[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"topic-en-cpasync-on-ampere-hide-hbm-latency-on-a100":3},{"cluster":4,"timeline":16},{"id":5,"slug":6,"title":7,"pinned":8,"status":9,"summary":10,"category":11,"language":12,"created_at":13,"merged_into":14,"article_count":15,"first_seen_at":13,"last_updated_at":13},"7d8283d2-69d2-4ab3-8687-b47467f678bd","cpasync-on-ampere-hide-hbm-latency-on-a100","cp.async on Ampere: Hide HBM Latency on A100",false,"active","Ampere’s cp.async moves data without stalling warps, cutting HBM waits from 450–600 cycles into overlapped compute on A100.","research","en","2026-07-05T09:02:07.867476+00:00",null,1,[17],{"id":18,"slug":19,"title":7,"summary":10,"category":11,"image_url":20,"cover_image":20,"published_at":21,"is_canonical_seed":22},"68bfa04a-94c4-4c8a-921c-61e93ab207aa","cuda-cp-async-ampere-hbm-latency-en","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775167612143-4qvu.png","2026-04-02T22:06:36.479+00:00",true]