AMD 拉上韩国的 FuriosaAI 和 MangoBoost,要把 CPU、GPU、NPU、DPU 拼进一个集群,看算力怎么搭配更省
#FuriosaAI
The talks show that LLM serving bottlenecks sit as much in memory architecture — CXL pooling, HBM bandwidth — as in GPU compute, a framing many benchmark writeups skip. Engineers building inference infrastructure can study the roofline analysis and LMCache MP Mode patterns for their own deployments. The event also signals a growing Korean accelerator and serving stack around FuriosaAI's RNGD and SK hynix memory.
韩国的 FuriosaAI 拿下 6.2 亿美元融资,自称 RNGD 芯片能效跑赢 Nvidia RTX Pro 6000,又一家想分 Nvidia 蛋糕的。
韩国这家做 AI 芯片的公司拿了 6.2 亿美元,估值涨了近 4 倍,专门冲着省电的推理芯片去打英伟达,连三星和 LG 都在用它的 RNGD。
FuriosaAI 晒出第三代推理芯片:32 PFLOPS 算力、48TB/s 带宽,做推理部署的可以看看规格。
FuriosaAI拿下瑞典数据中心7000块RNGD加速器大单,2027年上线,还跟英伟达AMD的GPU混着用。
FuriosaAI和超微搞了个16卡服务器托盘,双层主板设计,用Astera芯片把NIC和Switch比例提到16:1,能减少数据流动,效率更高。