Gazing into a crystal ball can be a risky business, but amidst the haze one thing seems clear about the future of non-volatile memory: 2017 will see a big push by Resistive Random-Access Memory (ReRAM) manufacturers into mainstream commercial applications, leveraging advantages such as very-high-storage density, low power consumption and long endurance that have been tantalizing potential users for several years.
ReRAM uses a passive two-terminal electrode that stores data by means of ions that change electrical resistance states. The ions dissolve and then precipitate again changing the electrical resistance, ReRAM is designed to express only the property of an electronic component that lets it recall the last resistance it had before being shut off. ReRAM is based on the "memory resistor" concept and operates by changing the resistance of a special dielectric material called a memristor (memory resistor), a term coined by University of California, Berkeley scientist Leon Chua in the early 1970s, whose resistance varies depending on the applied voltage.
A memristor’s dielectric material does not become permanently damaged or fail when dielectric breakdown occurs; instead, the dielectric breakdown is temporary and reversible. When voltage is applied microscopic conductive paths called filaments are created in the material. The filaments can be broken and reversed by applying different external voltages. It is this creation and destruction of filaments in large quantities that allows for storage of digital data. Materials that have memristor characteristics include oxides of titanium and nickel, some electrolytes, semiconductor materials, and even a few organic compounds.
The principal advantage of ReRAM over other non-volatile technology is high switching speed. Because of the thinness of the memristors, it has great potential for high storage density, greater read and write speeds, lower power usage, and cheaper cost than flash memory. Flash memory cannot continue to scale because of the limits of the materials, so ReRAM is a good candidate to eventually replace it. Devices with mounted ReRAM should be able to achieve high-speed rewriting and longer operational times in battery-powered equipment.
Market analysts expect sales of ReRAM chips to experience substantial growth this year. For one thing Intel and Micron will begin shipping a ReRAM product called Optane, their brand name for a technology called 3D Xpoint, which, according to Intel, is 10 times denser than DRAM and 1,000 times faster and more durable than flash storage.
Optane removes the need for bit-storing transistors and instead employs a latticework of wires that use electrical resistance to signify a 1 or a 0. The first 16GB and 32GB Optane storage products will work only on PCs with Intel’s latest (Kaby Lake) processors. Intel reports that the low-capacity Optane storage will ship in the second quarter of this year. The PCs that these chips are aimed at (such as Lenovo’s new ThinkPad T570) were announced last week at CES 2017 in Las Vegas. Some of Intel’s PC customers were quoted by the press attending the show as saying a PC with a hard drive as primary storage and an Optane cache could load the PC’s operating system (OS) and applications faster than an all-solid-state-drive (SSD) system. Intel has said Optane could be up to 10 times faster than conventional SSDs.
There are different approaches to implementing ReRAM, based on different switching materials and memory cell organization. Crossbar’s ReRAM technology, for example, is based on non-conductive amorphous silicon (a-Si) as the host material for a metallic filament formation. The switching mechanism is based on an electric field, making Crossbar ReRAM cell behavior stable across a wide temperature range.
Crossbar notes that its patented selector device solves one of the technical challenges faced by developers of high-density ReRAM called the sneak current (or leakage current). Its 3D ReRAM storage solutions are based on 1TnR arrays (1 Transistor driving n Resistive memory cell), making it possible for a single transistor to manage a very large number of interconnected memory cells, which in turn enables very high capacity solid-state storage. The company claims 100x lower read latency than NAND, 20x faster write than NAND and without any block erase design constraints and timing limitations.
ReRAM momentum has been building for some time. Last year Hewlett-Packard and SanDisk announced an agreement to jointly develop ReRAM for use in enterprise cloud infrastructures that could replace DRAM and would be 1,000 times faster than NAND flash. This product will be based on HP's memristor technology. SanDisk expects the first ReRAM chips to appear in enterprise storage products in 2018. Also last year Panasonic came out with electronic shelf labels (ESL) that are battery-less and use NFC and an E Ink display. The ESLs are powered by Panasonic's 8-bit MN101 MCUs that use 64kb ReRAM memory.
With all of the optimism it is necessary to put things into perspective. NAND flash is not going away: flash represents about $30 billion a year in sales while ReRAM will only capture about $100 million in revenue this year.
Still, using a single layer of memory cells NAND flash cannot be scaled down much farther on the die’s surface — cell-to-cell interference causes a reduction in the reliability of planar NAND flash products. NAND, however, isn’t sitting idly by. By stacking layers of data storage cells vertically, 3D NAND technology creates denser storage than traditional two-dimensional memory. Increasing the storage capacity brings significant cost savings, lower-power usage, reliability, speed, and overall performance gains. Stacking NAND cells with precision could have a dramatic impact by keeping flash storage solutions competitive if not dominant for a few more years. But the numbers game will catch up to it: by 2020 the amount of IoT devices is expected to increase to over 20 billion and many of those IoT devices will need extremely low-power, high-performance non-volatile memory to store data.
In addition, NAND flash program operation is relatively slow; current MLC/TLC NAND or 3D NAND flash needs about 600µs to 1ms to program an 8 to 16Kbytes page. NAND flash also has to be erased prior to being programmed. The NAND erase operation is slow, too, in the 10ms range and is done for a very large block size, 4-8Mbytes.
Recently researchers at Nanyang Technological University, Singapore (NTU Singapore) in collaboration with Germany’s RWTH Aachen University and the research center Forschungszentrum Jülich demonstrated that in addition to data storage, ReRAM can be used for logic operation and computation.
Conventional computing devices have to transfer data from memory storage to the processor unit for computation, while the new NTU circuit saves time and energy by eliminating these data transfers. The NTU researchers say that these types of devices could be at least two times faster than current processors. At present computer processors use the binary system, which is composed of two states – either 0 or 1. However, the prototype ReRAM circuit built by the research group led by Asst. Prof Chattopadhyay (NTU School of Computer Science and Engineering) and his collaborators created ReRAM that processes data in four states instead of two ( 0, 1, 2, or 3, known as Ternary number system). Because ReRAM uses different electrical resistance to store information, it should be possible, they reasoned, to store the data in an even higher number of states, hence speeding up computing tasks. By making the memory chip perform computing tasks, space can be saved by eliminating the processor, leading to thinner, smaller and lighter electronics. Their discovery was published in Scientific Reports, a journal produced by the Nature Publishing Group.