Previous | Next --- Slide 31 of 47
Back to Lecture Thumbnails
jkorn

Traditionally, we only have data pins connected on the perimeter of each chip, but now with this 3D stack, we have a much larger surface area to connect pins on the bottom, so we can transfer more data at once.

apadwekar

By stacking the DRAM, we also bring the memory closer to the processor likely lowering the memory latency.

pagerank

I am concerned that the heart may become a more important issue in 3D stacking architecture. Because now the layers in the middle cannot be sent the heart into the air any more and it has to transfer the heart to other layers. Maybe they are using some magic materials to solve this problem.

vasua

This stacked DRAM technology has actually hit the market, if in an expensive way. Recently (2014), Samsung introduced VNAND in an SSD form factor. I believe they've also integrated it into their newest phones, or have at the very least demoed such a system at CES.

https://www.extremetech.com/computing/189003-samsung-850-pro-review-3d-nand-and-ram-caching-result-in-the-fastest-most-durable-ssd-money-can-buy

hdd

Intel has introduced 3D-X-Point into the market which is the name of the architecture shown above. They claim that the memory system is 1000 times faster than current memory for small data sizes. However, one additional concern with going 3D is thermal dissipation per unit area. This increases significantly. So in order to reduce that, large heat sinks are kept on top of the memory/cpu architecture.