11月11日,华为通过麒麟官方公众号正式宣布麒麟9050系列芯片将搭载新一代达芬奇架构NPU 1。该系列芯片是首款采用“逻辑折叠”技术的τ芯片,其NPU性能提升140%,同时功耗降低66% 1。这一架构升级使得业界首个支持30B MoE大模型在端侧离线运行的技术成为现实,用户无需联网或依赖云端即可使用相关功能 1。
除了核心的AI算力突破,麒麟9050系列还落地了多项具体应用场景,包括AI屏幕朗读、Xmage智拍2.0、AI眼动翻页以及熄屏地图2.0等 1。华为半导体业务无线终端芯片产品管理部部长徐松余表示,该系列芯片旨在开启未来十年的技术新征程,并设定了在2031年实现P核频率突破5GHz的目标 1。
Huawei officially announced the launch of its Kirin 9050 series chips on November 11 1. These new processors feature a next-generation Da Vinci architecture NPU, which represents a significant leap in efficiency and performance for Huawei's semiconductor business 1. As the first "τ chip" to adopt "logic folding" technology, the Kirin 9050 series is positioned to kick off a new decade of technical advancement for the company 1.
A key highlight of the announcement is that the Kirin 9050 series supports the industry's first end-side offline operation of a 30 billion parameter Mixture of Experts (MoE) large model 1. This capability allows the model to run locally without requiring an internet connection or cloud support 1. The upgraded NPU delivers a 140% increase in performance while simultaneously reducing power consumption by 66% 1.
The integration of this advanced AI capability enables several specific user-facing features, including AI screen reading, Xmage Smart Photography 2.0, AI eye-tracking page turning, and Off-screen Map 2.0 1. Looking ahead, Xu Songyu, head of wireless terminal chip product management at Huawei Semiconductor, stated that the company aims to break through a P-core frequency of 5GHz by 2031 1.
评论
还没有评论,欢迎留下第一条。