A Beginner’s Guide
‘Semiconductor memory’ generally refers to digital data storage created within memory cells on a metal oxide layer that is manufactured on a silicon Integrated Circuit (IC) chip.
Much like our own human memories, the memory embedded in a semiconductor chip provides essential functions: encoding information, storing it, and retaining it so that it can be recalled for future use.
There are two basic types of semiconductor memory: volatile memory and non-volatile memory (NVM). They are each used in different parts of a system due to their specific properties.
Imagine if every time you went to sleep, you forgot everything you had done or learned previously. This is the nature of volatile memory – such as DRAM (Dynamic RAM) and SRAM (Static RAM) – where data is only retained so long as a device is connected to a power source.
DRAM is most often found as a system’s main memory, and its information must be refreshed continuously to retain data. SRAM does not need continuous refreshing like DRAM, but it is far more complex and expensive. It is used for applications that need more speed – primarily within the processors in a system.
DRAM and SRAM are basically used as working memory where system data can be accessed quickly and used while the system performs operations. But since the data is lost when the system is powered down, we need other types of memories to retain information. This is where Non-Volatile Memory comes in.
Unlike volatile memories which lose their information when the power is removed, Non-Volatile Memory (NVM) retains information such as program and data even when there is no power being supplied to the system. The term ‘NVM’ encompasses products such as hard disk drives (HDDs), solid state drives (SSDs), tape drives and computer memory sticks.
NVM chips that are widely used today include NAND flash, which is commonly used for on-device data storage in a broad range of automotive, and industrial applications; and NOR flash, which is widely used to store controller code and data in a range of applications.
Resistive Random Access Memory (ReRAM or RRAM) is an emerging NVM technology. It is designed for next-generation products that can benefit from its speed, low power consumption, endurance, and many other advantages.
Flash memory is a widely used NVM technology, and it represents a fast-growing segment of the overall memory market. However, embedded flash memory has limitations. As the semiconductor industry continues its ever-forward march to manufacture chips in smaller and smaller process geometries, embedded flash technology is unable to scale to the most advanced process nodes along with the rest of the chip.
The industry has alleviated some flash scalability challenges by moving from planar (two dimensional) flash to 3D stacking of flash arrays. But while this works for discrete chips, it adds a huge amount of complexity when the memory is embedded within a larger System-on-Chip (SoC), especially as the number of layers continues to increase. Given the limitations of embedded flash, a 28nm process may be the last where it can work.
What the industry needs is an embedded NVM technology that can be easily and cost-effectively manufactured at the most advanced process geometries. In this way, manufacturers can reap the cost and power benefits of continued scaling, while meeting ever-increasing performance requirements. ReRAM meets this need and presents a compelling successor to flash technology.
At the most basic level, the way semiconductor memory technologies work is by storing electronic data as binary information (‘1’s or ‘0’s) within a memory cell. The memory cells – which each hold one bit of data – are arranged into groups called ‘Words’ which each have memory addresses to identify them. These addresses are key for ‘Read’ and ‘Write’ operations. In a Write operation, data is stored into a Word, and in a Read operation, the chip recalls data from the Word, according to its identified address.
Memory technologies like DRAM, SRAM and flash all store data as an electrical charge. Each technology does this in its own way: with DRAM, each bit of memory is stored in a tiny capacitor; SRAM uses bi-stable latching circuitry; and flash memory stores the charge on a floating gate. Because all three of these technologies – DRAM, SRAM, and flash – store data as electrical charges, they all face challenges when it comes to scaling to the most advanced process technologies. These types of memory cells have gotten nearly as small as they can get. Memory technologies that store data as a charge also tend to be less tolerant to radiation, making it more complex to use them in a variety of medical, industrial, aerospace and other applications.
The industry is now looking for alternatives to storing data as an electrical charge. In particular, the future lies in storing data as resistance.
ReRAM and some other emerging NVM technologies like Phase Change Memory (PCM) and Magnetoresistive random-access memory (MRAM) store bits as resistance. Each of these technologies uses a different technique to reversibly change the resistance of a material (i.e., a chemical element). PCM uses heat, MRAM uses magnetization, and ReRAM uses the application of voltage through an electric current. This ability to store data as resistance enables these technologies to scale to more advanced geometries than those that store it as an electrical charge.
ReRAM is basically a ‘memristor’ technology. In such a device, resistance can be programmed (resistor) and that data can be stored (memory). With ReRAM, the application of positive and negative voltages creates either a 1 (low resistance state) or a 0 (high resistance state) to store data in the memory cell. You can read more about how ReRAM works on our Technology page.
ReRAM combines the advantages of both RAM and flash: ReRAM is a non-volatile, extremely fast, low-power and cost-effective technology that can endure a significantly higher number Program/Erase cycles than flash memory.
Since its inception, Weebit has focused on developing a ReRAM technology that not only has the best performance on key metrics, but is also designed to be extremely cost-effective and easy to integrate into any given CMOS fab. The company decided from the outset to achieve this goal by creating ReRAM technology based on materials and equipment that are already used in fabs. Using standard materials and tools means it is straightforward and affordable for fabs to integrate Weebit’s ReRAM with their existing manufacturing processes. In contrast, emerging memory technologies such as MRAM, FRAM and even conventional ReRAM require fabs to make large capital investments for special equipment and often exotic materials.
Almost every electronic product requires NVM, making the applications for ReRAM quite diverse. Memories are widely used today in consumer electronics, computers, smartphones, tablets and enterprise storage. There are also significant opportunities for growth in emerging segments such as Internet of Things (IoT) devices, autonomous vehicles, drones, robotics, wearables, neuromorphic computing, deep learning and machine learning, and the list goes on. Within these applications, ReRAM can be used in various systems, including power management ICs (PMICs), microcontrollers (MCUs), edge AI and many others. Learn more about the applications of Weebit’s technology.