The RAM in your desktop could run at much higher frequencies if it were soldered. The RAM the M5 Macbooks use runs at 9600 mbps. The base spec for DDR5 is 4800, and I rarely see people running at much higher than 6000. There's probably someone who's gotten a 9600 mbps overclock running on their desktop, but it's unatainable for a normal consumer outside of machines with soldered RAM.
"base spec" is doing the heavy lifting here, not "soldered". DDR5 modules supporting 8000 MT/s CMP/EXPO profiles are available. 9000 is available with CUDIMM modules.
The problem isn't the RAM chips. The problem is their path to the CPU.
Good luck finding a motherboard which will run at those speeds reliably. Sure, you can find a board which says it will do it, and it will sometimes manage those speeds, if you also have a CPU with a memory controller that can handle all the interference. When you solder the chips and thus make the traces much shorter and less susceptible to interference, that problem largely goes away.
Because GPUs have had it soldered since the beginning, while desktops have had it slotted since (close to) the beginning. There are some exceptions to that rule, but it mostly holds.
Not to mention that the bar of acceptibility is moving. We had RAM in slots in laptops for a long while, but it's (not so) slowly going away. You can still find laptops with SO-DIMMs, but it's not a given anymore.
That would be COAST as in 'cache on a stick', which wasn't that common.
They usally came in 8 to 16 discrete chips with pins on both of their sides, to be soldered on board, or plugged into sockets.
Plus one or two 'tag-rams' as chips of the same size.
All depending on the chipset, its cache-controller, and which capabilities the board-maker decided to use/expose/support. Which sometimes the BIOS would get wrong, still :-)
Yep. Soldered is just the best manufacturing process (besides on-package). Moving things close together simply delivers better outcomes and that will always trump openness concerns.
I would imagine that something like LPCAMM2 for GDDR is entirely possible and will never, ever be seen because people already accept replacing an entire $10000 GPU to get another 32GB VRAM on the next generation, so why give up that guaranteed revenue stream?
