Thanks, I thought it's talking about data compression. So clueless to not understand what the article talks about, as I was expecting things like hardware accelerated RAM.
CPU packages are starting to have issues with too much or too little pressure from the heatsink mounting to cause pin contact issues too. AMD Threadripper has a $30 torque wrench that is suggested to secure the package.
I took a look at the standard, and it's a 7x44 connector for each channel of memory. That gives 308 connection points, which is slightly more than 288 on a DIMM. Dual-channel modules (as shown in the article) simply double the number of pins to a 14x44 connector (616 pins)
I'm a big fan of the idea of CAMM in general. It's this, or we end up with locked-in amounts of RAM as per things like the M[1-3] Macbooks. The connectors won't be cheap, so I suspect uptake for laptops might not be so common, but this is a big help for desktops going forward.
Thickness seems to be a major selling point here. I wonder why there aren't any board-to-board edge connectors that would allow placing RAM expansion boards next to the motherboard, instead of on top of it. Given how small laptop motherboards are nowadays, it feels like there would be some room for this.
I think that one of the issues with edge-connectors (beyond thickness and distance as mentioned in another comment) is capacitance, which is one reason why CAMM is structured the way it is.
The issue is that the RAM needs to be close to the CPU and the edge that is the closest to the CPU is probably where designers want to put the cooling exhaust and radiator.
That's not always the case. My last three laptops have the RAM hanging off the edge of the mainboard. The SO-DIMM connectors are where the board stops, and there's "nothing" under the modules.
The CPU is actually in the middle of the MB, close to the two edges (top and bottom), but the cooling fan and heatsink are to one side, connected via some heat pipes.
The biggest selling point is signal integrity and routing length. CAMM(2) is designed to be much better at that than DIMMs, which is why if this ends up being successful we will probably see it (or something similar) on desktop and eventually server too.
Whatever room there is left is used for battery/ies today and I would bet money on that if you asked customers whether they wanted RAM-sockets back or increased battery life the majority would sadly (?) pick the latter.
This is a great development. It looks to greatly decrease the advantages soldered ram has compared to upgradable SODIMMs in laptops. It both allows for LPDDR5 ram and for slightly thinner designs. Hoping we'll see this in consumer laptops soon and for prices to get to about what SODIMMs cost right now maybe a year later?
I found the criticism of LTT to be unfounded beyond the mishap with the prototype cooler.
I'm more skeptical of them now but until I have good reason to believe they haven't followed through with their promises of improving their internal processes, I'll continue watching their videos.
[1] https://www.tomshardware.com/pc-components/ram/new-incredibl...