Here's what you need to know about the new memory MRAM!
Latest update time：2021-09-01 16:04
Reads：
Source: Content from "
CTIMES
", thank you.
Currently, several chip manufacturers are working on developing a new generation of memory technology called STT-MRAM, but this technology still faces many challenges in its manufacturing and testing.
STT-MRAM (also known as spin transfer torque MRAM technology) has gained market attention for combining the characteristics of several conventional memories in a single component.
In the years of development, it was found that STT-MRAM has the speed of SRAM and the stability and durability of flash memory.
STT-MRAM provides non-volatile storage in the chip through the magnetic properties of electron spin.
STT-MRAM attracts market attention
Although STT-MRAM technology seems to have its advantages, it is also highly complex, which is why its development process has taken longer than expected.
Samsung, TSMC, Intel, GlobalFoundries, etc. are continuing to develop STT-MRAM technology.
Nevertheless, chip manufacturers face some challenges in their wafer equipment, such as having to improve existing production equipment and upgrade it to support 28nm or 22nm or even newer nanometer processes.
Figure 1: MRAM structure diagram
In addition, testing will also play a key role in the production process.
STT-MRAM requires new testing equipment to test its magnetic field conditions.
In addition, more stringent testing processes are required at different locations in the production process, such as the production stage in the wafer fab, the test platform, or post-testing.
Even so, challenges remain.
MRAM testing creates new conditions when the MRAM chip is operated in a strong magnetic field.
In non-magnetic storage devices, this is not a concern.
However, for MRAM, the magnetic field in the environment becomes a new consideration.
Typically, strong magnetic fields are required to interfere with STT-MRAM during operation, which needs to be verified and solved.
The industry is currently paying close attention to STT-MRAM because this storage technology has begun to be introduced by customers in the embedded field during the product design phase.
STT-MRAM is not only able to run at high speed, but also has the feature of retaining data even when the power is turned off, and its power consumption is also very low.
Due to these characteristics, STT-MRAM is very suitable for application in the embedded memory market, and storage devices including PCs and mobile devices are also paying close attention to the development of STT-MRAM.
Higher density and lower power consumption
Compared with conventional components (Toggle MRAM), STT-MRAM can achieve higher density, less power consumption, and lower cost.
Generally speaking, the main advantage of STT-MRAM over Toggle MRAM is the ability to expand STT-MRAM chips to achieve higher density at a lower cost.
Because STT-MRAM is a high-performance memory that is enough to challenge existing DRAM and SRAM, it is very likely to become an important memory technology in the future.
It is expected that STT-MRAM can be expanded to a process below 10nm and challenge the lower cost of flash memory.
Figure 2: STT-MRAM architecture description
STT stands for Spin Transfer Torque Structure.
In STT-MRAM devices, a spin-polarized current is used to flip the spin structure of electrons.
This effect can be achieved in a magnetic tunneling junction (MTJ) or spin valve. STT-MRAM devices use STT-MTJs, which generate spin-polarized currents by passing currents through a thin magnetic layer.
This current is then directed into a thinner magnetic layer, which transfers angular momentum to the thin magnetic layer, thereby changing its spin.
Conventional STT-MRAM structures generally use planar MTJ (or iMTJ).
Some STT-MRAM devices use an optimized structure called perpendicular MTJ (pMTJ), in which the magnetic moment is perpendicular to the surface of the silicon substrate.
Compared with iMTJ STT-MRAM, perpendicular STT-MRAM is not only more scalable, but also more cost-competitive.
Therefore, STT-MRAM with pMTJ structure will be a better solution to replace DRAM and other storage technologies in the future.
Targeting the embedded memory market
MRAM has the property of rotation, where the rotation of electrons changes its direction through the applied current, and the time for the change of direction has quantum properties, which depends on the angle of rotation.
STT-MRAM is also prone to changes, which may cause some reliability problems.
The biggest challenge facing STT-MRAM is the so-called read interference.
Another problem lies in the process.
Today the industry is developing MRAM at 28nm or 22nm.
There is no doubt that STT-MRAM technology can be extended from the 2xnm node to the 1xnm node.
However, whether it can continue to expand to 7nm or 5nm remains to be seen.
Despite this, the development of STT-MRAM shows no signs of slowing down, and is targeting two major application areas, namely embedded memory and independent memory.
Currently, some manufacturers are focusing on the development of embedded MRAM.
To give an example of its importance, a microcontroller (MCU) usually integrates multiple components on the same chip, such as arithmetic units, SRAM and embedded flash memory.
This embedded flash memory has the non-volatile characteristics of NOR, and this NOR flash memory is usually used for program code storage.
Currently, the industry has launched 28nm MCU products using embedded NOR flash memory, and some manufacturers in the R&D stage have begun to use 16nm or 14nm chips.
However, some experts believe that it is difficult to expand embedded NOR flash memory in the process range below 28nm. Many people believe that 28nm or 22nm will become the limit of this flash memory because the high cost will limit its market acceptance.
And this is where embedded STT-MRAM fits in.
It is suitable for replacing embedded NOR flash memory at 28nm or 22nm and even beyond.
In addition to this advantage, STT-MRAM can also replace or enhance SRAM in MCU, microprocessor or SoC systems.
A new wave of storage is coming
According to the survey, the automotive market and the Internet of Things market are the areas with the highest growth momentum for MRAM.
Many experts predict that MRAM will bring the next wave of storage.
The characteristics of MRAM, including low power consumption and persistence, are the main reasons why MRAM has extremely high flexibility in many applications.
For example, MRAM can be used in extremely low-power designs, such as wearable devices, or RFID applications (such as smart tags or trackers), and also includes edge computing and cloud applications, which can also meet its performance requirements.
Another example is the data center, because power consumption accounts for the highest proportion of the overall operating cost of the data center.
Figure 3: MRAM is considered the most suitable storage technology for machine learning.
Currently, MRAM has three main application markets. One is used as embedded memory. The characteristics of MRAM are very suitable for use as embedded memory, especially when embedded or integrated in MCU.
In addition, high-density MRAM is suitable for use as system temporary memory, accelerating NAND flash memory, or as a substitute for SRAM applications.
In the future, MRAM is likely to replace DRAM.
MRAM is very suitable for mission-critical applications for enterprise customers, which can solve problems including power loss and file loss, because once these problems occur, they may seriously affect the client's usage.
MRAM and other next-generation memories are also considered to be the most suitable storage technologies for machine learning.
Today, most machine learning systems use traditional memory, which is very power-intensive.
According to research, a large part of the power consumed in the machine learning process is consumed in the simple data movement process rather than the actual computing function.
For the machine learning process, any performance improvement will help improve the ability of machine learning.
Therefore, compared with existing DRAM products, any reduction in power consumption and the long-term stability of the technology will help improve the overall performance of machine learning.
*Disclaimer: This article is originally written by the author. The content of the article is the author's personal opinion. Semiconductor Industry Observer reprints it only to convey a different point of view. It does not mean that Semiconductor Industry Observer agrees or supports this point of view. If you have any objections, please contact Semiconductor Industry Observer.
Today is the 2081st issue of content shared by "Semiconductor Industry Observer" for you, welcome to follow.
Recommended Reading
★ Ranking of the top 20 global industrial chip manufacturers: Two European giants are the highlights
★ Is the new round of “operations” in South Korea’s semiconductor industry effective?
★ Alibaba releases the world's most powerful AI chip, and its core cloud strategy finally takes shape!
Semiconductor Industry Observation
" The first vertical media in semiconductor industry "
Real-time professional original depth
Scan the QR code , reply to the keywords below, and read more
AI｜ RF ｜ Huawei ｜ CMOS ｜Wafer｜ Apple ｜ Storage ｜WiFi 6
Reply
Submit your article
and read "How to become a member of "Semiconductor Industry Observer""
Reply Search and you can easily find other articles that interest you!
Featured Posts
京公网安备 11010802033920号