What are some advantages of flash memory

Flash memory / flash memory


Flash memory or flash memory combines the advantages of semiconductor memory and hard drives. Like any other semiconductor memory, flash memory has no moving parts. As with a hard drive, the data is retained even after the power supply has been switched off.
Flash memory developed from EEPROM (Electrical Erasable and Programmable Read Only Memory). Depending on the literature, there are also the terms Flash-EPROM and Flash-ROM.
With the flash memory, the storage of data is functionally identical to that with the EEPROM. However, the data is read, written and deleted in blocks of 64, 128, 256, 1,024, ... bytes at the same time.

Computers with memory based purely on flash memory are the dream of every software developer and user. The computer would never have to boot for minutes to start up, but would be immediately ready for use within a few seconds. It would be switched off just as quickly. The next time you start it, the same programs and files would be loaded as before you switched off.

Flash memory cell with floating gate technology

The memory cell of a flash memory is very similar to the field effect transistor (FET). However, the gate (G) contains a charge trap called the floating gate. It is an electrically isolated semiconductor layer. The floating gate stores the charge like a capacitor. It is insulated from the drain (D), source (S) and control gate connections with an oxide layer. The oxide layer prevents the charge from flowing away from the control gate. In the de-energized state, the charge remains for a few years. It should be noted that the charge has to be refreshed at some point, which is done during operation. If a data memory is archived with flash memory, there is a risk that the memory content will be lost after a few years.

When the memory cell is erased, the charge jumps to the floating gate in a flash. It is charging. The current flow between source and drain is cut off. The transistor is then in the zero state.
To read the memory cell, a voltage is applied to the gate (G) of the transistor and the current flowing between the drain and source is measured. Depending on the current intensity, a certain state of the memory cell or the charge state of the floating gate can be assumed.

Difference between NAND and NOR Flash

The NAND and NOR architectures differ roughly in terms of storage density and access speed.

With NOR flash, the memory cells are connected in parallel. Access to the memory cells is random and direct. The access times are correspondingly short. The parallel connection guarantees a lower resistance between the current source and the evaluation circuit. NOR-Flash is used for the program memory of microcontrollers. For example, it is also used as non-volatile memory for the BIOS in PCs.

With NAND flash, reading and writing is only possible in blocks due to the internal serial interconnection. Due to the small number of data lines, NAND Flash requires less space. Since data on hard drives is also read and written in blocks, NAND flash is suitable as a mass storage device, such as memory cards, USB sticks and SSDs.
Compared to other types of non-volatile memory, NAND Flash allows higher storage densities at low cost and works with a significantly faster write speed and low power consumption. The differences in speeds are with the controllers.

Flash memory types

A memory cell typically only assumes two different states, which are expressed as charged and uncharged. If you are able to hold several states in a memory cell, then you can increase the storage density. In the case of flash memories, an attempt is made to map several states in a memory cell using different voltage levels.

 SLCMLCTLCQLCPLC
Bits per cell1 bit2 bits3 bit4 bit5 bits
Storable states2 (21)4 (22)8 (23)16 (24)32 (25)
  • SLC - Single Level Cell
  • MLC - Multi Level Cell
  • TLC - Triple Level Cell
  • QLC - Quadple Level Cell
  • PLC - Penta Level Cell

SLC Flash (Single Level Cell)

One bit per memory cell is stored in SLC flash. This enables around 100,000 write cycles to be achieved before a memory cell is worn out and defective.
Typically, memory cells are only used as SLC for a very large number of write cycles.

MLC-Flash (Multi Level Cell)

Two bits per memory cell are stored in MLC flash. This doubles the storage density compared to SLC flash, with the same silicon costs. However, MLC memory cells cannot be written to as quickly as SLC memory cells.
With 2 bits, a memory cell must be able to withstand several voltage levels (00, 01, 10, 11). To avoid reading errors, the reading process takes longer. Especially the more often a cell has been described. Therefore, MLCs need more error correction mechanisms. Accordingly, it takes longer to read. The reading process also puts physical strain on the cells.

Note: Originally, 10,000 write operations per MLC cell were assumed. Due to finer semiconductor structures, the number has dropped to 5,000 and is typically 3,000 (as of October 2013). Because of the charge differences for several bits, the memory cells are more prone to defects with around 3,000 write cycles per cell (with a 19 nm manufacturing process). Only through wear leveling and reserve memory cells can the overall life of the flash memory be extended, but this also increases manufacturing costs.

MLC flash is typically used for SSDs.

TLC Flash (Triple Level Cell)

TLC flash stores three bits per memory cell. Because of the higher storage density compared to MLC flash, the memory cells must be of high quality in order to be able to keep the different charge states stable for 3 bits, which makes production somewhat more expensive, but also has a higher data density.
Because this flash memory should be as cheap as possible, individual TLC cells can only write 1,000 times. This means that the service life of TLC flash is just about acceptable.
Because of the limited write cycles, TLC flash is mainly used for USB sticks and memory cards. TLC flash is also sometimes found in extremely cheap SSDs. TLC Flash is unusable in servers where there is frequent writing.

Advantages of flash memory

  • The stored data are retained even if there is no supply voltage. A trickle charge can be dispensed with. This means that energy consumption and heat generation are also lower.
  • Due to the lack of moving parts, Flash is silent, insensitive to vibrations and magnetic fields.
  • Compared to hard disks, flash memories have a very short access time. Read and write speeds are almost constant over the entire memory area.
  • The achievable memory size is open at the top due to the simple and space-saving arrangement of the memory cells.

Disadvantages of flash memory

  • limited write or delete processes (endurance)
  • limited duration of data storage (retention)

The main disadvantage of flash memory is the limited number of write or erase operations that a memory cell can tolerate.

The reliability of flash memory also suffers as the memory cells lose their charge over time. The higher the temperature and the amount of charge per cell, the faster this will happen. During a reading process, the cell content can already be tilted. The finer the structures, the greater the risk. Flash memory is therefore less suitable for long-term archiving.

Endurance and retention

Endurance refers to the maximum number of erasure or memory cycles that a non-volatile data memory (NVRAM) can withstand per memory cell until noticeable errors occur in memory operations.
The fact that the number is limited at all is due to today's flash memories (the current technology for implementing the NVRAM principle) that the memory cells for the programming and erasing operations are exposed to high voltages (10 to 18 V), which causes damage to the Structure of the cell occur. The damage becomes less, the lower the voltages for programming and erasing can be selected. This minimization requirement means that a central insulation structure in the memory cell has to be made as thin as possible. However, this in turn has negative effects on the duration of data storage (retention).

Why does a flash memory cell break?

The floating gate is charged with a high voltage during write access. This is necessary to overcome the oxide layer (insulation). The oxide layer is damaged in the process. A little more with each write access. At some point it no longer insulates and the storage cell becomes unusable.
Since a write process writes memory blocks between 16 and 128 kByte at the same time, memory cells that do not require any changes are also used. This means that a large number of memory cells are rewritten even if the memory contents are changed slightly.

Flash memory durability and reliability

The following findings are available on the durability and reliability of flash memory: The number of possible write or erase cycles does not allow any direct conclusions to be drawn about the durability or reliability. In contrast to conventional hard disks, there is no direct assignment between the memory cells and the sectors of the file system. In general, the flash controller distributes the write access evenly over all flash memory cells. The data in the cells, which are occupied by rarely changed data such as the operating system and programs, are shifted from time to time in order to get back to less worn cells. At the same time, all memory cells are regularly refreshed so that no data loss creeps in over time due to the loss of charge in the memory cells.
All processes and mechanisms that extend the service life of the memory cells fall under the term "wear leveling".

In general, it can be assumed that SSDs will last longer in everyday desktop use than stated by the manufacturers. 3,000 to 100,000 storage or delete cycles are completely sufficient for most applications. If all storage cells are loaded equally with wear leveling, the service life is extended permanently.

Further development of flash memory

The development of flash storage media is moving in the direction of long durability, short access times and high speeds. The aim is to reduce the number of write and read errors to a level so that the durability of inexpensive MLC flash moves into the range of SLC flash.
Possible solutions are 3D-V-NAND, iSLC or eMLC, in which several MLC memory cells are stacked in several layers. Instead of the typical planar architecture, a vertical arrangement is used. With monolithic 3D-V-NAND chips, there are over 60 layers on top of each other. Stacking saves silicon space, which is why the individual cells can be larger. In this way, the storage capacity, service life and speed are increased and the data storage of the individual memory cells is extended.

However, there are other requirements for flash memory. It should be as big and cheap as possible and remain switched off when idling until a specific request is received. Reliability and speed come second. Such flash memories are interesting for data archives, where the user can wait a little longer to access them.

Flash memory applications

Flash memory chips are ubiquitous. They are found in many devices that we use every day.

Flash memories are not suitable for archiving. Conventional hard drives are better suited for this.

Flash alternatives

The semiconductor industry is looking for alternatives to flash. But the reason you're not really making any headway is that 3,000 or 100,000 save or delete cycles are sufficient for most applications. In addition, the number of accesses to the memory cells is reduced with additional technical effort.

Overview: semiconductor memory

Other related topics:

Everything you need to know about computer technology.

Computer technology primer

The computer technology primer is a book about the basics of computer technology, processor technology, semiconductor memory, interfaces, data storage, drives and important hardware components.

I want that!

Everything you need to know about computer technology.

Computer technology primer

The computer technology primer is a book about the basics of computer technology, processor technology, semiconductor memory, interfaces, data storage, drives and important hardware components.

I want that!