大家在 Qwen3.6 27b 上的推理速度如何？

I'm getting ~13 tps on Q8_0, with a context window of 128000, K Q8_0, V Q8_0 this is on 3x GPUS (1x2060super 8gb, 2x5060ti 16gb), via llamacpp unsure if this is slow or to be expected? */llama-server --port 8080 --model */llama.cpp/Qwen3.6-27B-Q8_0/Qwen3.6-27B-Q8_0.gguf -mm */Qwen3.6-27B-Q8_0/mmproj-BF16.gguf -np 1 --temperature 1.0 --top-p 0.95 --top-k 20 --min-p 0.0 --presence-penalty 1.5 --repeat-penalty 1.0 --chat-template-kwargs '{"preserve_thinking": true}' --cache-type-k q8_0 --cache-t