Moore Threads Just Launched GPU With 50x Better Ray Tracing That Could End Nvidia’s GPUs!

Author: Evolving AI - Bewertung: 2x - Views:17

China’s Moore Threads just went from “interesting startup” to a direct threat to Nvidia’s entire GPU stack. In this video, we break down how the company unveiled two new chips — the Lushan gaming GPU and the Huashan AI accelerator — both built on its next-gen Huagang architecture and backed by a full CUDA-style platform called MUSA. Lushan isn’t just an upgrade; Moore Threads is claiming up to 15× higher gaming performance than its old cards and a 50× jump in ray tracing, full DirectX 12 Ultimate support, and massively boosted on-card AI compute, with memory configs that could go all the way to 64 GB — at prices expected to undercut Nvidia’s RTX lineup in China. On the AI side, Huashan is where things get serious. It uses a chiplet design with dual compute dies, eight stacks of high-bandwidth memory, and an MTLink 4.0 interconnect that Moore Threads says can scale training clusters to more than 100,000 GPUs — with compute density up 50%, energy efficiency up to 10×, and memory bandwidth rivaling or even beating Nvidia’s Blackwell B200 in some workloads. All of that is tied together by the MUSA software stack, built to be China’s answer to CUDA, along with a “physical AI” partnership with 51WORLD/51Sim to power massive autonomous driving and world-simulation rigs. This isn’t just about one GPU launch — it’s the opening move in a parallel Chinese GPU and software ecosystem designed to break China's dependence on Nvidia. And if you want the real story behind the world’s fastest-moving tech and AI breakthroughs, make sure to like and subscribe to Evolving AI for daily coverage.