STT-MRAM memory could enable super efficient computers and sensors

Most parts of present computer systems are made of volatile devices, and the power to supply them to avoid information loss causes huge energy losses. We can eliminate this meaningless energy loss by utilizing the non-volatile function of advanced spin-transfer torque magnetoresistive random-access memory (STT-MRAM) technology and create a new type of computer, i.e., normally off computers . Critical tasks to achieve normally off computers are implementations of STT-MRAM technologies in the main memory and low-level cache memories. STT-MRAM technology for applications to the main memory has been successfully developed by using perpendicular STT-MRAMs, and faster STT-MRAM technologies for applications to the cache memory are now being developed. The present status of STT-MRAMs and challenges that remain for normally off computers are discussed.

Normally off computers could use less than 1% of the power

The room temperature (RT) tunnel magneto-resistance (TMR) effect found in Al-O based magnetic tunnel junctions (MTJs) has enabled a new type of non-volatile memory, i.e., the magneto-resistive random access memory (MRAM). The concept of “ instant-on computers” has attracted attention around 2000 as an application of MRAMs. MRAMs were expected to reduce the start-up time of computers and to reduce user frustration. MRAMs play an important role only when computers start up in instant-on computers. However, we believe that the potential of MRAMs is not limited to start up and they have hidden potential to change the computer architecture. The researchers proposed the concept of ” normally off computers” in 2001 from this point of view.

Suppose that you are typing on a keyboard. During the approximately 100 ms to move your finger from one key to the next, the computer needlessly wastes energy waiting for your input. This is because most parts of present computers are made of volatile devices, i.e., transistors and dynamic RAMs (DRAMs), which lose information when powered off. The present computers are designed on the premise that power will always be supplied, i.e., they will be normally on. If computers are redesigned so that power consumption is zero during any short intervals when users are absent from the job without them even being aware of it, very energy efficient computers such as mobile personal computers running on solar batteries or hand-cranked dynamos can turn into a reality.

We need high performance non-volatile devices that do not require a power supply to retain information to create normally off computers and simultaneously guarantee sufficiently high speed operation to manipulate the information. The main memory, for example, requires performance as fast as 10 to 30 ns (image below) and density as high as 1 Gbit per chip.

Many problems still remain to be solved to achieve normally off computers. The problems are not only limited to materials and MTJ devices but circuits, memory architectures, operating systems, and peripherals, which should also be redesigned. This paper reports efforts to attain normally off computers and discusses the challenges that remain.

Layered structure of computer systems. Typical access times for smartphone, personal computer, and supercomputer systems are shown.
Citation: J. Appl. Phys. 115, 172607 (2014);

Journal of Applied Physics

Other Applications

In real high-end computer systems, a variety of peripheral circuits and devices are needed in addition to the architecture outlined in the image above. Data access frequency is low and fast operation speed is not required because they are located far from the processor core. There are also low-end processors that are not required to be so fast as high-end ones. Some devices such as configuration data memory for field-programmable-gate array (FPGA) and one-time-programmable memory for securing data do not need fast operation. Restrictions imposed by the dilemma become less stringent for such devices. The standby-power free advantage of STT-MRAM based non-volatile logic, for example, will find a variety of applications such as in sensor-network and health-care systems.

However, being free from the dilemma simultaneously means that the most important advantages of STT-MRAMs are lost, i.e., fast read/write speeds and infinite endurance against other types of non-volatile devices such as ferroelectric RAMs, phase change RAMs, resistive RAMs, and flash memories. Non-volatile logic devices based on these non-volatile memories are already commercially available. STT-MRAM based logic devices are required to demonstrate better cost performance against rivals. Once the mass production process for STT-MRAMs for main memory applications is established, it can easily be diverted to lower performance STT-MRAM based circuits. That will provide them with excellent cost performance.

There is also the need to develop normally off displays because displays in computer systems typically consume 20%–40% of the power of the systems. Although STT-MRAMs will not play important roles in displays, a variety of normally off type displays, such as electrostatic paper displays and micro-electro mechanical systems (MEMS) displays, are currently being developed

The concept and technical requirements for normally off computers were discussed. Spintronics based normally off computers have not yet been introduced 12 yr after the concept was first proposed. However, marvelous advances made in the last decade and collaboration with the field of computing science are now making normally off computers a reality.

If you liked this article, please give it a quick review on ycombinator or StumbleUpon. Thanks