In this blog post, we’ll examine the basic structure of flash memory and the differences between SLC and MLC technologies, and explore how multi-level cell technology has evolved to increase storage density.
We live in the information age. Vast amounts of information are digitized and exchanged at a rapid pace. We no longer have to sit in front of a bulky computer to access information. People carry information in their pockets as they go about their daily lives, exchanging data while walking down the street or riding the subway. Devices like USB flash drives, cell phones, and MP3 players have become such an integral part of our lives that calling them “high-tech” feels almost out of place. Flash memory is widely used as a storage medium in these portable digital devices. Flash memory is a storage device that allows digital information to be written and erased electronically. There are several types of flash memory, classified by the method used to store information in a single memory cell; the most common types include SLC, MLC, TLC, and QLC.
First, let’s look at Single Level Cell (SLC) technology, which is the basic storage method for flash memory. Digital information comes in various forms—such as documents, images, and music—but when viewed without context, it is ultimately represented as a sequence of 0s and 1s. Flash memory, which stores this information, does not contain any complex machinery; rather, it consists of a structure made up of countless tiny electronic components called transistors. The transistors used in flash memory are a slightly modified version of standard transistors; they can store or release electrons, and their electrical properties change depending on how many electrons are present. In other words, a transistor acts like a container for electrons, and we can determine how many electrons are inside that container using electrical methods. SLC flash memory utilizes this principle to distinguish between two states and store digital information. In SLC, a single cell stores 1 bit of information, and each cell distinguishes between two states to represent either a 0 or a 1.
In contrast, there is Multi-Level Cell (MLC) technology. While SLC flash memory uses two states per cell to represent 1 bit—that is, 0 or 1—MLC flash memory distinguishes multiple states based on the amount of charge stored in a cell, thereby representing more information. The method commonly referred to as MLC today stores 2 bits per cell, allowing it to distinguish four states. In other words, it stores the four combinations—00, 01, 10, and 11—in a single cell.
However, since states are distinguished based on the amount of charge, MLC flash memory requires more sophisticated detection technology than SLC flash memory. In SLC technology, because only two cell states are distinguished, the difference between each state is relatively large. In contrast, in MLC technology, because the amount of charge is divided into multiple levels, the difference between the states that need to be distinguished becomes smaller. The more levels there are, the smaller the differences become, making it increasingly difficult to distinguish between states. MLC can distinguish four states and represent 2 bits of information in a single cell. Compared to SLC, this means that twice as much information can be stored with the same number of cells.
This multi-level storage method did not stop at MLC. Subsequently, Triple-Level Cell (TLC) technology, which stores 3 bits per cell, and Quad-Level Cell (QLC) technology, which stores 4 bits per cell, were also commercialized. TLC distinguishes between eight states per cell, while QLC distinguishes between sixteen states. Therefore, even when using the same number of cells, the amount of information that can be stored in a single cell can be increased. Currently, in NAND flash memory, SLC, MLC, TLC, and QLC are utilized as storage methods, each with different applications and characteristics.
However, SLC has the advantage in terms of read/write speeds and durability. When reading data from memory, flash memory that uses MLC or even more states than that finds it harder to distinguish cell states compared to SLC. Consequently, more processing steps are required to determine the state.
The same applies when programming a specific state into a cell. Unlike SLC, which only needs to distinguish between two states, methods requiring precise adjustment of charge levels across multiple steps demand finer control. Furthermore, as the number of storage states increases, so does the importance of error correction technology for accurately detecting and correcting errors. In fact, with NAND flash, as we move from SLC to MLC, TLC, and QLC, the read and program processes become more complex, and the need for error correction increases. It’s similar to pouring water into a cup: if you pour with the intention of filling it to the brim, the water flows freely and the cup fills quickly; but if you need to fill it precisely to a specific line, you must pour little by little, which takes much longer.
Thus, both SLC and MLC flash memory have their respective strengths and weaknesses. Nevertheless, the reason so many people were captivated by the idea of MLC was that its drawbacks were considered acceptable. If one were to name a single feature that distinguishes flash memory from other storage devices, it would be the fact that it stores and reads information electrically. Traditional disk-based storage devices require mechanical movement to read and write data. In contrast, flash memory reads and writes information electrically, so it does not require such mechanical movement. Even though MLC flash memory is relatively slower and has lower write endurance than SLC, the advantage of being able to store and read information without mechanical movement remains unchanged. Furthermore, in current NAND flash technology, methods that store more bits in a single cell—such as TLC and QLC—are widely used in actual high-capacity storage devices.
While there is certainly a difference in height between an elementary school student and an adult, this difference is negligible when compared to the Earth. Similarly, when compared to storage devices that require mechanical movement, the speed difference between SLC and MLC flash memory may be significant depending on the specific conditions, but it does not alter the inherent advantages of the electrical storage method provided by flash memory. Ultimately, what matters is striking a balance between how much information can be stored in a single cell and how well speed, durability, and reliability can be maintained in the process.
However, MLC flash memory holds such great potential that these drawbacks are worth accepting. The key to this potential lies in the technology that distinguishes the amount of charge stored in a transistor across multiple levels. A single transistor can store a very large amount of charge. Although transistors are very small objects compared to the things we encounter in daily life, they are not entirely invisible. However, when viewed in the microscopic world—a realm of things invisible to the human eye—transistors can be considered relatively large structures. In contrast, electrons are very small particles even in the microscopic world. Therefore, it is only natural that a single transistor can store a very large amount of charge. If we can use these characteristics to distinguish storage states more precisely, we can increase the amount of information that can be stored in a single cell.
In fact, flash memory technology has been steadily advancing in this direction. While SLC and MLC were the main technologies in the past, TLC and QLC have since emerged, and QLC technology is now being used in commercial products. Since QLC stores 4 bits in a single cell, it must distinguish between 16 different states. Major semiconductor companies, including Samsung Electronics and Micron, are currently developing and mass-producing high-density QLC NAND flash, and they continue to expand storage capacity through 3D NAND technology, which vertically stacks memory cells in three dimensions.
Consequently, the amount of information that a single transistor or memory cell can hold has increased significantly compared to the past. However, the number of states cannot be increased indefinitely. As the number of cell states increases, the differences between each state become smaller, making it more difficult to distinguish and correct subtle electrical fluctuations, interference, and errors. In actual NAND flash, it has been confirmed that as the number of stored bits increases, so do the challenges regarding performance, durability, and error correction.
That does not mean there are no ways left to increase storage capacity. In addition to technologies for more precisely controlling and distinguishing the amount of electrons, other advancements are underway, such as techniques for storing more states in a single cell and 3D NAND technology, which stacks memory cells vertically. KIOXIA explains how 3D NAND increases storage density by vertically stacking memory cells and is currently utilizing TLC and QLC technologies in its products.
Therefore, while expressing eight or sixteen states with a single transistor was once considered a future possibility, QLC—which utilizes sixteen states—has now become a reality. Although technology capable of distinguishing even more states may develop in the future, this does not guarantee that a single transistor will be able to store an unimaginable amount of information, nor does it ensure that a device the size of a standard USB drive will be able to hold a near-infinite amount of data. This is because the storage capacity of a device is not determined solely by the number of states a single cell can distinguish, but rather by a combination of factors such as cell size, stacking structure, error correction technology, data retention characteristics, and durability. Nevertheless, the evolution of flash memory—which has continuously increased the amount of information that can be stored in a single small cell—is moving in a direction that brings us significantly closer to the future once imagined as an “infinite-capacity USB drive.”
