[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-cuda":3},{"tag":4,"articles":11},{"id":5,"name":6,"slug":7,"article_count":8,"description_zh":9,"description_en":10},"603dae7f-ab7d-4827-a3cb-4abe85e1f058","CUDA","cuda",15,"CUDA 是 NVIDIA GPU 的平行運算平台與程式模型，核心在 SM、warp、shared memory、HBM 延遲隱藏與資料搬移優化。它直接影響 AI 訓練、推論、科學模擬與高效能計算的效能上限。","CUDA is NVIDIA’s parallel computing platform and programming model, centered on SMs, warps, shared memory, and latency hiding with HBM. It shapes performance in AI training, inference, scientific simulation, and other GPU-heavy workloads.",[12,21,28,36,44,51,58,65],{"id":13,"slug":14,"title":15,"summary":16,"category":17,"image_url":18,"cover_image":18,"language":19,"created_at":20},"f9efd9e5-c8e9-4cb1-9f30-443cbdb4d845","cuda-architecture-sms-cores-memory-zh","CUDA 架構怎麼跑：SM、核心、記憶體","CUDA GPU 把工作拆給 SM、數千個核心和分層記憶體。這篇用台灣開發者看得懂的方式，拆開它為何特別適合平行運算。","tools","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775197313894-6e3x.png","zh","2026-04-03T06:21:37.918394+00:00",{"id":22,"slug":23,"title":24,"summary":25,"category":17,"image_url":26,"cover_image":26,"language":19,"created_at":27},"65281366-d5a8-4cae-b397-5c0b839f3e01","nvidia-forum-su7-cuda-lattice-engine-zh","NVIDIA 論壇聊 SU(7) CUDA 晶格引擎","NVIDIA Developer Forums 一篇貼文把 7×7×7 晶格、shared memory、warp 與 bank conflict 放在一起談。重點不是 SU(7) 名字多炫，而是 CUDA 真的吃不吃這套。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775178415223-azaq.png","2026-04-03T01:06:28.438192+00:00",{"id":29,"slug":30,"title":31,"summary":32,"category":33,"image_url":34,"cover_image":34,"language":19,"created_at":35},"d458f7db-1e28-4cf1-9bd8-ad9c95dee997","cuda-cp-async-ampere-hbm-latency-zh","Ampere 的 cp.async 怎麼藏 HBM 延遲","A100 上一次 HBM2e 載入約要 450 到 600 cycles。Ampere 的 cp.async 讓資料直進 shared memory，搭配 pipeline 把等待時間藏進計算裡。","research","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775167621432-n9fo.png","2026-04-02T22:06:36.022671+00:00",{"id":37,"slug":38,"title":39,"summary":40,"category":41,"image_url":42,"cover_image":42,"language":19,"created_at":43},"327cb21a-e255-4225-93de-fa6369880bb0","will-nvidia-stock-recover-after-deepseek-zh","DeepSeek 之後，NVIDIA 股價會回來嗎","DeepSeek 讓市場重新算 AI 成本，但 NVIDIA 仍握有 GPU、CUDA 與資料中心供應鏈。本文用數字、競品與產業脈絡，拆解股價能否回升。","industry","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775161858604-gxgw.png","2026-04-02T20:30:36.102518+00:00",{"id":45,"slug":46,"title":47,"summary":48,"category":17,"image_url":49,"cover_image":49,"language":19,"created_at":50},"e97caa94-b5de-452f-ae23-ac5c2b2854b3","cuda-in-2025-why-gpus-still-win-zh","2025 年 CUDA 為何還是強","CUDA 已經 18 年，卻仍是 NVIDIA GPU 的核心軟體堆疊。從 AI 訓練、氣象模擬到蛋白質計算，這套工具鏈為何還是開發者首選？","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775149438491-u7kw.png","2026-04-02T17:03:37.713589+00:00",{"id":52,"slug":53,"title":54,"summary":55,"category":17,"image_url":56,"cover_image":56,"language":19,"created_at":57},"d233c90c-e7d8-418d-a8dc-f76080f1b968","turboquant-fast-cold-starts-rust-gpu-zh","TurboQuant、冷啟動與 GPU Rust","TurboQuant 把 KV cache 壓到 4.6 倍，GPU state restore 盯上 32B 模型冷啟動，Rust 也更深入 CUDA 開發。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775146380823-5d5u.png","2026-04-02T16:12:38.23896+00:00",{"id":59,"slug":60,"title":61,"summary":62,"category":17,"image_url":63,"cover_image":63,"language":19,"created_at":64},"83e2a967-1919-4771-857f-37fb8d4cfd00","cuda-asinf-accuracy-no-performance-hit-zh","CUDA asinf() 更準，速度沒掉","NVIDIA Developer Forums 上有人替 CUDA 12.8 的 asinf() 做精度優化，指令數仍維持 26 條。這篇看它怎麼在 GPU 數學裡，硬拚準度與效能。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775142948311-udy5.png","2026-04-02T15:15:32.933149+00:00",{"id":66,"slug":67,"title":68,"summary":69,"category":41,"image_url":70,"cover_image":70,"language":19,"created_at":71},"ea6be18b-c903-4e54-97b7-5f7447a612e0","nvidia-gtc-2026-big-ai-announcements-zh","NVIDIA GTC 2026 重點拆解","NVIDIA 在 GTC 2026 一口氣端出 1,000 場 session、2,000 位講者，還把 AI 工廠、推論基礎設施、Agent 平台與實體 AI 全部綁成一套銷售方案。這場大會重點不是單一 GPU，而是從晶片到軟體的整包系統。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1774516049779-pr7v.png","2026-03-26T07:14:26.62638+00:00"]