3 ms·My $600 Mac Mini Runs a 35B AI Model4 points by danebalia 7mo agobigyabai 7mo ago> The 35B Trick (Your SSD Is the New GPU Memory) Wave "bye bye" to your write cycles.RobMurray 7mo agowhy? it's mostly reads. the weights are static.bigyabai 7mo agollama-cpp's process is, but macOS itself will swap hard when 10-14gb of memory is paged for LLM inference. Dense models especially would thrash zram.