摩尔线程发布“庐山”GPU芯片,AI性能提升64倍

Core Viewpoint - Moore Threads announced the mass production of its new GPU architecture "Huagang" series chips, "Wushan" and "Lushan," set to launch next year [1][4]. Group 1: Product Features - The "Huagang" architecture will utilize a new generation instruction set, supporting asynchronous programming models and efficient thread concurrency [1][4]. - The computing density of the "Huagang" architecture will increase by 50%, with energy efficiency improving by 10 times [1][4]. - The architecture supports large-scale intelligent computing clusters of over 100,000 cards, featuring a newly invented asynchronous programming model to enhance computing resource utilization [1][4]. Group 2: Performance Enhancements - The "Lushan" high-performance graphics rendering chip will achieve a 15 times improvement in 3A game rendering, a 64 times increase in AI performance, and a 50 times enhancement in ray tracing performance [1][4]. - In addition to gaming, the "Lushan" chip will support all CAD and CAE graphic design rendering applications [1][4]. - The "Wushan" GPU chip has achieved leading capabilities in floating-point computing power, memory bandwidth, memory capacity, and high-speed interconnect bandwidth, surpassing many international mainstream chips [1][4].

Moore Threads Technology-摩尔线程发布“庐山”GPU芯片,AI性能提升64倍 - Reportify