The MRAM flash memory has been in development since the 90s, but it was not until the year 2019 that it came to light that Intel was working very seriously on it with the aim of taking it to the production ramp. But what is MRAM memory? Will it ever replace current Flash and DRAM memories?
Although the development of this type of flash memory dates back almost 20 years, the continued development of existing technologies for Flash memory and DRAM has prevented the generalization of its use, although its defenders think that its advantages are so evident that it will come the day when they finally undo Flash and DRAM memories and become the memory that we will all use on our PCs.
What is MRAM flash memory?
MRAM stands for Magnetoresistive Random Access Memory , or magnetoresistive random access memory. It is a new type of memory that does not change the way of managing data, and yet changes everything. And it is that to begin with it is a non-volatile memory , which means that when you turn off the PC the data will continue to be stored, in addition to having an incredible performance rate per wafer, making it much cheaper to manufacture.
Unlike DRAM or Flash memory, data is not stored in electrical charges or current flows, but by means of magnetic storage elements made up of two ferromagnetic disks, each of which maintains a magnetic field.
These two discs are separated by an insulating layer, and the binary data (ones and zeros) are interpreted because one of these discs is placed on a magnet, while the other moves to match the field of the other, changing from positive to negative. An MRAM memory chip is made up of a mesh of these cells.
What advantages does this type of memory have compared to the current one?
The first advantage we have already told you: since it is not volatile , it can maintain information even if there is no power involved, when we turn off the PC (according to Intel, it is able to retain data for 10 years at 200ºC). We have also talked about the second advantage: its performance per wafer is much higher (99.9%), so it is cheaper to manufacture and by far. In addition to this, according to Intel the access times are much lower (around 1 ns), so it also provides better performance .
These MRAM memories, which are still in development, are being manufactured with Intel’s 22 nm FFL process, a node with high efficiency. According to the manufacturer’s documentation, each MRAM bit cell for each transistor (0.0486 um2) together with the magnetic tunnel junction occupies an area of only 216 x 225 nm2, which means that the dimensions of the “target devices” would be between 60 and 80nm, allowing for huge voltage variations .
This is very important because the higher the voltage, the shorter the access times are, and that is that at 0.9V the reading detection is only 4 ns, while at 0.8V the time rises to double, 8 ns.
When and how will MRAM arrive?
In theory, MRAM is already ready for mass production, so we have to wait to see how Intel intends to introduce it (and hopefully it will not do as with its failed Optane memory) because they have not yet presented any plan for it.
Yes, it is known that it has been in production for a year, or at least that’s what they said at the beginning of 2019, so of course they have something fat on their hands, or they have decided to finally discard it, something that does not seem likely after all the efforts put into its development and how promising its operation and, above all, its profitability.
