Core Insights - Anfu Technology (603031.SH) has introduced FP8, which stands for "Float Point 8-bit," offering significant advantages over traditional FP16 and FP32 formats in terms of storage and computational efficiency [1] Group 1: Technology Advancements - FP8 reduces memory usage for model weights and activation values by half during large model training, allowing for more data to be cached in the same size on-chip storage [1] - The computational speed of FP8 is enhanced by 2-3 times compared to FP16, leading to improved overall efficiency [1] Group 2: Product Development - The upcoming Fuxi architecture B0 chip from Xiangdi Xian is designed as a heterogeneous chip for AIPC, supporting FP8 operations [1]