Non-volatile memory (NVM) plays an essential role in a wide variety of systems ranging from mobile computing platforms to embedded systems. Flash memory has dominated this space for a number of years, with gate sizes continually shrinking to allow it to meet performance demands. There are limits to minimum gate size, however, forcing the industry to see alternatives. An emerging nanotechnology-enabled approach, with a particular focus on a recent development called split-gate thin-film storage (SG-TFS), provides a promising alternative. Let’s take a step back and look at the super category of NVM technologies, and then explore the world of flash.
There are memory technologies that cannot retain the stored information when the product is not powered; hence, they are called volatile memory technologies. NVM is a semiconductor memory technology that can retain the stored information even when it is not powered; hence it’s called non-volatile. Examples of NVM include read-only memory (ROM), flash memory, ferroelectric RAM (FRAM), and most types of magnetic computer storage devices.
In most electronic systems, there are two types of memory technologies used: A primary storage technology (volatile), and a secondary storage technology (NVM; see figure 1).
Figure 1: This simple lighting module MCU uses two types of memory: primary (volatile) and secondary (NVM).
The most popular type of primary storage technology today is called static random access memory (SRAM), which is a volatile form of memory technology, hence when the system is shut down, anything contained in SRAM is lost. NVM is typically used for secondary and long-term storage, when retention of the stored information is needed during power-down. If retention of the information is not required, for example due to the presence of an external memory drive, dynamic RAM (DRAM) is often used. This discussion will focus on the former case, in which NVM is required for secondary storage.
The SRAM primary storage provides read and write times close to the clock period of the system’s processor, and thus is used where minimal latency is required. The NVM typically requires several clock cycles to read or write, but it is chosen for secondary storage because it retains its contents during power-down and provides minimal cost per bit; hence, many systems use a moderate amount of RAM in combination with a larger amount of NVM (typically flash memory) to optimize both cost and performance.
There are a few technologies under development that are more comparable to SRAM, or at least to DRAM, in terms of performance while also meeting the non-volatility requirement for the secondary memory . These include magnetoresistive RAM (MRAM), phase-change memory (PCM), and resistive RAM (RRAM). While these technologies show good potential, they do not typically match the sub-5-ns read/write times needed to fully replace SRAM, and have not yet achieved the maturity and cost-per-bit to fulfill the NVM secondary memory requirement. There is also ferroelectric RAM (FRAM), which has shown good advantages over conventional NVM concerning endurance, write time and write power, but again FRAM does not even approach SRAM in read/write performance, and has been difficult to scale beyond the 130-nm node, or to use in applications requiring temperatures higher than 85ºC. There is thus an ongoing need for innovative solutions that will allow continued scaling of flash memory as the choice for secondary storage.