Moore Threads has unveiled its next generation Huagang GPU architecture, targeting both gaming and AI and signaling how quickly Chinese graphics vendors are trying to catch up with Nvidia and AMD. The new Lushan gaming GPU is said to deliver up to 15x higher AAA gaming performance and 50x faster ray tracing compared to the company’s current MTT S80 and S90 cards, although these are internal claims without public third party benchmarks so far. Support for DirectX 12 Ultimate, second generation hardware ray tracing and a unified rendering plus AI engine is meant to address earlier issues with features, drivers and game compatibility that kept Moore Threads firmly behind Western brands.
On the data center side, the Huashan AI GPU uses chiplets and HBM and is pitched as offering performance somewhere between Nvidia’s Hopper and Blackwell generations, with higher claimed memory bandwidth than B200 and big gains in compute density and efficiency. In practice, China’s best accelerators are still estimated to be roughly 1.5 to 3 years behind the top US GPUs in effective performance and ecosystem maturity, but the gap has clearly narrowed compared with the past decade.
