I went with M5 Max at 128GB -- but I run local LLMs, I'm deeply disorganized on web browsing, and I multitask excessively, and the 32GB on my M1 Max turned out to be a problem almost from the start even as the processor tended to keep up. I'm pleased to say that in baseline use -- let's say big Photo directory merges, multi-tabbed web browsing, general multitasking, MS OneDrive getting in the way, Teams calls -- the M5 Max manages to stay at around 35-40C without running the fans at all. The Teams part of that is particularly striking because any sort of video conferencing quickly stressed the M1. For AI, of course, the fans come on but the other day with Llama 3.1 70B and roughly a 50,000 word document that I was norming and a large context window, the fans never really got past 3200rpm and the temperature never really got past 70C. I'd say the average of that particular model running process over a two hour period was about 55C and with fan use ranging anywhere from about 300rpm to 3200. And the keyboard never got more than lukewarm. It is astonishingly cooler running than my M1 Max and this has really surprised me in view of how aggressively the M5 can draw power (as I discovered when trying to run LM Studio, which is either violently incompatible with the M5 Max or I had something horribly misconfigured or both; fans at 6000rpm and 97C and an abysmal token rate -- that's the point at which I shifted to Ollama and Terminal for the task I just described). Note also that while I was running Llama 70B it was also synchronizing my Photos directory and uploading thousands of images I'd recovered off of a backup I thought I had lost, and it really didn't slow down on that Photo upload during the AI work.
The M1 Max by the way does fine with throwing 45MP raw photo files around and the like. My sense for my own purposes is that without the AI use, a 48GB Pro would solve the multitasking bottlenecking. But throwing in recoding video files and AI and that tipped the balance up to an M5, and a flash sale lured me up to the configuration I got.
What strikes me about your memory usage is on the one hand that huge requirement for the Windows VM which is obviously having to be manipulated quite a bit given 24GB and showing clear signs of stress on memory from your Activity Monitor, but on the other hand very little pressure from elsewhere compared to what I'm used to. My hunch is that you'd probably be well served by a 48 to 64GB Pro, but that you should ask around as well because the poster above with the 96GB Max is showing a far smaller allocation to the VM than you are -- which tells me the larger memory allocations are taking strain off the process to the point the computer is likely running cooler, quieter and more efficiently.
