2× Radeon R9700 — Qwen 3.6 27B Q8 MTP on llama.cpp

User shares multi-GPU setup with 2× Gigabyte Radeon R9700 GPUs on a ThinkStation P7, running Qwen 3.6 27B at Q8 quantization with multi-token prediction via llama.cpp. Post includes performance benchmarks and server log numbers for reference.