Introduction to MRAM (Magnetoresistive Random-Access Memory) [MiniTool Wiki]
What Is MRAM?
What is MRAM? It is the abbreviation of magnetoresistive random-access memory, a type of non-volatile random-access memory that stores data in magnetic domains. MRAM has the potential to become a universal memory - it can combine the density of storage memory with the speed of SRAM while maintaining non-volatility and high power efficiency at all times. Currently, the memory technologies in use (such as flash RAM and DRAM) have practical advantages, and so far they have kept MRAM in a niche position in the market.
Data in MRAM is stored by magnetic storage elements, rather than stored as electric charge or current flows. The element is formed by two ferromagnetic plates, each of which can maintain magnetization and is separated by a thin insulating layer.
One of the plates is a permanent magnet set to a specific polarity. The magnetization of another plate can be changed to match the external magnetic field to store the memory. This configuration is called a magnetic tunnel junction and is the simplest structure of MRAM bits.
The simplest way to read is to measure the electrical resistance of the cell. A particular cell is (usually) selected by powering the associated transistor, which switches the current from the power line through that cell to the ground.
Due to the magnetoresistance of the tunnel, the electrical resistance of the cell varies with the relative direction of the magnetization in the two plates. By measuring the resulting current, the resistance inside any particular cell can be determined, and thus the magnetization polarity of the writable board can be determined.
Generally, if two plates have the same magnetization alignment (low resistance state), they are considered “1”, and if the alignment is antiparallel, the resistance will be higher (high resistance state), which means “0”.
Suggested uses of MRAM include the following equipment, such as aerospace and military systems, digital cameras, laptops, smart cards, mobile phones, cellular base stations, personal computers, battery-backed SRAM replacement, data logging specially memory (black box solution), media players and book readers.
Comparison with Other Systems
The main factor that determines the cost of a memory system is the density of the components that make up the storage system. Smaller components, fewer components, means that more “cells” can be packaged on a single chip, which in turn means that more cells can be produced from a silicon wafer at a time. This increases production, which is directly related to cost.
MRAM is physically similar to DRAM in appearance and usually does require the use of transistors for write operations (though not absolutely necessary). Scaling transistors to higher densities will inevitably lead to lower available currents, which may limit the MRAM performance of advanced nodes.
Since capacitors used in DRAM lose their charge over time, memory components using DRAM must refresh all cells in their chip 16 times per second, reading each cell and rewriting its content. As the size of DRAM cells decreases, it is necessary to refresh the cells more frequently, resulting in greater power consumption.
On the contrary, MRAM does not need to be refreshed. This means that it not only retains memory when the power is turned off, but also does not continue to consume power. Although in theory, the reading process requires more power than the same process in DRAM, in practice the difference seems to be very close to zero.
However, the write process requires more power to overcome the existing magnetic field stored in the junction, and the power required for the reading process ranges from three to eight times.
MRAM is usually touted as non-volatile memory. However, the current mainstream high-capacity MRAM, spin-transfer torque memory provides better retention at the cost of higher power consumption (that is, higher write current). In particular, the critical (minimum) write current is proportional to the thermal stability factor Δ. The retention force is directly proportional to exp(Δ). Therefore, the retention rate decreases exponentially as the write current decreases.
Static random-access memory (SRAM) is currently the only performance that can compete with MRAM and has a considerable density. SRAM consists of a series of transistors arranged in flip-flops, as long as power is applied, they will maintain one of two states.
Since transistors have very low power requirements, their switching time is very short. However, because SRAM cells consist of several transistors, usually four or six, their density is much lower than that of DRAM. This makes it expensive, which is why it is only used for a small amount of high-performance memory, especially the CPU cache in almost all modern central processing unit designs.
Although MRAM is not as fast as SRAM, even so, it is interesting enough. Given its much higher density, CPU designers may prefer to use MRAM to provide larger but slower caches rather than smaller but faster caches. How to make this trade-off in the future remains to be seen.
The endurance of MRAM is affected by write current, just like retention and speed and read current. When the write current is large enough for speed and retention, the possibility of MTJ breakdown needs to be considered. If the ratio of read current/write current is not small enough, read disturbance may occur, that is, a read error occurs during one of the multiple switching cycles.