03:39
2026-10-11
forum.level1techs.com
artificial-intelligence
Running LLMs on a repurposed Kintex-7 PCIe card: openTPU
An open-source project called openTPU is running local LLM inference on a repurposed Inspur YPCB-00338 FPGA accelerator card built around a Kintex-7 xc7k480t with 4 GiB of DDR3 across two channels and…