8 GPUs in one box. Four Maxsun dual B60 cards in the lab, 192GB total VRAM. Running Intel Battle Matrix with the LLM Scaler stack for multi-GPU vLLM, pipeline or tensor parallel, speculative decoding, and smart quantization. #AI #LocalInference #LLM #IntelArc #BattleMatrix #vLLM #WorkstationAI #GPUs #OnPrem #RAG #LabLife