Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Huawei Ascend 950DT 1,024-card cluster goes live in Wuhan

AI summary

IT之家 AI reported that a 1,024-card Ascend 950DT liquid-cooled AI supernode commissioned by Hubei Science and Technology Investment has gone live in Wuhan. Project representatives said the deployment uses LingQu architecture, provides 1 EFLOPS of FP8 compute and 256 TB of globally unified memory, and is intended to support trillion-parameter model training, industry-agent development and inference; no independent benchmarks were provided.

In-depth insight

The deployment adds a large-scale domestic AI computing installation for model training and agent workloads in Hubei, although its performance claims remain unverified independently.

IT之家 10 月 29 日消息,以“数智荆楚、跃升有为”为主题的华为中国行 2026· 湖北省人工智能创新发展大会昨日(28 日)在武汉举办。

会上,由湖北科投投资建设的昇腾 950 千卡超节点智算集群正式点亮,全国首个昇腾 950DT 1024 卡智算集群在武汉上线。

据介绍,此次落地的昇腾 950DT 1024 卡液冷超节点,单节点搭载 1024 张昇腾 NPU 加速卡,采用计算柜与互联柜全液冷设计,FP8 算力达 1EFLOPS,配备 256TB 全局统一内存,卡间互联带宽达 TB 级,端到端通信微秒级时延;依托灵衢架构,集群内各卡可直接互访,全面支撑万亿参数大模型训练与智能体推理。

湖北科投透露,该集群从项目启动到落地点亮,仅用时 1 个月。

IT之家获悉,该超节点将为湖北省内大模型训练、行业 AI 智能体开发调优与运行提供本地算力支撑,补齐湖北高端算力供给短板。

相关阅读:

#Huawei#Hubei Science and Technology Investment#Ascend 950DT#LingQu architecture#Ascend NPU

Original · 1 min readRead the original
Group chat

No comments yet. Start the conversation.