[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-nsight-compute":3},{"tag":4,"articles":10},{"id":5,"name":6,"slug":7,"article_count":8,"description_zh":9,"description_en":9},"81d372a3-1278-4e24-aa71-4d307d830ce0","Nsight Compute","nsight-compute",1,null,[11],{"id":12,"slug":13,"title":14,"summary":15,"category":16,"image_url":17,"cover_image":17,"language":18,"created_at":19},"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","zh","2026-04-02T22:06:36.022671+00:00"]