Workflow
GPU架构
icon
Search documents
摩尔线程,重大发布!
Zheng Quan Shi Bao· 2025-12-20 07:54
摩尔线程最新发布。 12月20日,摩尔线程董事长兼CEO张建中在摩尔线程首届MUSA开发者大会上发布新一代GPU架构——"花港"。据介绍,"花港"基于新一代指令集,算力 密度提升50%,能效提升10倍,可支持十万卡以上规模智算集群。 此前,摩尔线程曾发布预告称,12月20日至21日,摩尔线程首届MUSA开发者大会(MUSA Developer Conference)将在北京中关村国际创新中心举办。 张建中将首次系统阐述以MUSA(Meta-computing Unified System Architecture)为核心的全栈发展战略与未来愿景,并重磅发布新一代GPU架构、推出涵 盖产品体系、核心技术及行业解决方案的完整布局,分享多领域落地案例与生态建设进展。 摩尔线程还分享了其图形进化与前沿探索:实现硬件级光线追踪加速与自研AI生成式渲染技术,具身智能、科学智能(AI4S)、AI for 6G等前沿领域的 深度布局。 摩尔线程主要从事GPU及相关产品的研发、设计和销售,被市场视为"中国版英伟达"。12月5日,摩尔线程顶着"国产GPU第一股"称号登陆A股市场,上市 首日就大涨超400%,此后股价更是一路上涨, ...
摩尔线程,重大发布!
证券时报· 2025-12-20 07:50
摩尔线程最新发布。 12月20日,摩尔线程董事长兼CEO张建中在摩尔线程首届MUSA开发者大会上发布新一代GPU架构——"花港"。 据介绍,"花港"基于新一代指令集,算力 密度提升50%,能效提升10倍,可支持十万卡以上规模智算集群。 此前,摩尔线程曾发布预告称,12月20日至21日,摩尔线程首届MUSA开发者大会(MUSA Developer Conference)将在北京中关村国际创新中心举办。 张建中将首次系统阐述以MUSA(Meta-computing Unified System Architecture)为核心的全栈发展战略与未来愿景,并重磅发布新一代GPU架构、推 出涵盖产品体系、核心技术及行业解决方案的完整布局,分享多领域落地案例与生态建设进展。 在今天的演讲中,张建中表示,摩尔线程全功能GPU主要包括四大功能引擎——AI计算加速引擎、图形渲染引擎、物理仿真和科学计算引擎、超高清视频编 解码引擎。类似于英伟达的CUDA,摩尔线程的MUSA是支撑这些引擎的系统架构。 据悉,摩尔线程基本上一年迭代一代产品,MUSA架构已经从2022年的"苏堤"、2023年的"春晓"、2024年的"曲院",进化到2 ...
摩尔线程发布新一代GPU架构 能效提升10倍
(文章来源:上海证券报) 人民财讯12月20日电,摩尔线程董事长兼CEO张建中12月20日在摩尔线程首届MUSA开发者大会上发布 新一代GPU架构——"花港"。 据介绍,"花港"基于新一代指令集,算力密度提升50%,能效提升10倍,可支持十万卡以上规模智算集 群。 ...
摩尔线程发布第五代GPU架构“花港”:能效可提升十倍 明年量产
Mei Ri Jing Ji Xin Wen· 2025-12-20 04:21
Group 1 - The core focus of the news is the announcement by Moer Thread, a leading domestic GPU company, regarding the production of its fifth-generation GPU architecture "Huagang" set to begin next year [2][4] - Moer Thread's CEO Zhang Jianzhong introduced several products based on the "Huagang" architecture, including the AI training and inference product "Huashan," the high-performance graphics rendering product "Lushan," and the "Changjiang" intelligent SoC chip [2] - Moer Thread differentiates itself by offering a full-featured GPU that integrates AI chips, GPGPU, and graphics chips, while being compatible with NVIDIA's ecosystem (CUDA) [4] Group 2 - The "Huagang" architecture features a new instruction set, with a 50% increase in computing density and a 10-fold improvement in energy efficiency, capable of supporting intelligent computing clusters of over 100,000 cards [4] - The GPU is equipped with the first-generation AI generative rendering architecture (AGR) and a second-generation ray tracing hardware acceleration engine, supporting DirectX 12 Ultimate, which is considered the gold standard for next-generation gaming [4]
摩尔线程发布新一代GPU架构花港,算力密度提升50%
Xin Lang Ke Ji· 2025-12-20 04:14
Core Viewpoint - The article highlights the launch of the new GPU architecture "Huagang" by Moore Threads, which promises significant advancements in computing power and efficiency [1] Group 1: New GPU Architecture - Moore Threads introduced the "Huagang" architecture at the 2025 MUSA Developer Conference, featuring a new instruction set and support for asynchronous programming models [1] - The architecture boasts a 50% increase in computing power density and a 10-fold improvement in energy efficiency [1] Group 2: Chip Production and Performance - The "Huagang" architecture will support large-scale intelligent computing clusters with over 100,000 cards [1] - The company is set to launch a series of chips based on this architecture, named "Huashan" and "Lushan," which will soon enter mass production [1] - The "Lushan" chip is reported to enhance AI performance by 64 times [1]
摩尔线程发布新一代GPU架构,能效提升10倍
转自:证券时报 据介绍,"花港"基于新一代指令集,算力密度提升50%,能效提升10倍,可支持十万卡以上规模智算集 群。 人民财讯12月20日电,摩尔线程董事长兼CEO张建中12月20日在摩尔线程首届MUSA开发者大会上发布 新一代GPU架构——"花港"。 ...
摩尔线程发布第五代GPU架构“花港”:能效可提升十倍,明年量产
Mei Ri Jing Ji Xin Wen· 2025-12-20 03:54
Group 1 - The core focus of the news is the announcement by Moore Threads, a leading domestic GPU company, regarding the production of its fifth-generation GPU architecture "Huagang" set to begin next year [1][4] - Moore Threads held its first MUSA Developer Conference in Beijing, where CEO Zhang Jianzhong highlighted the company's advancements and product launches [1] - The company has introduced several products based on the "Huagang" architecture, including the AI training and inference product "Huashan," the high-performance graphics rendering product "Lushan," and the "Yangtze" intelligent SoC chip [1] Group 2 - Moore Threads differentiates itself by focusing on full-featured GPUs that integrate AI chips, GPGPU, and graphics chips, while also being compatible with NVIDIA's ecosystem (CUDA) [4] - Since 2022, the company has been iterating its GPU technology at a rate of one generation per year, having previously launched four generations of products: "Sudi," "Chunxiao," "Quyuan," and "Pinghu" [4] - The "Huagang" architecture features a new instruction set with a 50% increase in computing density and a 10-fold improvement in energy efficiency, supporting large-scale intelligent computing clusters [4]
摩尔线程新一代GPU架构“花港”发布:能效提升10倍,算力密度提升50%
Xin Lang Cai Jing· 2025-12-20 03:43
新浪科技讯 12月20日上午消息,今日举办的摩尔线程2025MUSA开发者大会上,摩尔线程创始人、董 事长兼首席执行官张建中发布了新一代GPU架构"花港"。据悉,该架构将采用全新一代的指令集,支持 异步编程模型和高效的线程同比;算力密度将提升50%,能效提升10倍。 此外,"花港"还支持支持十万卡以上规模智算集群,为了强强算力利用率,该芯片还发明了新一代的异 步编程模型。基于摩尔线程最新一代GPU架构"花港"的系列芯片——华山、庐山,也将很快量产上市。 (文猛) 责任编辑:尉旖涵 新浪科技讯 12月20日上午消息,今日举办的摩尔线程2025MUSA开发者大会上,摩尔线程创始人、董 事长兼首席执行官张建中发布了新一代GPU架构"花港"。据悉,该架构将采用全新一代的指令集,支持 异步编程模型和高效的线程同比;算力密度将提升50%,能效提升10倍。 此外,"花港"还支持支持十万卡以上规模智算集群,为了强强算力利用率,该芯片还发明了新一代的异 步编程模型。基于摩尔线程最新一代GPU架构"花港"的系列芯片——华山、庐山,也将很快量产上市。 (文猛) 责任编辑:尉旖涵 ...
摩尔线程,发布新一代GPU架构
财联社· 2025-12-20 03:35
摩尔线程今日发布新一代GPU架构花港,具备新一代指令集, 算力密度提升50%,能效提升10倍,集成全精度端到端加速技术,支持10万 卡以上规模智算集群, 搭载第一代AI生成式渲染架构和第二代光线追踪硬件加速引擎。 ...
摩尔线程发布“庐山”GPU芯片,AI性能提升64倍
Xin Lang Cai Jing· 2025-12-20 03:10
新浪科技讯 12月20日上午消息,今日举办的摩尔线程2025MUSA开发者大会上,摩尔线程创始人、董 事长兼首席执行官张建中宣布,基于摩尔线程最新一代GPU架构"花港"的系列芯片——华山、庐山,将 于明年量产上市。 据张建中介绍,"花港"将采用全新一代的指令集,支持异步编程模型和高效的线程同比;同时,算力密 度将提升50%,能效提升10倍。此外,"花港"还支持支持十万卡以上规模智算集群,为了强强算力利用 率,该芯片还发明了新一代的异步编程模型。 基于花港架构的"庐山"高性能图形渲染芯片,将实现3A游戏渲染15倍的提升,AI性能提升64倍,光线 追踪性能提升50倍。除支持游戏体验外,还支持所有CAD、CAE等图形设计渲染。 此外,基于该架构的GPU芯片"华山",在浮点算力、访存带宽、访存容量和高速互联带宽方面,取得了 多项领先甚至超越国际主流芯片的能力。(文猛) 责任编辑:宋雅芳 新浪科技讯 12月20日上午消息,今日举办的摩尔线程2025MUSA开发者大会上,摩尔线程创始人、董 事长兼首席执行官张建中宣布,基于摩尔线程最新一代GPU架构"花港"的系列芯片——华山、庐山,将 于明年量产上市。 据张建中介绍,"花 ...