The internet of things (IoT) has the potential to drive applications worth trillions of dollars – from consumer wearables and household appliances, to industrial equipment and automobiles.
According to IDC, close to 30 billion devices may be part of the IoT ecosystem by 2020, generating as much as 40 zettabytes of cloud-based data.
Several factors are driving the rapid growth of IoT. Moore’s law is still at work creating faster, smaller, cheaper devices. Other technological advancements like low power and lower cost communications that can interconnect tiny sensors with cloud-based applications enable new business models that harvest greater value from sharing information.
But there are obstacles. More connected devices means more data is created every second that is stored, processed, uploaded and shared. The abundance of sensors including temperature, pressure, direction, speed, weight, paces, heart beats, light intensity, etc. will generate a flood of information to transmit. And continuous wireless data transmission is a power-hungry operation requiring far too much energy for battery-powered applications.
Next generation IoT devices will be low-cost, highly integrated systems requiring specialized SoCs with more functions on chip. Most will need to run on a single battery charge for years, or even harvest energy from their environment. This is particularly critical in remote applications, such as smart sensors deployed across wide areas like cities, where regular battery changes would be impractical.
Meeting the energy budget will necessitate a variety of energy-saving strategies and many devices will be designed to spend most of their time in standby or other energy saving modes, active only to perform necessary functions. Designers will be looking for ways to reduce not only the volume of data to be transmitted, but also the frequency and duration transmission to conserve power.
Innovative memory technologies can help address the most critical IoT energy challenges. Lower power and lower voltage operation, monolithic integration, faster read and write times, nonvolatility and higher capacity are all ways that memory technology can help IoT devices to achieve greater energy efficiency.
With such restrictive energy budgets, memory will need to operate on far less power and be more integrated than what’s possible with today’s memory technologies, which are too power hungry, slow, unreliable and difficult to manufacture.
Nonvolatile memory technology is an obvious choice for lowering IoT device energy consumption. By design, nonvolatile memory can be completely powered down, yet retain all stored information.
More powerful, yet lower-power microcontrollers make pre-processing a viable way to reduce the volume of sensor data to be transmitted. However, efficient processing necessitates greater local memory capacity for stored data to be processed and programs to execute.
To reduce the frequency of data transmissions, designers will make greater use of local data buffering, as batching data for transmission allows the frequency of transmissions to be significantly reduced. Faster read and write times enable data to be sent more quickly, reducing the duration of each transmission, and optimizing on/off duty cycles.
More memory will also be needed to ensure interoperability, as the lack of a globally accepted IoT connectivity standards will require devices to support multiple protocol stacks. Today’s on-chip memory technologies do not scale economically to greater densities, so SoC designers are forced to rely on external chips for this additional memory.
Software is never perfect and market requirements are constantly changing, which results in a constant flow of new and improved features, bug and security fixes, and other updates common in today’s software driven hardware. When billions of IoT devices are in the field, it will be vital to ensure that reliable over the air (OTA) firmware updates are possible, requiring additional memory capacity, allowing two or more firmware images to be stored for each device.
Resistive random-access memory (RRAM) is widely hailed as the most promising technology in the race to develop new, more scalable, high-capacity, high-performance and reliable storage solutions. RRAM cells typically employ a switching material sandwiched between two metallic electrodes that can exhibit different resistance characteristics when a voltage is applied across it. Significant performance differences can be achieved depending on the switching materials and memory cell organization chosen.
Unlike flash, RRAM, such as that pioneered by Crossbar, Inc., is byte addressable and can be architected with small pages that can be independently erased or reprogrammed. On-chip storage drastically simplifies complexity of the microcontroller by removing a large portion of background memory access required for flash-related data management. It also enables the use of wide memory buses that break the bandwidth bottleneck between computing cores and storage. With a write amplification equal to one, RRAM achieves visible benefits in read and write latencies, lower energy consumption and increased lifetime of the storage solutions.
At the memory cell level, RRAM improves programming performance and power consumption and, on a system level, on-chip storage memory reduces energy consumption by 50X. Lower, more predictable latencies also reduce power consumption by shortening the execution time of code fetching or data streaming.
Monolithic integration of storage memory eliminates the need for a variety of mechanical connectivity components and methods of varying complexities that can lower yields and increase overall fabrication cost. Moreover, in the case of Crossbar’s solutions, RRAM technology can easily integrate with CMOS logic circuitry and be manufactured using existing CMOS fabs. More on-chip storage enables data logging applications to achieve energy savings of up to 40X compared to frequent wireless data uploads.
RRAM solutions can enable radical innovations in the connected IoT device world. Low energy, fast, nonvolatile memory that can be easily integrated in very large capacities on a single SoC along with logic, analog and RF components that can operate for years without a battery charge or change providing the power savings needed to enable the future of IoT devices across consumer and industrial applications.