@mxk @cstross @ela I disagree to a huge extent, as #arm64 is inherently more efficient and the gains on paper and benchmarks are not even remotely as relevant to daily use and long-term sustainability, rather #PowerSavings are more relevant and #LPDDR5 exists, which eats less than #LPDDR5X / #GDDR7X / #DDR5 RAM.

- That's literally why you don't see #HBM even on Gaming & Workstation GPUs anymore where the low latency & high bandwith would justify the costs.

- Plus GPUs for the most part come with one VRAM config. Not a dozen different ones...

The integration density of soldered-down RAM may be justified in phones and tiny embedded devices where there's no may to put a SODIMM inside the chassis - not even at the edge of the mainboard.

- But it's not as if the #iPad or any #MacBook were as tightly packed as an #iPod and certainly the #MacbookNeo and Desktop-#Macs don't have to deal with fitting a pants-pocket compatible form factor and having no meaningful way to dissipate heat at an extremely low delta-t° of 22° or less (at 20°C ambient and a maximum 42°C body contact temperature, not accounting for human body heat!)…

It's a DELIBERATE CHOICE by #Apple to prioritize #AntiRepairDesign.