Think high capacity SSDs are far off on a distant horizon? Not so. New manufacturing techniques mean that petabyte, high capacity SSDs are probably a question of when, not if.
The world of SSDs – and what we think of as a “high capacity SSD” – is going to change radically over the next few years. That’s because a single SSD will soon be able to store in such vast quantities of data that the very term “high capacity SSD” may end up being wholly inadequate and in need of replacement.
That’s good news for many businesses, because the high capacity SSD market has been quite dull for the last few years and technology advances moderate. The main change has been a move from single level cell (SLC) technology to multi-level cell (MLC) and enterprise multi-level cell (eMLC) technology. That’s made it possible to offer cheap consumer grade SSDs to sit alongside the more expensive enterprise class ones, and also allowed a general fall in the price per GB stored on high capacity SSDs.
High capacity SSDs with MLC “just as good”
By segmenting the market into SLC, MLC and eMLC devices, manufacturers could sell enterprise SLC flash at higher prices. They could argue that it was more reliable, and because more flash cells were needed to provide a given capacity the higher prices demanded for high capacity SSDs for enterprise use could easily be justified. By including advanced controllers to provide extra reliability, and plenty of spare storage cells that could be swapped in to replace bad blocks when necessary, high capacity SSDs for the enterprise ended up being very expensive indeed.
But things have moved on in the high capacity SSD space, and the stigma around anything that isn’t made of SLC has gone. It’s now pretty standard for high capacity SSDs for enterprises to make use of MLC or even TLC (triple level cell) flash. That’s because manufacturing advances mean that MLC is arguably just as good as SLC for all practical purposes. “Today’s SSDs are more reliable and have better wear than ever,” says Jim Handy, a solid state storage expert and semiconductor analyst at Objective Analysis. “Today’s MLC is better than yesterday’s SLC.”
What’s more, falling processing power means that increasingly powerful error correction systems can be used in controllers, such as powerful but computationally complex error correction algorithm called Low Density Parity Check (LDPC). (LDPC was originally invented in the 1960s, but it required more processing power than was available at the time so it was largely ignored at the time.)
Time for high capacity SSDs with TLC
That means that TLC – which is more prone to errors than MLC but offers even more capacity – can now be used without reliability problems. And so you can argue that there’s really no reason for manufacturers not to use TCL in high capacity SSDs aimed at the enterprise market, says Handy. “I do believe that eventually everything will be using TLC,” he says.
A triple level cell, remember, has the ability to cram three bits of data in each cell, using a transistor that can store eight different charge levels. Because it can store more data in the same number of cells, TLC is very much cheaper per gigabyte than SLC (and cheaper than MLC), and it also enables more storage capacity to be built into a given space: it provides much higher storage density.
3D NAND needed for high capacity SSDs
So TLC is a key to making high capacity SSDs, and it’s one reason that what constitutes a high capacity SSD is about to change. But that’s not the end of the story. The action now is all around 3D NAND chips, which stack NAND in vertical layers like sheets of pastry in a layer cake, again providing far more storage capacity in a given volume. That means very high capacity SSDs with extraordinary storage density can be mooted.
Making 3D NAND is a challenge, because fabricators face difficulties lining up the layers exactly so that vertical tunnels can be created though the layers, says Handy. None the less, the current state of the art is 48 layers, with Samsung shipping 48-layer 3D V-NAND chips already. These result in individual NAND chips with a capacity of 256Gbits (32GB). A high capacity SSD crams in a number of these NAND chips to produce a standard sized SSD with a total capacity which might be as much as 15TB.
32TB high capacity SSD
But Samsung has already demoed a 32TB high capacity SSD using 64-layer 3D NAND, which indicates that it can now make individual 64-layer chips which have a capacity of 512Gbit (64GB). In fact its 2.5-inch form factor high capacity SSD for enterprises will be made up of 32 stacks of 1TB of storage. Each of these 1TB storage stacks will be made up of 16 of these new 512Gbit chips.
This makes sense, because Samsung’s largest current high capacity SSD, which has a capacity of just under 16TB, is made in a similar way – using 32 stacks of 1/2TB of storage. Each stack is just 1/2TB of storage because they are made up of 48-layer 256Gbit chips instead of the newer 64-layer chips which have twice the capacity.
High capacity SSD limits
An obvious question then is what the limit is for high capacity SSDs? Samsung’s 32TB high capacity SSD is not slated for release before next year, but the company and its rivals are bound to be working on even high capacity products.
At the moment it is proving very hard to make NAND chips with more than 64 layers, but it seems inevitable that 96 and even 128 layer chips will be feasible sooner or later, once manufacturing techniques develop. For example Toshiba has talked about “super stacking” technology, which can stack more than 100 layers.
And simply moving to a larger form factor – like the 3.5-inch standard – could make a huge difference. In fact Seagate has also announced a 60TB high capacity SSD, which may simply be using a large number (over 1,000) 32-layer NAND chips inside. If that can be upgrade by using 64-layer chips then a high capacity SSD approaching 100TB would certainly seem to be feasible.
High capacity SSDs will breach 100TB
Further out, Toshiba has already talked about a high capacity SSD with 100TB capacity at the Flash Memory Summit held in August in Santa Clara.
How would it produce this? The answer is by going back to the cell level, and introducing quad-level cell (QLC) technology which, as the name suggests, uses four bits per cell to store data in 16 different voltage states. This could be combined with super-stacking technology to make high capacity, multi-layer stacks built in to a high capacity SSD.
(Such a high capacity drive might consume 0.1W when not in use. To provide a similar storage capacity with 12 8TB nearline disks you would need about 100W while the disks are idling. That means that a 100TB high capacity SSD used as an archive could offer a much lower TCO than one built with 12 nearline hard drives, Chris Mellor points out in The Register.)
But there are other possibilities too – for example using die shrinking technology to increase storage density so that more storage chips can be fit in a given form factor of high capacity SSD. And indeed this does seem to be happening: Western Digital (which acquired SanDisk in May 2016) and Toshiba are making 64-layer 3D NAND.
What’s interesting about the BiCS3 NAND they are making is that it will use 64-layer TLC chips that still have a 256Gbit capacity. The previous generation 48-layer BiCS2 chips also have a 256Gbit capacity, so it would seem that Western Digital and Toshiba are experimenting with smaller chips with more layers, resulting in the same storage capacity. The next step would then be to use these smaller dies to make the larger 512Gbit 64-layer chips.
1PB high capacity SSDs – when, not if
And then there’s the possibility of simply designing the layout of products more efficiently to get more storage on board to produce very high capacity SSDs. That’s the case with Seagate’s 2TB XM1440 M.2 high capacity SSD drive, which was previously a 960GB drive. How did Seagate double the capacity? The newer high capacity SSD doesn’t have higher density storage packages – it simply has used better placement to fit eight packages instead of four on the gum stick format device.
What’s clear from all this is that while we may be used to 256GB, 512Gb and even 1TB high capacity SSDs, things are about to change very rapidly. Using an ever increasing number of layers of triple (or even quadruple) level cell flash – combined with powerful, efficient controllers – ultra high capacity SSDs with capacities approaching a Petabyte are probably more a matter of when, not if.