[kube-system]

It depends on the market segment of the device. Many of the segments that ARM devices compete with are also segments where sockets aren't common in x86 either. I mean, we all know Ampere devices typically have socketed RAM, but that's because different compromises were made in those devices vs. the compromises made in a thin low-power laptop.

The thin-and-light segment has some pretty tough competition, and the best sellers are not the devices with the most sockets.

If you're choosing to put an ARM processor in a laptop, you're probably doing so because it helps you accomplish some of the things this market wants: a smaller thermal solution, lower power consumption, slick packaging, etc. Adding sockets doesn't really help you commercially position that particular product any better, it only hurts you.

[ChuckNorris89]

>The thin-and-light segment has some pretty tough competition, and the best sellers are not the devices with the most sockets.

Again, the original comment I addressed was the fallacy that it's technically impossible to have socketed RAM on SoCs, which is false, not whether the device with more sockets is better seller which is probably also false, but again, this was not the main topic and we can continue to discuss forever on the pros and cons of RAM sockets, so, having made my technical point clear, I will end the conversation here.

[kube-system]

I think that's an unfair interpretation of that comment. They most likely meant that you couldn't just lift the DRAM off the package and put it on a standard module, all else equal. It would require other significant design compromises to accomplish that. LPDDR4X socketed modules don't even exist.