3D NAND flash, layers, performance, density and evolution

The Boss

PC

3D NAND flash, layers, performance, density and evolution

density, Evolution, Flash, layers, NAND, performance

The first 3D NAND chip was released in 2013, it was Samsung’s first V-NAND with a capacity of 16 GB and made up of 24 layers. Over time, the number of layers has evolved and today we can find 176-layer 3D NAND configurations. Is there a limit to the number of layers in 3D NAND?

What is NAND flash memory?

NAND Flash

NAND Flash memory is a type of non-volatile RAM, it means it retains its information after losing power. Like RAM, it is organized by memory chips, but unlike RAM, it has lower bandwidth and higher latency.

It is therefore a storage memory that is used like hard disks, which, like RAM, grows from year to year thanks to the effects of Moore’s Law. Something that is accelerated by the deployment of 3D NAND memory. Which uses dozens of stacked 2D memory chips, making the storage capacity many times that of RAM.

Why is 3D NAND needed?

SK-Hynix-96-Layer-QLC-4D-NAND-Flash-3

The main difference between classic 2D NAND memory and 3D NAND memory is in the way the cells communicate. In 2D NAND, the memory cells where the bits are stored lie with each other on a single chip. Thus the capacity of each chip is determined on the one hand by the type of cell used and on the other hand by the number of cells that can be placed on a chip.

For these reasons, 2D NAND is limited by two factors. The first of these is the best known and concerns the capabilities of the manufacturing node used to manufacture the memory. That is to say the number of transistors per mm2. On the other hand, the second reason is less understood, since it relates to the reduction of the distance between the transistors and the electric fields which are created.

3D NAND organization

In other words, the closer the cells are, the more danger there is that the electrical charge of one cell ends up corrupting the data of a neighboring cell. Hence the need to implement 3D NAND systems based on the vertical intercommunication of several 2D NAND chips. The idea is to reduce the cell density in each of the battery chips to prevent the electrical charge in one of the cells from destroying the contents of the others.

It is precisely this fact that makes the 2D NAND chips used to build 3D NAND from a less advanced manufacturing node and therefore with a lower density in terms of the number of transistors than those which are purely two-dimensional.

How many layers does 3D NAND memory support?

3D NAND

3D NAND memory, however, differs from 3D RAM like HBM in the way the memory stack is organized, because starting with the memory chip at the base of the stack, the rest of the chips on top are smaller. size, until the point is reached where the 2D NAND memory chips that make up the stack are too small to scale any further. It must be taken into account as we explained in the previous section that the use of new nodes and the increase in cell density forces certain levels of storage capacity to increase the number of layers in order to avoid contamination of the power supply between the cells.

On average, manufacturers add 30-50% more layers with each generation of 3D NAND. Currently, 96-layer 3D NAND memories are the bestseller with more than half of the market share, followed by 128-layer memories with 15%. Generational evolution? On average every 18 months, it will not be until the end of 2022 that we will see the next generation of 3D NAND memory with 256 layers.

Increased capacity thanks to QLC and PLC cells

qlc_nand_density

Powerline cells are the least reliable of all, allowing storage of up to 4 bits per cell but limiting the product lifecycle. However, the need for storage capacity for many people has increased the demand for QLC cells, which in 2021 was 15% of the demand, but is expected to reach 50% by 2025.

The process should be repeated with the 5-bit cells or PLC that Intel submitted as a proposal a few months ago. Ultimately, this is the same process that has happened before with TLC and later with QLC. Although predictions are made from the use of QLC cells and technological advancements for 2025.

What situation will we be in? By 2025, we are expecting to be able to hit the cost of 2 cents per gigabyte of storage. This translates into a 20% reduction in costs over one year on average and therefore 30% between generations. By 2025, we should expect 3D NAND setups with 384 layers.

When will 3D NAND replace hard drives?

SSD hard drive manufacturers

Conventional hard drives are superior to NAND Flash memory in one point, their ability to store data. It is true that one can make a list of things in which SSDs are currently superior, such as the fact that they do not depend on mechanical parts, have faster seek times and, most importantly, more bandwidth. important.

The question is, in what year will the cost of 3D NAND be as cheap as a hard drive? Well, the predictions point to 2033, so it’s still over a decade for hard drives to outperform 3D NAND memory.

The reason we’re going to have to wait so long is that the cost of the fabrication nodes that are going to be deployed in the coming years is going to be more and more expensive, which is going to be a factor in that regard. However, that doesn’t mean we’ll continue to use hard drives for years to come, especially as SSDs grow in popularity and have benefits beyond storage capacity.

Leave a Comment