The domestic GPU market is experiencing explosive growth, and now another domestic manufacturer has announced new achievements.
According to Anfu Technology on their interactive platform, the new generation “Fuxi” architecture chip developed by Xiangdi has completed tape-out verification, demonstrating excellent performance in graphics rendering and parallel computing capabilities.
It is reported that after receiving investment from Anfu Technology, Xiangdi has accelerated the iteration of new products. The new generation Fuxi architecture GPU will adopt a 5nm process, achieving a computing power of 160 TFLOPS (FP32), and integrating 12GB of HBM2 memory.
According to official statements, the new generation “Fuxi” architecture chip developed by Xiangdi is currently in tape-out, and this architecture will bring two new products, both of which meet leading domestic performance and technical specifications.
The Fuxi A0 focuses on filling the gap in high-end domestic rendering products; the Fuxi B0 is a chip that integrates GPU and NPU, targeting edge model deployment and the rapidly emerging AIPC market. The Fuxi B0 will fully support the edge deployment needs of mainstream models such as LLAMA, ChatGLM-6B, Stable-Diffusion, Sora, and DeepSeek R1 1.5B/7B.
