CMP170HX mining cards tested for local LLM inference

A Reddit tester put four 8GB CMP170HX mining cards through local LLM inference, finding that many models fit entirely on a single card and several small models can run simultaneously as long as their combined VRAM allows.