They advertise local LLMs which will be servery limited with 16GD of RAM. Plus the GPU could in theory provide decent gaming performance but again might suffer from the RAM limit.

Most people can totally live with 16gigs but it is kind of a waste for the horsepower. They know what they are doing. Apple is a master in upselling.

Though personally I don't mid the aggressive upsellign as long as the quality is there. Problem is, the hardware quality is great but the software side is severely lacking and getting worse.

If anything, it's less, because you're giving up more RAM to the GPU.

Which, I mean, I love unified memory, as one of those weirdos that does do local LLM stuff and am contemplating if it's time to upgrade my m2 max.

But if you needed 32gb then you still need at least 32gb now. Unless swap on nvme disks is enough for you - and it isn't for me.

RAM is still RAM, the switch from crusty HDDs to fast NVMe SSDs may have helped to smooth things over when you spill into swap but it's not going to do miracles.

RAM isn't just RAM though. Unified memory on Apple Silicon provides significantly better memory management and efficiency compared to trad RAM

Says your rose-tinted marketing-loving glasses. Unified memory only means you lose some to graphics. Only your GPU wins this bargain, not your usual workloads.

Other workloads win because on Apple Silicon the GPU memoy remains accessible by the CPU so there only needs to be one copy of computed resources for the GPU to use.