Consider this: a widely available machine, up to 1.5TB of system RAM, room for 4 passively cooled GPUs with 128GB of VRAM, in desktop or rack format. That machine was released in 2019, then discontinued in favor of one that had only a max of 192GB shared memory. This would be the local LLM machine r
