[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"aihot-art-94733":3},{"itemId":4,"vertical":5,"category":6,"source":7,"score":8,"title":9,"summary":10,"analysis":11,"url":12,"coverUrl":13,"direction":13,"marketSignal":13,"publishedAt":14},"94733","ai","产品发布\u002F更新","IT之家",56,"芯展速AI90固态硬盘显存化方案：首Token延迟缩短至1\u002F50","芯展速推出AI90方案，将SSD纳入显存体系，使首Token延迟缩短50倍，吞吐量提升5.1倍。","芯展速推出AI90固态硬盘显存化方案，将首Token延迟缩短至1\u002F50，为AI推理加速提供了新思路。\n· AI90方案将高性能SSD纳入GPU显存体系，形成HBM+DRAM+SSD三级存储，KV Cache可卸载到SSD。\n· 首Token延迟从秒级降至亚秒级，提速50倍；吞吐量提升5.1倍；显存节省39%。\n· 结合智能P2P多卡互联，8卡RTX 5090集群推理加速可达5.8倍，支持128K+上下文。\n· 配套的PT200Z AI SSD采用pSLC闪存和PCIe Gen5，DWPD高达100，满足高写入负载。\n影响\u002F看点：这一方案通过存储创新缓解了显存瓶颈，有望降低大模型推理的硬件成本，推动AI应用普及。","https:\u002F\u002Fwww.ithome.com\u002F0\u002F978\u002F942.htm",null,"2026-07-20 11:57:36"]