The included images are a bit confusing. It would have been better to link to the video explaining this technology on the source site, or if that wasn't possible to screenshot a some screenshots from that video to add as illustration for how the technology works.

I am guessing here, but this only works for low voltage slow storage because stacking memory modules without enough cooling path would cook some of the center layers, just like the AMD 3D cache lowers the heat transfer away from the CCD. GPU memory would be the hardest chips to layer, desktop memory a little maybe but still. Maybe slow but many channel server memory could be an option, just like AMD epyq 12 channel DDR4 4800 is ridiculously fast compared to even dual channel 10.000 memory. It could be 24 small single chip memory sticks that is cooled by high airflow anyway.

This is pretty awesome, assuming they make it work well. Imagine having 1 TB of RAM in your computer, without it costing three arms, five legs and a couple of kidneys. (and software designed to take advantage of so much fast storage)

I would be skeptical of these claims; They have never produced anything and are making, frankly, outlandish claims. You're telling me this no-name company with no proven track record was able to achieve something that companies like Intel, Samsung, Micron, Hynix, etc. have not figured out in their decades of research into volatile memory?

For some perspective on the NEO Semiconductor density claims, Samsung's current highest density DDR5 DRAM ICs are 16Gb in capacity. The South Korean tech firm, an important player in both NAND and DRAM, is expected to launch 32Gb ICs imminently. Combined with chip stacking, that should enable 1TB memory modules in late 2023 / early 2024. That's a lot earlier than NEO's 2030 launch timeframe, and chip stacking for DRAM could go even higher.

-source The longer I chew on this article and do research, the more these guys seem like patent trolls and/or some sort of investment scam.

wavetrex:

This is pretty awesome, assuming they make it work well. Imagine having 1 TB of RAM in your computer, without it costing three arms, five legs and a couple of kidneys. (and software designed to take advantage of so much fast storage)

lol @ 1 TB of system ram when my pc fails at even using 32 Gigs.

TLD LARS:

I am guessing here, but this only works for low voltage slow storage because stacking memory modules without enough cooling path would cook some of the center layers, just like the AMD 3D cache lowers the heat transfer away from the CCD. GPU memory would be the hardest chips to layer, desktop memory a little maybe but still. Maybe slow but many channel server memory could be an option, just like AMD epyq 12 channel DDR4 4800 is ridiculously fast compared to even dual channel 10.000 memory. It could be 24 small single chip memory sticks that is cooled by high airflow anyway.

This architecture picture better shows how the stacking is supposed to work. Now whether this increases the possibility of heat dissipation or not is not for me to say (not a hardware engineer), but the vertical channels through the layers might help. https://neosemic.com/wp-content/uploads/2023/04/bitline.png