Looking at the movements of the 3C product market in the past five years, it's safe to say that the reputation of Solid State Drive (SSD) applications have rapidly grown within the industry. SSDs are able to provide up to ten times the read/write speeds of any traditional memory hardware in terms of transferring data and information for computer systems. With the increased momentum stemming from relevant businesses, and with prices of flash memory continuing to shrink, SSD products are becoming an ideal, highly sought-after memory solution for improving computational efficiency.
To facilitate a better understanding of the gradually maturing memory market, we have analyzed a series of performance indicators of SSDs, and provided recaps on some of their more critical applications. Through the information presented below, we hope to help readers develop a better understanding of the evolution of solid state drives and discover the most suitable types of SSD solutions for their computers.
Understanding SSDs from the component perspective
The SSD hardware market, upon first glance, is filled with a plethora of business vendors, a trend which can make the learning of SSDs a bit overwhelming to handle. In an effort to make matters easier, we have managed to organize as well as single out the key aspects that are the most worthy of mentioning.
In a nutshell, a SSD's most important components are its flash memory and controller. The three most popular global suppliers for the former component are currently Samsung, Toshiba, and Intel. With regard to the controllers—which our reports from the previous two years demonstrate are highly critical to hardware performance—the number of suppliers tend to be slightly more varied. Given how major memory manufacturers typically promote only one or two flash memory products, finding a perfectly suitable flash memory/controller combination can be a relatively uneasy, if not highly cumbersome, task.
The high-end SSD market, at present, is dominated by SanForce, Marvell, and Indilinx, all of which use distinctive flash memory components. Intel's flash memory is not necessarily always the best option for vendors and consumers, despite its apparent popularity. In the grander scheme of things, factors such as the compatibility between a controller and flash memory, a device's basic internal structure, and hardware performance are all likely to affect how well SSDs are accepted by the general consumer audience.
What is Toggle? ONFI?
Consider, for a moment, the relationship between a typical website visitor and the written content of a webpage. The readers of the web content will be able to discern what a webpage is conveying based on their understanding of the language used and the purpose of the web medium. Conceptually, we can view the "language" and "website" aspects of the aforementioned scenario to be analogous to a computer's "protocol" and "interface."
SSD controller and Flash memory can be best understood in terms of the relationship between a transmission interface and its communicative standard. The interaction between Toggle and ONFI essentially operates on the same kind of principle and rules. With an incoming wave of signals, a flash memory receives the read-write commands of its controller, which then ensures that all the necessary protocols and actions are carried out correctly. This is similar to what happens with DDR (Double Data Rate), a standard that enables the transfers between a controller and flash memory component to be performed at up to twice the original speed. (The actual frequency required for 200 MT/s transfer rates, for example, is only around 100 MHz). With the proper utilization of an interface's controller and flash memory, it is generally possible to develop any SSD products with approximately twice the original performance capacity.
One general issue that has garnered quite a lot of attention lately is the lack of compatibility among different formats. Samsung and Toshiba supports only the Toggle format, for instance, whereas ONFI –or Open NAND Flash Interface- is generally favored by Intel, Micron, Phison, SanDisk, Hynix, Sony, and Spansion. Even though it is not entirely impossible for different controller vendors to be able to adapt to either of the two aforementioned standards, for parties like the assembly manufacturers, having to handle more than one distinctive format can be a bit of a hassle. Success in this area, on the whole, will likely depend on finding appropriate ways to reconcile—as well as integrate—a variety of different standards.
At the moment, both the ONFI 3.1 and Toggle 2.0 flash memory formats are able to support transfer rates up to 400MT/s. If the eight-part Flash chips are assembled into a single memory unit, it will become strong enough to support even greater bandwidths. The only elements left that would need any enhancing are the flash memory's reading and writing speeds.
It can be argued that a vendor's overall ability to manage its brand image—along with its products' qualities— are as critical as the effectiveness of the controller it chooses. Also of great importance are issues such as how to appropriately apply, integrate, and simultaneously manage different component parts. A further understanding of these elements can be helpful in terms of improving a flash memory's overall lifespan.
The state of SLC and MLC
Interestingly, both SLC and MLC are gaining renewed interest despite having been extensively covered in the past. In assessing the two formats from the manufacturing as well as application perspective, we hope to provide readers with a further insight into their applicability, pros and cons, and future prospects.
SLC (Single-Level Cell) and MLC (Multi-Level Cell), simply put, refer to two different ways of storing data through flash memory. In the case of SLC, data is stored inside a single cell and in the simple "0" and "1" format. This comes with two major benefits: First, the delays associated with the read and write speeds become almost negligible, which enables for higher endurance levels and, in turn, longer data support. Second, power consumption is reduced by a significant amount. As a direct result of this, the SLC's application can be extended to a wider assortment of hardware products, including the smaller and more portable mobile devices. Where SLC is currently being criticized is its limited storage capacity. Even with the implementation of the newest manufacturing processes, the storage provided by SLC is still less than that of a typical systematic product.
MLC is different from SLC in that it divides a cell's space into four separate layers, each with the capacity to store two bits of data. In MLC, an individual cell's number of electrons determines the content of the information stored, whereas the controller ensures every cell is able to read and write the data properly. MLC technically provides up to eight times more capacity than SLC, which in effect opens up the way for more flash memory applications in the market. Its problems can be divided into two major categories. The first is its relatively limited lifespan; the flash memory's data re-writing processes occur only after a certain number of electrons have passed over to the floating gate. Its power levels tend to drain rather quickly given the division of an individual cell into four different parts and the need to perform up to eight times the work.
The second flaw of MLC concerns its reading and writing efficiency. In terms of the data read by each controller, the average delay response time is roughly 10ns for SLC, and 44ns for MLC. The delay writing response time for MLC, on the other hand, is around 400ns, approximately three times the amount for SLC. The relatively weak life span of MLC generally worsens as the manufacturing processes become more and more intricate. This is in large part due to the narrowing of the floating gate and the reductions taking place inside the cell. Problems like these are likely to be resolved only through major advancements within the field of material technology.
A number of notable solutions –some directly intended for the above issues-- are worthy of mentioning here. On the hardware end, the most popular options available include eMLC (Enterprise MLC) and "MLC/SLC mode." As suggested earlier, standard MLC utilizes four separate layers when storing any forms of data. "MLC/SLC mode," by contrast, makes use of only two of the layers, and allows each cell 2 bits of space for data storage. Through this arrangement, the reading delay response times can be minimized, while endurance is enhanced. The downside is that a large portion of the layers (up to 50%) remains unused for the saving and transference of data. This is generally perceived to be one of the necessary compromises needed to ensure stability and efficiency.
The nature of eMLC has been open to various interesting interpretations. Some perceive it to be a more sophisticated version of a MLC-based solution. Others believe it is just another term for "MLC/SLC." One way to resolve the confusion is to look to the product's endurance level. Ordinary products have limited endurance, no matter how strict the standard. For the typical MLC, the average wear leveling count is generally around 3000 cycles (5000 for those with better qualities or made using older processes). About 80 to 90% of the products with a wear leveling count of 8000 cycles are of the MLC-SLC variety. The SLC products, while good to use, tend to deter consumers due to their higher price points. In the end, even if it's difficult to decide what solution should be chosen (or which one is the best), the creators of the aforementioned designs still deserve a lot of credit for what has been accomplished so far.
|