Solid State Drives, better know as SSDs are rapidly increasing in popularity. While the cost per Gigabyte is not competitive with traditional hard disk drives (HDDs) yet, the boost in performance, reduced power consumption and the lack of moving parts contribute to the rapid adoption rate in both consumer and enterprise computers. The large amount of companies fighting for market-share keeps competition fierce which drives prices down and stimulates innovation.
In such a fast market is might be hard to keep up with all the terms and abbreviations, here are five terms you should know to stay up to date. If you want more information or have other questions, be sure to join our SSD Forum.
With the increase of SSD speeds the demand for higher bandwidth between the drive and the rest of computer became evident. Traditionally SSD drives connect to the computer using the SATA interface. While new specifications of this interface are released once in a while, it’s relative low bandwidth is often a bottleneck.
SSDs would be much faster when not limited by the SATA interface. Therefore SSD manufacturers started to look at the PCI Express (PCIe) interface to connect SSDs. This interface provides much higher bandwidths. The latest SATA specification (3.2) brings the SATA speed to 2 GB/s while the currently most commonly used SATA specification (3), only allows speeds up to 0.6 GB/s. That’s only a fraction of the bandwidth PCI Express provides, which can achieve speeds up to 8 GB/s, four times more than the most recent SATA specification. It might be no surprise that PCIe based interfaces rapidly became popular.
Unfortunately this resulted in different implementations by different manufacturers, all requiring different drivers. Every SSD manufacturer had to create their own drivers for every OS, like Windows, Linux, FreeBSD and OSX. NVMe provides an interface controller standard and by implementing a native driver in the operating system all NVMe SSDs work out of the box. This means manufacturers don’t need any resources to develop drivers themselves.
Once the SSDs is plugged in to a NVMe compatible OS it will simply work. Currently most operating systems have a NVMe driver or have announced to include one soon.
DevSleep is part of the SATA specification and its goal is to make SATA devices (read SSDs) to go into a low power mode while becoming immediately available on request. This means that a SSD can be nearly entirely powered off. The standard defines that all circuitry is powered off when in DevSleep and first demonstrations have shown a decrease of power consumption of a whopping 99.8% When you wake the system the SSD powers up in a mere 20 ms which does not give a noticeable delay.
LSI has announced that their SandForce controllers will support the technology, but also Intel announced SSDs with the feature.
M.2 / NGFF
M.2 or Next Generation Form Factor (NGFF) is standard form-factor developed by Intel to reduce the different sizes used for mSATA SSDs. This allows for standardized small card-sized SSDs where the standard provides lenghts between 30 mm and 110 mm. Manufacturers can use both sides of the card to put their NAND chips on and height of all cards is the same. The specification supports both the mSATA and the PCIe interface. Intel has made the standard part of the Ultrabook specifications.
In order to make SSDs cheaper, the amount of NAND memory chips that can be made from a single wafer has to go up. These NAND memory chips make up the largest cost of a SSD and lots of efforts have been made to decrease the size of these chips will retaining the same capacity. However the more data dense the chips become, the more error prone they get. NAND memory is made up of cells that are currently as small as 19 nm. The data is stored by trapping electrons controlled by applying a voltage.
The smaller the cells get, the more they suffer from interference by electrical effects. You can consider this effect the same when you hold your mobile phone near a speaker. The more disturbance, the larger the chance for data errors and the more unreliable the SSD. To accommodate smaller cell sizes the SSD controller needs better error correction algorithms which make SSDs slower. Also using different materials to trap the electrons have been tried but manufacturers often stated that they are near the limit with traditional NAND.
Therefore all major manufacturers are pursuing new technologies to create higher density chips. All of them are looking at three dimensional NAND but Samsung has been the first to announce commercial production of 3D NAND which Samsung calls V-NAND.
The company claims twice the density compared to regular NAND, two times higher performance, ten times more reliablity and half the power consumption.
The technology works by stacking a new type of NAND cell. These cells are not only put next to eachother but also on to eachother. This means that Samsung can make larger cells (they now make 30 nm cells) but due to the fact they increase the chips in height can still hold more data. Currently the company is able to stack 24 layers of NAND on eachother. If they are able to add more layers or decrease cell size, they can improve on that number.
More information on V-NAND can be found in the video below.
ReRAM (Resistive Random Access Memory) is a much anticipated product that we likely won’t see in consumer drives in the near future. But there’s already a lot of talk about the technology so it deserves a place on this list. ReRAM will be the follow up of NAND memory, it promises non-volatile storage at the speeds of RAM memory.
Recent announcements claim that it should be 20 times faster than NAND flash while consuming less power. Currently ReRAM is in heavy development and several companies have all announced their own form, using different materials. The data density of the chips should be high, a Terabyte of data should fit on the size of a post-stamp
The technology works by creating resistance instead of trapping electrons, like in NAND. By applying an electric current to a material, the resistance of that material is changed. The resistance state can then be measured and either on or off, or 1 or 0. Much of the research on ReRAM focusses on finding the right materials and measuring the resistance state of the cells.