V100 32G SXM2 adapted to PCIE. Running llamacpp with Q4KM Qwen 3.6 27B or Gemma 4 31B. I use them when I feel of privacy is important or I just want to mess around.