Years behind a substitute and there is no way for the market to turn to a memory type other than CMOS. A few months ago we put a bit of light in the dark by talking about what was supposed to be a substitute for normal DRAM, MRAM. But it is that everything progresses at a rate that even this can become obsolete in just a few years and without debuting. A series of materials with which it is working can do more and better, antiferromagnetic .
It has just started and as such it is still under study, but it is that the data is so promising that MRAM and all the effort of the industry can be left in borage water if the developments finally go to fruition.
This MRAM or magnetoresistive random access memory is really close to its final phase and it was planned that in a few years the first models would be on a production ramp and perhaps they would begin to be implemented on servers, since the speed is similar to that of RAM , it is not volatile, consumes less energy and would be cheaper to manufacture. But … Maybe I don’t even have a chance.
It all starts with a series of very specific and completely new materials
MRAM, as we saw in its corresponding article, has a very curious property: the storage of information is done as electron spins that look towards one direction, all in the same way.
When a current crosses or approaches them, its magnetic field can cause the spins to change direction, where the magnet registers the change and depending on it marks a zero or a one. Obviously, MRAM has the classic disadvantage of an HDD: they cannot be influenced by external magnetic fields.
Besides they have to grow in density and be more precise when registering the changes, but they are on the right track for it. As we say, your problem is precisely antiferromagnetic materials, a new wave that can change everything.
These materials are composed of a multitude of other commons, such as tin, platinum or manganese, and when combined they result in a new material that achieves a new property: the spins do not all face the same place, but instead point one against the other. others and cancel each other.
The first results are promising, all the advantages and none of the problems
This orientation of the spins starts with the premise that they can record bits of information without problems, but does not include a magnetic field like MRAM and therefore there is no magnet to record the changes.
Therefore, this type of materials is not influenced by magnetic fields and also the change and rotation of the spins is done at a really incredible speed, in picoseconds at working frequencies that are science fiction like the terahertz . This is much faster than MRAM and infinitely faster than conventional RAM, which works in nanoseconds and Gigahertz as we well know.
The first results in the laboratory affirm that these materials could increase the speed of writing by three orders of magnitude today and other things being equal. The next step after discovering antiferromagnetic compounds has been to influence them at will, which has required various studies that have ended up affirming and demonstrating that it is possible to use a low electric current to control the spins of the electrons.
Curiously, and being a paradigm in the industry that they are opening, they have discovered that there are such a number of options to manufacture different antiferromagnetics that it is both a solution and a problem due to the number of these.
Weyl’s fermions, a first step in creating the first memory with antiferromagnetics?
From the University of Tokyo, a compound called Mn3Sn was discovered that generates what they have called Weyl fermions.
The condition of these fermions is very easy to measure and allows a device like a new type of memory to be simpler in construction, which should save manufacturing costs for the industry. Due to the amount of materials being discovered within antiferromagnetics , researchers have to continue studying the properties of each one to opt for a winner and thus be able to guide the industry towards the construction of a type of memory that is manufactured and compatible with CMOS .
As far as is known, large companies in the computer sector are already knocking on the doors of universities that continue with these types of projects, so we are not talking about a technology that the industry looks at suspiciously, but rather as an opportunity to promote the performance of the biggest bottleneck of any PC or server in the world.
We also do not know when the first memories with these materials will be ready, but if the investment of any giant in the sector occurs, everything will accelerate and the competition will do the same for their benefit, we will simply have to wait to see what all this leads to.