Oriental Compute Core Debuts DF1000 AI Chip With 6.4 TB/s Memory Bandwidth on July 13
According to Security Times, Oriental Compute Core launched the DF1000 software-defined 3D AI chip in Shanghai on July 13. The chip delivers 520 TFLOPS at BF16 and reaches 6.4 TB/s of memory bandwidth, exceeding Nvidia H200's 4.8 TB/s. It uses DRAM-logic hybrid-bonded 3D packaging to address AI
NVDA-13.84%
GateNews·07-13 23:54
