Moore Threads, a prominent Chinese chipmaker, has announced the launch of its latest GPUs, Lushan and Huashan, revealing significant advancements in both gaming and AI applications. These innovations were unveiled during the MUSA 2025 summit, highlighting the company’s commitment to pushing the boundaries of performance and technology.
Unveiling the Lushan Gaming GPU
Moore Threads introduced the Lushan GPU, designed to revolutionize gaming and content creation. Built on the advanced Flower Harbor (Huagang) architecture, it promises a substantial leap in performance. According to reports, the architecture features an enhanced compute unit, increasing compute density by 50% and improving energy efficiency by 10%. The Lushan GPU boasts an impressive 15x increase in AAA gaming performance and a staggering 50x boost in ray tracing capabilities.
Notably, the architecture is compatible with the latest APIs, including DirectX 12 Ultimate, correcting past shortcomings. The new design also incorporates AI generative rendering under the UniTE unified rendering architecture, alongside a cutting-edge hardware ray-tracing engine, paving the way for advancements in Neural Rendering and Path Tracing.
Huashan AI GPU: A New Frontier
The Huashan GPU targets AI applications, comprising two chiplets with eight HBM sites. Reports indicate its Floating-Point compute performance is comparable to NVIDIA’s Blackwell B200, while exceeding in memory access capacity and matching its bandwidth. The architecture’s asynchronous programming model and ultra-large-scale interconnect support the scaling of over 100,000 GPUs in a cluster using MTLink technology.
Furthermore, the upcoming GPUs will offer quadruple the memory capacity compared to previous models, with up to 64 GB expected in the new solutions. Moore Threads plans to release the Lushan gaming lineup and corresponding AI products by 2026, marking a significant milestone in their technological evolution.
