SambaNova 在超算大会展示 Intel GPU 与 SambaRack SN50 组合推理部署
SambaNova 在 SC 大会现场跑了个分离式推理 demo:Intel GPU 管预填充,自家 SambaRack SN50 管解码,还全是风冷。做推理部署、纠结液冷成本的工程师可以去看看他们怎么拆分 prefill 和 decode 的。
SambaNova 在 SC25 超级计算大会 2246 号展位现场演示了一套分离式推理方案:@Intel GPU D-Series 负责预填充(prefill),SambaRack SN50 负责解码(decode),整套系统采用全风冷散热,无需液冷基础设施。
这套架构把预填充和解码两个阶段拆到不同硬件上运行,Intel GPU D-Series 处理 prompt 的并行计算,SambaRack SN50 专门负责逐 token 生成,是 SambaNova 在大会现场跑起来的实际 demo,而非概念展示。
Things are poppin' off at Booth 2246.
The demo we blogged about yesterday is running live: @Intel GPU D-Series on prefill, SambaRack SN50 on decode, all air-cooled. Come see it at @supercomputing! https://t.co/GoCFpFeuOI
This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include SambaNova, Intel, SambaRack SN50, AI inference.
Source attribution: curated from SambaNova (@SambaNovaAI).