Welcome to the era of ash
raditional hard dis drives may very well still have a place in modern computing, but the future belongs to !s.
Six years ago, a 256GB SSD would set you back nearly $600. Today, the same capacity drive can be had for as little as $125 and that same $600 can get you a 1TB SSD with change to spare. Prices of SSDs have fallen dramatically and that has had a profound effect on the computing landscape.
SSDs are now becoming increasingly commonplace. A recent report by TrendForce said that the adoption rate of SSDs in notebooks will hit 50% this year. Put another way, one out of every two new notebooks will have an SSD. Due to falling costs, SSDs are also becoming increasingly popular in the enterprise space, especially in performance-critical applications. Many research rms believe that shipments of SSDs will surpass HDDs in the next two to three years. So what is it about SSDs that gives them a leg up over its mechanical rivals? Let’s look at some key considerations now.
HDD VS. SSD Performance
We all know performance is one of the biggest advantages that SSDs have over HDDs. Even the most basic SSD will run circles around the fastest, most advanced HDDs. SSDs are therefore well suited to performance-critical applications or tasks where high performance and throughput is required. For consumers, SSDs provide signicantly shorter booting up and loading times. In the enterprise space, SSDs are suitable for data analytics, virtualization, and cloud services. rvices.
To give you an idea of justst how fast SSDs are, consider the tablee opposite page:
HDDs used to have the edge over SSDs because of their higher capacities. But this is no longer the case. Samsung announced a 15TB SSD in 2016 and followed up with a 30TB SSD just earlier this year that comes in a slightly thicker 2.5-inch form factor. For businesses who need more, Nimbus Data’s ExaDrive DC100 is a 100TB SSD that comes in standard 3.5-inch form factor. In comparison, the largest capacity HDD currently holds only 14TB of data.
SSDs are also considerably moree power r e cient than HDDs. For mobile devicesdevices, , using SSDs in place of HDDs can extend d battery life. In the enterprise space,ce, this crucially means they generate lessss heat so less robust cooling is required,, which h translates to easier maintenance and lower running costs.
Shock and vibration
SSDs have no moving parts, which makes them signicantly more resistant to shocks as compared to HDDs. HDDs can be easily damaged if they are jolted when in use. And since SSDs have no moving parts, they don’t vibrate. This makes them easier to deploy in data centers. On the other hand, su cient care must be taken to ensure that data racks and servers have the structural rigidity required to withstand the vibration caused by multiple HDDs spinning simultaneously.
SSDs have a lifespan. There are only so many times you can write to a cell. However, SSD endurance and reliability have been steadily improving and are now as good as, if not better, than HDDs. One way of looking at reliability is MTBF or mean time between failures. Enterprise-grade SSDs have an MTBF of 2 to 2.5 million hours, similar to enterprisegrade HDDs.
For consumers, a better way of looking at reliability is endurance. Take, for example, a 512GB version of Samsung’s new SSD 860 Pro. That has a rated endurance of 600TBW, which means you can write 600TB of data to it before it fails. In most client usage scenarios, most users will write, at most, about 30GB of data to their drives a day. Even if you are a heavy user and write 50GB of data per day, 600TBW is enough for roughly 32 years.
Even though prices of SSDs have fallen signicantly. They are still considerably more expensive than HDDs. Samsung’s 15TB enterprise-grade PM1633a SSD, for example, retails for over US$10,000. On the other hand, a 12TB enterprise-grade HDD typically costs about US$500. Things are not much better on the client side. A 1TB SSD will set you back S$500, but a 1TB HDD can be had for as little as S$70.
THE ARRIVAL OF 3D NAND
Though the rst SSD was shipped in 1991, SSDs in the client space only arrived about ten years ago. Back then, it was widely believed that SSDs could never compete against HDD when it came to capacity and reliability. The reason for that was because of the limitations of traditional NAND architecture.
In the past, ash memory manufacturers sought to increase memory density, and hence capacity, by either using smaller manufacturing processes to cram more cells onto the die or by increasing the number of bits
that can be stored per memory cell. Both approaches have their problems.
There is a limit to how much cells one can squeeze onto the die. Since each cell has to hold a charge, squeezing too many cells onto the die results in less space between each cell. Interference from neighboring cells can disrupt or even inadvertently change the state of the cell that it is next to and cause the data to become corrupt.
On the other hand, one can keep the size of cells constant but increase the amount of data stored on it. There are, however, two key downsides to this approach and they are performance and endurance. By increasing the number of bits per cell, you are increasing the number of voltage states for the drive controller to check. This reduces drive performance.
Additionally, by having more stages in each cell, the margin between individual stages is reduced. And since writing to the cell gradually reduces the layer of silicon oxide within that acts as an insulator between the various stages, this effectively means that each cell has a shorter lifespan.
These are the problems facing traditional NAND architecture. And in 2014, Samsung made a breakthrough with what they call 3D NAND or V-NAND (Vertical NAND), and this has gone on to become one of the most important innovations in the ash memory business.
Samsung’s solution to the problems facing NAND is elegant and effective. As its name suggests, 3D NAND overcomes the limitation of traditional two-dimensional planar NAND by stacking cells upon each other. This not only allows more cells to be placed on a single die, it also creates more space between cells, allowing Samsung to roll back to a less aggressive manufacturing process node, thereby improving performance and endurance.
Samsung’s rst 3D NAND SSD to go into production was the SSD 850 Pro from 2014 and its 3D NAND memory consisted of 24 stacked layers of cells. As of now, Samsung has already put into production 3D NAND with 64 layers of cells. This has allowed them to cram 4TB into a standard 2.5-inch form factor and 2TB on a standard M.2 2280 memory drive. And at the Flash Memory Summit last year, Samsung has shared that they are already working on 3D NAND with 96 layers of cells. Today, the vast majority of SSDs employ 3D NAND, which is a testament to the importance of this NAND technology.
Beyond 3D NAND, ash memory manufacturers are already exploring different types of memory. Intel and Micron have a new type of memory called 3D XPoint, which offers significantly greater performance and endurance than NAND memory. It is marketed to address the performance gap between NAND and DRAM.
Not to be outdone, Samsung is also working on a new type of memory that they call Z-NAND. Like 3D XPoint, Z-NAND is a high performance, high endurance memory that slots between NAND and DRAM. And like Intel and Micron, Samsung has not shared details about this new memory, but the rst drive to use Z-NAND, the enterprise-focused SZ985, looks mighty promising.
3D NAND or V-NAND (Vertical NAND) has gone on to become one of the most important innovations in the ash memory business.
3D NAND STRUCTURE