$400 Qwen 3.6-27B Setup - Dual RTX 3060 - 30-50 t/s

In three lines$400 budget setup with dual RTX 3060 (24GB total) running Qwen 3.6-27B. Decode speed 30-50 t/s on llama.cpp with Q4_K_S quantization. Legacy i7-4770K platform with PCIe 3.0 x8 dual support, performance-equivalent to modern boards. Limitation: tensor parallel disables KV cache quantization, context capped at 64k.

Summary generated by Claude — human-verified