Non-volatile memory is one of the most important aspects of computing but not many people could give you a quick definition of this technology. As opposed to volatile memory, which requires its memory data to be refreshed periodically, non-volatile memory is capable of saving data even without any power. Non-volatile memory is what enables you to use your computing device, whether a smartphone or a laptop, without having to download and install the same programs every time you want to use them. Volatile memory such as random access memory (RAM) stores data while in use by other applications, but it’s the non-volatile memory card that actually stores files and data when they’re not in use.
Many readers will be aware that the data capacity afforded by a RAM chip has much to do with the overall speed of a computing system. More RAM means quicker recall of stored data. However, RAM wouldn’t be as necessary if the non-volatile memory chip was faster and allowed users to access data sooner on command. Currently, the most commonly used non-volatile memory chip are negative-AND (NAND) flash, which utilize a specific configuration of memory cells so that cells can share source and drain regions, reducing the overall size of the NAND chip. Unlike mask programmable read-only memory, which is used on some integrated circuits and microcontrollers, NAND memory can be rewritten many times, making it a good fit for consumer computing devices which often have their data rewritten daily.
NAND flash memory is not a young technology, however. The use of flash memory, named for its ability to write data or erase data quickly, was pioneered in the early 1980s by Dr. Fujio Masuoka, at the time a factory manager with Tokyo-based Toshiba Corp. (TYO:6502) He was tasked with developing dynamic random access memory (DRAM) chips but was interested in creating a product that could store data even when the power is off. Over the course of decades, flash memory became more valuable with the rise of solid state devices (SSDs) for consumers such as tablets or flash memory sticks. In 2013, Samsung Electronics Co. (KRX:005930) was the world’s largest manufacturer of NAND flash memory chips, enjoying sales of more than $5.5 billion buoyed by a nearly $8 billion global market for SSDs.
The replacement for NAND memory chips may be on the way. A partnership between Intel Corporation (NASDAQ:INTC) and Micron Technology, Inc. (NASDAQ:MU) has resulted in the development of a new non-volatile memory called 3D XPoint (pronounced “crosspoint”). The companies say that the non-volatile memory is both 1,000 times faster and has 1,000 times greater endurance than conventional NAND flash. It would be the first major upgrade to non-volatile memory options since NAND flash was first commercialized in 1989.
The improved characteristics of 3D XPoint compared to NAND flash are on an entirely different order of magnitude. The increased endurance means that 3D XPoint would be able to undergo millions of write cycles whereas NAND can only handle tens of thousands before the component starts to deteriorate. Both Intel and Micron compare the speed gains of the memory chip to reducing the amount of time that it takes to travel on a flight from San Francisco to Beijing from 12 hours down to 43 seconds.
Remarkably, the architecture of the Intel/Micron memory chips is such that no transistors are required. Individual two-layer wafer cells, each containing a memory cell and a selector, are connected through a series of wires which are layered in a perpendicular fashion, traveling vertically across memory cells on one side and horizontally across on the other. Instead of using a transistor to switch an electric signal, 3D XPoint achieves reading and writing by varying the voltage transmitted to individual cells from each wire.
A very good metaphor describing why a faster programmable memory chip is important comes to us from the online technology news publication Gizmag. The process of computing can be compared to a relay race team trying to pass along data like a baton. Processors have gotten faster and faster over the course of years but computing speeds can still be held back by slow data recall. If 3D XPoint technology works as advertised, it would be akin to taking the slowest member of that relay team and replacing that runner with Olympic gold medalist Usain Bolt. Today’s processors are incredibly fast and could complete a great deal of computational work than we typically call upon them to perform, but the relatively lower speed of NAND serves as a bottleneck on operations.
The new memory chip technology is also 10 times denser than conventional memory options utilized in computing devices. One individual wafer of these chips could store up to 128 gigabytes of data while operating at speeds measured in nanoseconds instead of microseconds. As devices continue to reduce the size of their form factors, this innovation would continue to support that while enabling incredible speeds of computation.
It’s not as though NAND flash memory will be phased out tomorrow and 3D XPoint still isn’t as fast as DRAM for read/write operations. But this new chip could have an incredible impact to functionality in just about every area of computing in a way that’s much more complex than a simple increase in processing speed. The specifications of 3D XPoint make it powerful enough that, when coupled with some form of RAM, it could support the quick recall of vast stores of data required for immersive social gaming experiences or quick genomic analysis of medical patients. Although DRAM still edges out 3D XPoint in terms of speed, the incredible increase in the number of write cycles that can be sustained enables the Intel/Micron invention to provide both system memory and storage memory, replacing both NAND flash for storage and DRAM for random access memory.
This development comes at a crucial time during these early days of the Internet of Things. Memory speeds were already proving to be a constraint on processor operations in 2013, when the world generated a total of 4.4 zettabytes, or the equivalent of 1,000,000,000,000,000,000,000 bytes. By 2020, that annual global data generation rate is expected to climb to 44 zettabytes, an increase by a power of 10. By 2050, when it is expected that 50 billion devices will be outfitted with computing processors for digital services, the amount of data generated every year could skyrocket. 3D XPoint, which can write up to 40 terabytes per day, will be much more suited for that atmosphere than SSD NAND, which can write up to 40 gigabytes in a day. We’ve already talked about how Bluetooth wireless communications technology is well situated to benefit from the boom in Internet-enabled devices. Where Intel and Micron’s non-volatile memory innovation is concerned, the timing could not have been more impeccable.