Where Weebit ReRAM Can Make a Difference
As nearly every electronic product requires NVM, the applications for Weebit ReRAM are numerous and varied. While the most popular application sectors for memories are computers, consumer electronics, smartphones, tablets and enterprise storage, there are tremendous growth opportunities in innovative and emerging segments such as Internet of Things (IoT) devices, automotive systems, industrial automation, autonomous devices, drones, robotics, neuromorphic computing, deep learning and machine learning systems, wearables, and many others.
Mixed-Signal, Analog and Power ICs
Analog ICs represent a diverse set of technologies including sensors, power management ICs (PMICs), mixed-signal ICs, RF, MEMS, and others. These devices increasingly require embedding low-density NVM for different purposes. First, such mixed signal designs require specialized NVM to store trimming and configuration data, as well as unique IDs. These should be programmable to allow product variations, and should be field upgradeable as well.
In addition, some analog and power management designs today require the integration of a microcontroller (MCU) for added intelligence, including smart sensing and measurement on end nodes in areas such as medical and fitness devices, industrial automation and other ultra-low power IoT applications. These highly integrated devices not only need to store data and boot code, but must also run firmware updates, requiring high-performance, low-power NVM.
Analog and mixed-signal ICs need to support many device components, most of which are built in the Front-end-of-line (FEOL) of the manufacturing process, and they must support many process nodes and derivatives. Since embedded flash is also integrated in FEOL, companies must often make compromises with power and analog components in order to accommodate flash on the same wafer. This leads to degraded performance, larger size, and higher cost.
ReRAM is integrated in the Back-end-of-line (BEOL) of the manufacturing process, allowing full optimization of such components, so there is no impact on design rules. This also simplifies adopting ReRAM in a new fab as it can be adopted once for a geometry and it will work with all the different variants of that geometry, unlike flash which needs to be adapted to each variant.
As a low-cost, high-performance, highly scalable BEOL technology, Weebit ReRAM enables key requirements for embedded NVM in analog ICs.
Aerospace and Defense
ICs for aerospace and defense have unique requirements for robustness, reliability at high temperatures, and tolerance to radiation (rad-hard) and electromagnetic fields. As these products are often required to last for years – mostly without maintenance – longevity is another key trait. Memory must be reliable for the lifetime of the product.
Weebit ReRAM has significantly better endurance than flash, ensuring it can support products with long lifetimes. It is also able to maintain its reliability at a broad range of temperatures, from (-55)0 Celsius up to 1750 Celsius. ReRAM cells are inherently immune to various types of radiation and electromagnetic fields. In fact, Weebit ReRAM can withstand 350x more radiation than flash. These features make Weebit ReRAM ideal for aerospace and defense applications.
Every day more devices become connected to each other, to networks, and to the cloud – delivering valuable insights into everything from our personal health to our farms to city infrastructure and beyond.
The IoT depends on massive numbers of small sensors reliably delivering streams of data from often remote or inaccessible locations. To enable these billions of battery-operated and energy harvesting IoT edge devices, embedded NVM must have:
- Ultra-low power consumption
- Low total cost of ownership
- Scalability to low process geometries for maximum cost reduction
- High data retention and endurance even in high temperatures or other severe conditions
- Speed and accessibility to enable the shortest possible active power usage
Weebit ReRAM enables these requirements for embedded NVM in IoT, with significantly higher performance and lower power than embedded flash, and much lower cost compared to alternative emerging NVM technologies. Weebit ReRAM also maintains high endurance and data retention in harsh field conditions.
As the world becomes increasingly connected, electronic payments gain in popularity, and new connected devices and technologies enter the mainstream at a rapid pace, security risks continue to rise. Whether it’s for smartcards, IoT devices, gateways, or other connected systems, the industry needs added layers of security – from software to the deepest embedded hardware – to prevent and circumvent increasingly clever hacking techniques.
The best security solutions begin during the semiconductor manufacturing process. Solutions such as physical unclonable functions (PUFs), true random number generators (TRNGs) and other such technologies are important components of SoC security. Integrated during the manufacturing process, a PUF generates signatures and keys unique to each chip, and a TRNG generates unpredictable strings of data to enable security algorithms.
ReRAM has inherent physical attributes that make it an ideal solution for such mechanisms. These features also aid security when ReRAM is embedded as a traditional NVM, keeping memory content, including data, logs, and code, safe from hacking. Since it is difficult to intrude, read or modify ReRAM, the technology shines as a secure solution for PUFs, TRNGs and embedded NVMs.
Edge Artificial Intelligence
Regardless of the specific application, storing weights for artificial Neural Networks (NNs) requires significant on-chip memory.
Depending on the network size, requirements typically range between 10Mb – 100Mb. For AI edge products where low power consumption is so important, what’s needed is small, fast, on-chip (embedded) NVM.
Although it is common and simple for near-memory computation, SRAM won’t work for these applications because it is extremely large and volatile. This volatility means it must stay connected to power, consuming a great deal of power and also risking data loss in the event that power is unexpectedly cut off. Given its size, it would also require additional off-chip NVM, leading to memory bottlenecks and power waste. On-chip flash memory is also far from ideal. As NVM, it can persistently hold weights even during power-off, but it can’t scale below 28nm as embedded on-chip memory. This means a separate chip is needed – leading to memory bottlenecks.
ReRAM is 4x smaller than SRAM so more data can reside locally. It scales well below 28nm, it is non-volatile, and it enables quick memory access. Weebit ReRAM is ideal for advanced edge AI chips.
Today’s vehicles integrate many hundreds of chips to control the various functionality – from the electrical control units (ECUs) inside of the engine, to powertrain, transmission and other control systems to the electronics that handle automotive driver assistance systems (ADAS), infotainment and other advanced features. Most of these chips require some form of non-volatile memory (NVM).
As vehicles become increasingly intelligent, in-vehicle systems must boot quickly and respond instantly to stimuli. They must also be able to handle frequent over-the-air (OTA) upgrades to ensure all systems are running smoothly with the latest feature and security updates. These systems need NVM with superior endurance that can execute code quickly, reliably, and securely, even in harsh environmental conditions.
Weebit ReRAM delivers the endurance, fast switching speed, high-temperature reliability (up to 175oC), and longevity needed for automotive solutions. Unlike embedded flash, Weebit ReRAM can scale to the most advanced process nodes, and it is easy and cost-effective to implement.
Storage Class Memory (SCM)
There is a big gap in today’s systems between the main memory (DRAM), which is very fast but volatile, and storage (flash/SSD), which is non-volatile but slow.
A new type of memory called Storage Class Memory (SCM) has emerged to address this gap in storage architecture. SCM, also called Persistent Memory, represents a new tier of memory/storage. It is faster and lower power than flash and less expensive than DRAM. It is also non-volatile, and is targeted to accelerate storage in data centers, cloud computing and mass storage devices.
As data usage and storage needs increase, the industry is looking for new memory technologies to address SCM. Weebit ReRAM is a perfect fit for SCM since it not only reduces cost compared to DRAM, but can deliver higher performance, better endurance and lower latency than NAND flash.
The use of Neural Networks is leading to exciting developments in artificial intelligence, with the ability to increase machine learning accuracy in areas like speech recognition and image classification.
However, reaching this high degree of precision using traditional computing architectures takes a significant power toll. Since much of this power consumption is related to data movement between a system’s computing elements and memory modules, the industry is investigating new technologies that can reduce this data movement. The solution lies in integrating dense, low-power NVM closer to the computing elements – also called In-Memory Computing - and Weebit ReRAM is an ideal candidate.
In addition, the next wave of AI and Machine Learning architectures will take a new approach – Analog Computing, or Neuromorphic Computing – whereby the computation is done within the storage element in an analog fashion. These architectures are designed to accurately emulate the brain’s operation and can therefore achieve orders of magnitude better power efficiency.
The Weebit ReRAM cell functions similarly to a synapse in the brain, making it a promising solution for neuromorphic computing. Many institutes are now studying this domain, which has the potential to become a huge market in the future.
We are collaborating with research partners – both academia and industry – to explore the possibilities of using ReRAM for neuromorphic computing. This includes ongoing projects with the Non-Volatile Memory Research Group of the Indian Institute of Technology Delhi (IITD), the Institute of Nanoscience and Nanotechnology (INN) NCSR ‘Demokritos’, the Politecnico di Milano and the Technion – Israel Institute of Technology, as well as CEA-Leti. Together with Leti we were the first in the industry to demonstrate ReRAM-based spiking neural networks (SNN).