":1664,"type":544,"tunes":1665},{"text":543},{},{"id":547,"data":1667,"type":552,"tunes":1668},{"body":549,"title":550,"variant":551},{},{"id":555,"data":1670,"type":552,"tunes":1671},{"body":557,"title":558,"variant":559},{},{"id":562,"data":1673,"type":566,"tunes":1674},{"title":564,"maxLevel":565,"minLevel":47},{},{"id":569,"data":1676,"type":572,"tunes":1677},{"text":571,"level":47},{},{"id":575,"data":1679,"type":595,"tunes":1686},{"steps":1680,"title":593,"orientation":594},[1681,1682,1683,1684,1685],{"label":579,"description":580},{"label":582,"description":583},{"label":585,"description":586},{"label":588,"description":589},{"label":591,"description":592},{},{"id":598,"data":1688,"type":544,"tunes":1689},{"text":600},{},{"id":603,"data":1691,"type":572,"tunes":1692},{"text":605,"level":47},{},{"id":608,"data":1694,"type":658,"tunes":1710},{"rows":1695,"title":646,"layout":647,"columns":1706},[1696,1698,1700,1702,1704],{"id":612,"label":613,"values":1697},{"failure":615,"deadline":616,"resource":617},{"id":619,"label":620,"values":1699},{"failure":622,"deadline":623,"resource":624},{"id":626,"label":627,"values":1701},{"failure":629,"deadline":630,"resource":631},{"id":633,"label":634,"values":1703},{"failure":636,"deadline":637,"resource":638},{"id":640,"label":641,"values":1705},{"failure":643,"deadline":644,"resource":645},[1707,1708,1709],{"id":650,"label":651},{"id":653,"label":654},{"id":656,"label":657},{},{"id":661,"data":1712,"type":572,"tunes":1713},{"text":663,"level":47},{},{"id":666,"data":1715,"type":544,"tunes":1716},{"text":668},{},{"id":671,"data":1718,"type":677,"tunes":1719},{"url":673,"title":674,"excerpt":675,"ctaLabel":676},{},{"id":680,"data":1721,"type":572,"tunes":1722},{"text":682,"level":47},{},{"id":685,"data":1724,"type":544,"tunes":1725},{"text":687},{},{"id":690,"data":1727,"type":552,"tunes":1728},{"body":692,"title":693,"variant":694},{},{"id":697,"data":1730,"type":572,"tunes":1731},{"text":699,"level":47},{},{"id":702,"data":1733,"type":647,"tunes":1741},{"content":1734,"stretched":729,"withHeadings":15},[1735,1736,1737,1738,1739,1740],[706,707,708],[710,711,712],[714,715,716],[718,719,720],[722,723,724],[726,727,728],{},{"id":732,"data":1743,"type":552,"tunes":1744},{"body":734,"title":735,"variant":694},{},{"id":738,"data":1746,"type":572,"tunes":1747},{"text":740,"level":47},{},{"id":743,"data":1749,"type":544,"tunes":1750},{"text":745},{},{"id":748,"data":1752,"type":677,"tunes":1753},{"url":750,"title":751,"excerpt":752,"ctaLabel":753},{},{"id":756,"data":1755,"type":595,"tunes":1765},{"steps":1756,"title":783,"orientation":594},[1757,1758,1759,1760,1761,1762,1763,1764],{"label":760,"description":761},{"label":763,"description":764},{"label":766,"description":767},{"label":769,"description":770},{"label":772,"description":773},{"label":775,"description":776},{"label":778,"description":779},{"label":781,"description":782},{},{"id":786,"data":1767,"type":572,"tunes":1768},{"text":788,"level":47},{},{"id":791,"data":1770,"type":544,"tunes":1771},{"text":793},{},{"id":796,"data":1773,"type":552,"tunes":1774},{"body":798,"title":799,"variant":559},{},{"id":802,"data":1776,"type":572,"tunes":1777},{"text":804,"level":47},{},{"id":807,"data":1779,"type":544,"tunes":1780},{"text":809},{},{"id":812,"data":1782,"type":677,"tunes":1783},{"url":814,"title":815,"excerpt":816,"ctaLabel":817},{},{"id":820,"data":1785,"type":658,"tunes":1800},{"rows":1786,"title":850,"layout":647,"columns":1797},[1787,1789,1791,1793,1795],{"id":612,"label":824,"values":1788},{"gfx":826,"agent":827},{"id":829,"label":830,"values":1790},{"gfx":832,"agent":833},{"id":650,"label":835,"values":1792},{"gfx":837,"agent":838},{"id":840,"label":841,"values":1794},{"gfx":843,"agent":844},{"id":653,"label":846,"values":1796},{"gfx":848,"agent":849},[1798,1799],{"id":853,"label":854},{"id":640,"label":856},{},{"id":859,"data":1802,"type":572,"tunes":1803},{"text":861,"level":47},{},{"id":864,"data":1805,"type":544,"tunes":1806},{"text":866},{},{"id":869,"data":1808,"type":647,"tunes":1816},{"content":1809,"stretched":729,"withHeadings":15},[1810,1811,1812,1813,1814,1815],[873,874,875],[877,878,879],[881,882,883],[885,886,887],[889,890,891],[893,894,895],{},{"id":898,"data":1818,"type":572,"tunes":1819},{"text":900,"level":47},{},{"id":903,"data":1821,"type":544,"tunes":1822},{"text":905},{},{"id":908,"data":1824,"type":677,"tunes":1825},{"url":910,"title":911,"excerpt":912,"ctaLabel":913},{},{"id":916,"data":1827,"type":572,"tunes":1828},{"text":918,"level":47},{},{"id":921,"data":1830,"type":595,"tunes":1838},{"steps":1831,"title":942,"orientation":594},[1832,1833,1834,1835,1836,1837],{"label":925,"description":926},{"label":928,"description":929},{"label":931,"description":932},{"label":934,"description":935},{"label":937,"description":938},{"label":940,"description":941},{},{"id":945,"data":1840,"type":572,"tunes":1841},{"text":947,"level":47},{},{"id":950,"data":1843,"type":658,"tunes":1860},{"rows":1844,"title":987,"layout":647,"columns":1857},[1845,1847,1849,1851,1853,1855],{"id":954,"label":955,"values":1846},{"role":957,"responsibility":958},{"id":960,"label":961,"values":1848},{"role":963,"responsibility":964},{"id":966,"label":967,"values":1850},{"role":969,"responsibility":970},{"id":619,"label":972,"values":1852},{"role":974,"responsibility":975},{"id":640,"label":977,"values":1854},{"role":979,"responsibility":980},{"id":982,"label":983,"values":1856},{"role":985,"responsibility":986},[1858,1859],{"id":990,"label":991},{"id":993,"label":994},{},{"id":997,"data":1862,"type":552,"tunes":1863},{"body":999,"title":1000,"variant":559},{},{"id":1003,"data":1865,"type":572,"tunes":1866},{"text":1005,"level":47},{},{"id":1008,"data":1868,"type":544,"tunes":1869},{"text":1010},{},{"id":1013,"data":1871,"type":647,"tunes":1879},{"content":1872,"stretched":729,"withHeadings":15},[1873,1874,1875,1876,1877,1878],[1017,1018],[1020,1021],[1023,1024],[1026,1027],[1029,1030],[1032,1033],{},{"id":1036,"data":1881,"type":572,"tunes":1882},{"text":1038,"level":47},{},{"id":1041,"data":1884,"type":1048,"tunes":1887},{"link":1043,"meta":1885},{"image":1886,"title":1046,"description":1047},{"url":13},{},{"id":1051,"data":1889,"type":1048,"tunes":1892},{"link":1053,"meta":1890},{"image":1891,"title":1056,"description":1057},{"url":13},{},{"id":1060,"data":1894,"type":1048,"tunes":1897},{"link":1062,"meta":1895},{"image":1896,"title":1065,"description":1066},{"url":13},{},{"id":1069,"data":1899,"type":1048,"tunes":1902},{"link":1071,"meta":1900},{"image":1901,"title":1074,"description":1075},{"url":13},{},{"id":1078,"data":1904,"type":1102,"tunes":1911},{"items":1905,"title":1101},[1906,1907,1908,1909,1910],{"id":1082,"answer":1083,"question":1084},{"id":1086,"answer":1087,"question":1088},{"id":1090,"answer":1091,"question":1092},{"id":1094,"answer":1095,"question":1096},{"id":1098,"answer":1099,"question":1100},{},{"id":1105,"data":1913,"type":1125,"tunes":1919},{"title":1107,"entries":1914},[1915,1916,1917,1918],{"term":1110,"anchor":1111,"definition":1112},{"term":1114,"anchor":1115,"definition":1116},{"term":1118,"anchor":1119,"definition":1120},{"term":1122,"anchor":1123,"definition":1124},{},"Post erfolgreich abgerufen",{"items":1922,"source":1989,"manualIds":1990,"manualMatchedIds":1991},[1923,1929,1935,1942,1949,1956,1963,1970,1976,1983],{"id":1924,"slug":1925,"title":1926,"excerpt":1927,"featuredImage":14,"publishedAt":1928},"398","upcoming-nintendo-amiibo-new-figures-release-windows-and-functional-significance","即将推出的任天堂amiibo——新模型、发售窗口与功能意义","amiibo是配备NFC芯片的实体手办，可与任天堂游戏机进行通信。在兼容游戏中，它们能解锁游戏内物品、装饰元素或小型游戏功能。本目录概览主要关注官方已公布的即将推出的amiibo，包括其预计发布时间窗口、功能价值以及在更广泛的产品批次中的定位。","2026-02-27T15:54:00.000Z",{"id":1930,"slug":1931,"title":1932,"excerpt":14,"featuredImage":1933,"publishedAt":1934},"456","what-would-a-spark-like-gaming-pc-actually-change-unified-memory-ai-compute-and-the-classic-gpu-split","一台类似Spark的游戏PC究竟会带来什么改变？统一内存、AI计算与经典的GPU分工","\u002Fuploads\u002F2026\u002F10\u002Fwhat-would-a-spark-like-gaming-pc-actually-change-unified-memory-ai-compute-and-the-classic-gpu-split-1792340021752-bia7n6.webp","2026-10-18T12:12:00.000Z",{"id":1936,"slug":1937,"title":1938,"excerpt":1939,"featuredImage":1940,"publishedAt":1941},"434","google-io-2026-gaming-ai-xr-and-the-next-wave-of-collectibles","Google I\u002FO 2026 与游戏：AI、XR 以及下一波收藏品浪潮","Google I\u002FO 2026并非一场典型的游戏活动，但其AI、XR、Gemini、Google Play和创意工具方面的公告指向了游戏文化的一次明显转变。本文解释了AI发现、XR伴侣层、Gemini驱动的媒体以及更快的角色可见性可能如何塑造游戏、手办、收藏品和未来的粉丝生态系统。","\u002Fuploads\u002F2026\u002F05\u002Fgoogle-io-2026-gaming-ai-xr-and-the-next-wave-of-collectibles-1779257974958-jjj07f.webp","2026-05-20T00:17:00.000Z",{"id":1943,"slug":1944,"title":1945,"excerpt":1946,"featuredImage":1947,"publishedAt":1948},"454","why-ai-is-reshaping-gaming-memory-hbm-gddr-vram-and-the-fight-for-capacity","为什么AI正在重塑游戏内存：HBM、GDDR、VRAM与容量之争","现代游戏背后的内存问题不再仅仅是GPU需要多少千兆字节。AI基础设施正在改变供应商优先考虑哪些内存产品，以及消费者买家能负担多少。","\u002Fuploads\u002F2026\u002F10\u002Fwhy-ai-is-reshaping-gaming-memory-hbm-gddr-vram-and-the-fight-for-capacity-1792335495408-ymehaw.webp","2026-10-18T10:56:00.000Z",{"id":1950,"slug":1951,"title":1952,"excerpt":1953,"featuredImage":1954,"publishedAt":1955},"457","what-does-future-proof-gaming-hardware-mean-in-the-neural-rendering-era","在神经渲染时代，“面向未来”的游戏硬件意味着什么？","过去，面向未来意味着购买比当下所需更多的 CPU、GPU 和 VRAM。神经渲染让这条规则变得不那么可靠。这里有一个更好的决策框架。","\u002Fuploads\u002F2026\u002F10\u002Fwhat-does-future-proof-gaming-hardware-mean-in-the-neural-rendering-era-1792340303825-012uhb.webp","2026-10-18T12:16:00.000Z",{"id":1957,"slug":1958,"title":1959,"excerpt":1960,"featuredImage":1961,"publishedAt":1962},"453","render-less-infer-more-is-ai-breaking-the-traditional-gaming-hardware-upgrade-cycle","渲染更少，推理更多：AI正在打破传统游戏硬件升级周期吗？","人工智能正使高性能芯片和存储器在游戏之外更具价值，同时也为游戏提供了新的渲染、压缩和预测方式。本分析探讨这两种趋势是否会汇聚成一种新的硬件模式。","\u002Fuploads\u002F2026\u002F10\u002Frender-less-infer-more-is-ai-breaking-the-traditional-gaming-hardware-upgrade-cycle-1792323493963-l2x8qb.webp","2026-10-18T03:34:00.000Z",{"id":1964,"slug":1965,"title":1966,"excerpt":1967,"featuredImage":1968,"publishedAt":1969},"455","what-if-a-game-could-predict-its-next-memory-access-ai-prefetching-and-the-future-of-stutter","如果一款游戏能预测它的下一次内存访问会怎样？AI预取与卡顿的未来","游戏引擎已经知道玩家、摄像机和世界将去向何处。研究表明，学习式GPU预取能够预测内存页。如果将这些想法结合起来会发生什么？","\u002Fuploads\u002F2026\u002F10\u002Fwhat-if-a-game-could-predict-its-next-memory-access-ai-prefetching-and-the-future-of-stutter-1792339100514-lyx0yw.webp","2026-10-18T11:57:00.000Z",{"id":1971,"slug":1972,"title":1973,"excerpt":1974,"featuredImage":14,"publishedAt":1975},"402","ll-amiibo-compatible-games-complete-overview-of-bonuses-and-gameplay-integration","所有兼容amiibo的游戏——奖励与游戏玩法整合完整概览","amiibo手办可与Wii U、任天堂3DS及任天堂Switch平台上的众多任天堂游戏互动。不同游戏中的附加价值各有差异：有时奖励仅为外观装饰，有时则会直接影响游戏玩法。以下将按游戏平台分类，系统梳理兼容作品及扫描手办可解锁的核心内容。","2026-02-27T18:13:00.000Z",{"id":1977,"slug":1978,"title":1979,"excerpt":1980,"featuredImage":1981,"publishedAt":1982},"433","pre-order-alert-good-smile-company-figma-doom-the-dark-ages-doom-slayer-dx-edition","Pre-Order Alert: Good Smile Company Figma Doom: The Dark Ages – Doom Slayer DX Edition","The new Good Smile Company Figma Doom: The Dark Ages Doom Slayer DX Edition is more than a routine figure drop. It connects collector demand, franchise identity, and the wider appeal of Doom as one of gaming’s most durable icons.","\u002Fuploads\u002F2026\u002F05\u002Fpre-order-alert-good-smile-company-figma-doom-the-dark-ages-doom-slayer-dx-edition-1779227794098-251pwj.webp","2026-04-22T04:37:00.000Z",{"id":1984,"slug":1985,"title":1986,"excerpt":1987,"featuredImage":14,"publishedAt":1988},"399","beyond-scanning-creative-uses-and-modifications-of-amiibo","\"超越扫描——amiibo的创意用途与改造\"","amiibo是任天堂自2014年起推出的搭载NFC技术的角色手办。从技术层面看，它们是封装在塑胶模型内的微型数据载体。在实际应用中，这些产品介于玩具、收藏品与交互设备之间。将其扫描接入游戏主机仅是它们生命周期中的一环。多年观察可见，许多拥有者将其视为具备独立创作潜能的实体物件。","2026-02-27T16:36:00.000Z","fallback",[],[]]