>>103674078
Yes. That's what the --cpu-moe or -n-cpu-moe flag does in llama.cpp. Otherwise, you have to manually define it and that is a pain from personal experience. See pic related.
>>103675641
MOE models are more dependent on CPU and memory speed than GPU at these high speeds but a 1060 is old enough that I will say yes, it is liable to just melt down and do what you said.