Not to over-hype this storage chip tech, but if I could get away with calling my first-born '3D NAND', I totally would
Let's take a quick tour under the hood – and spell out why your biz needs this
Comment Anyone who’s ever watched the original Star Trek series will probably remember Spock and Kirk playing three-dimensional chess – a great chance for the Enterprise’s science officer to show off his prowess with logic as he contemplated a range of complex moves.
It’s that same complexity that underlines one of the hottest technologies right now, the introduction of 3D NAND, the next, er, logical step-up from today's 2D NAND.
Conventional NAND memory chips are long established as the mainstays of computer storage, however, they are reaching their limitations. We all know the numbers by now. How companies are having to handle the inexorable rise in data – according to IDC, the world will be crunching through 175 zettabytes of data by 2025 – and dealing with the corresponding dizzying growth in storage requirements.
Businesses are all looking to make better use of their data, which is why they are paying closer attention to concepts such as machine learning and artificial intelligence to analyse that information more effectively, all while the Internet of Things has added to the overall data load. Just about everywhere is seeing some significant changes: it’s not just a question of increasing compute power to process all this info, there’s the need for a steep rise in storage capacity, too.
And that’s where 3D NAND comes in. It uses the same underlying technology as 2D NAND though, as the name suggests, it consists of vertical stacks of silicon, all housed within individual chip packages, with the aim of providing more storage capacity at a lower cost.
That's good news for organizations, as adopting the technology may lead to considerable savings. Companies can potentially get more business intelligence for their money.
It’s a technology that’s certainly making its presence felt. Research indicates that 3D NAND is set to dominate the flash memory market. According to figures from Allied Market Research, the global 3D NAND flash memory market size is expected to swell to nearly $100bn by 2025. That’s an annual average rise of 35 per cent from the $9bn of a couple of years ago, with most of the growth coming from the Asia-Pacific market.
While 3D NAND sounds straightforward, the move from 2D NAND introduces a new set of challenges. Piling layers on top of each other increases storage capacity, and introduces plenty of manufacturing difficulties along the way. It’s almost like building a sub-millimetre skyscraper of silicon with all the demands that pulling together such a structure entails.
'Unstoppable' Chinese NAND fabber YMTC to unleash 64-layer flash flood before skipping ahead to 128 – analystREAD MORE
There are several issues at play here but most of them are related to the ways that the layers themselves are made and then put together. Let’s look at what’s inside, and at the challenges.
First, there are all those tiers of stacked silicon film, connected by vertical chimneys that are lined with NAND gates: 64-layer chips are relatively commonplace, 96-layer NANDs were introduced last year, and, in the past couple of months, eggheads have produced 128-layer 3D NAND chips. It’s not certain how many layers are going to be possible, though there are certainly going to be a lot more – manufacturers are already talking about 256 and 512-layer versions.
Stacking itself creates risk. As more layers are placed on top of each other, to form the 3D tower of storage, the chance of defects increases – one minuscule miscalculation or mis-alignment in one level can have a knock-on effect up or down the stack, resulting in serious consequences in terms of overall use and reliability.
For instance, the act of stacking alternating layers of silicon film to form this tiny tower can stress and damage the material. Next, you need connectivity to the gates. Traditionally, tungsten is used for these electrical lines. With multiple layers, the interconnects have to be thin, to keep the design practical, which can lead to high-tensile stress, causing additional distortion.
There are a couple of options here: one is to use low-fluorine tungsten to deposit the metal; the other is to use an entirely different metal – cobalt, as one example, has been used as a replacement here. There’s another conductivity problem, too: as the stacks get higher, there’s an issue in getting the electrons moving through the gate channels. Manufacturers are looking at different approaches to resolving this – one way that’s being investigated is the use of germanium doping to improve conductivity.
The silicon in each layer must be etched. Accurate etching is an issue with all aspects of silicon technology though the multiple layers of 3D NAND complicate matters somewhat. There needs to be consistent degree of etching from top to bottom.
Another characteristic getting in the way of the etching process is the stack's aspect ratio – that is, the relationship between height and width of the pile. The higher the aspect ratio, the more difficult it is to maintain consistency in the etching. Currently, 3D NAND has an aspect ratio of around 60 to one – the ratio between the height and width of a 3D stack – and that’s set to increase as the height of multi-layer stack increases. In the planar NAND world, the more usual aspect ratio is about 15 to one. To put that in perspective: skyscrapers have an aspect ratio of considerably less – the highest slenderness, to the best of our knowledge, is 24 to one.
Why does all this matter? It shows the sort of challenges manufacturers are facing, and it also shows the degree of innovation that’s going on. The deployment of this high-density technology will bring enormous value to enterprises, offering real financial savings, and providing a springboard for future tech innovation.
So, where do we go from here? The manufacturers have their work cut out in bringing rock-solid 3D NAND to the enterprise market in a cost-effective manner – this stuff has got to work reliably in the data center – though we’re already seeing the fruits of those labors. Factories are working to improve the manufacturing process, even as physical size is reduced, and lower the risk factor. It’s also clear we will see NAND prices continue to fall for some time ahead, which is good news for people juggling storage budgets.
It’s all very encouraging. There’s still time to go until 3D NAND becomes the de-facto storage medium, though all the signs are there that the technology is heading in that direction. There’s probably not been a better time for CIOs to consider going down the solid-state route. In other words, here's to more gigabytes for your system-purchasing buck. ®