The promise of resistive memories
As early as the 1960s, the resistivity of some types of materials has been studied by research organizations around the globe to determine whether these materials could be used as non-volatile memory technology.
The ability to store data as resistance – rather than storing it as an electrical charge like DRAM and flash do – is an appealing proposition. Some of the memory cells that store data as a charge face challenges in scaling to the most advanced process geometries along with the rest of an SoC. They also tend to be less tolerant to radiation, making them more complex to use in a variety of medical, industrial, aerospace and other applications.
Over the years, academic institutions and research centers have continued to make progress in the development of resistivity-based memories. Then, in the early 2000s, when it became clear that there would be a limit to the scalability of existing memories, the industry began to invest in these this research. The quest for next-generation memories had begun.
The Emergence of ReRAM
With the idea of replacing flash non-volatile memory (NVM) technology, we began to see industry developments around Resistive RAM (ReRAM or RRAM) starting in the early 2000s from companies such as IBM, HP, Panasonic and others. We’ve also seen some companies make major announcements about ReRAM which never materialized into functioning products. Many of these efforts eventually stalled, and 3D stacking of standalone flash entered the picture around 2015, making it possible to delay dealing with flash’s scaling problem.
However, the industry knew that they were just buying time, and that ultimately cost, power and other issues would force the adoption of a new NVM. The limitations of flash are increasingly clear. Even with 3D stacking, advanced packaging and chiplet architectures, there are significant cost, power, and security difficulties to overcome. It’s just not economically feasible to embed flash memories into SoCs beyond 28nm for most applications. Flash also suffers from other technical limitations in terms of performance, reliability, power consumption and cost.
Weebit enters the scene
Weebit was founded in 2015 with a focus on creating a new memory that would be scalable for future generations of electronics. The company has worked closely since that time with our R&D partner, CEA-Leti, one of the world’s most advanced microelectronics research institutes, to create a truly innovative NVM technology based on over a decade of research.
Above: The Weebit Nano logo from 2016, with a focus on the ‘future’
A May 2015 article in TechTarget discusses the potential future for memristor devices (like ReRAM), noting, “The technology certainly has the potential to replace flash, but whether it immediately becomes a successor will ultimately depend on the technology’s economic viability.”
This economic viability is at the forefront of Weebit’s development efforts, since even the most advanced technology can’t succeed if it isn’t affordable and easy for customers to integrate and manufacture. Since the beginning, Weebit has focused on creating technically excellent ReRAM technology that is commercially viable. This has meant focusing first on developing IP solutions for the embedded market using standard materials and processes. In this way, we can intercept the market need for flash where it is first emerging (in the embedded space) and do so with fab-friendly and easy-to-integrate technology.
Weebit has made steady progress toward this goal, with its first demonstrations of small ReRAM arrays in 2017 – showing functional RRAM cell samples and verifying the ability of the cells to maintain their memory behavior. This was followed in 2018 by a demonstration of a working 1Mbit ReRAM array and the production of the first packaged units containing memory arrays based on Weebit ReRAM.
Above: Different Weebit taglines have focused on ReRAM emerging at some undetermined point in time
The path to commercialization
In 2020, commercialization activities came to the fore for Weebit. The company’s technology was mature, and the technology stabilization process was completed, making it possible for Weebit ReRAM to transfer to a production fab. Then in 2021, we demonstrated production, testing and characterization of fully functional 1Mb ReRAM arrays in a 28nm FD-SOI process, followed by the introduction of silicon demonstration wafers integrating Weebit’s embedded ReRAM module. All of this led to our first commercial deal with SkyWater Technology.
With each milestone, we’ve been getting closer and closer to production. This year, we demonstrated our IP module for the first time. Most recently, we fully qualified our ReRAM module with CEA-Leti and received silicon wafers of our demo chip from SkyWater Technology – the first Weebit ReRAM wafers from a production fab. At the same time, we are continuing to accelerate our development toward more advanced process geometries such as 22nm FD-SOI.
Importantly, ReRAM isn’t just a one-off technology from Weebit. The industry is getting behind ReRAM technologies as a way forward for many applications. We see companies working on their own forms of ReRAM such as foundry TSMC, which is said to be in production with its own ReRAM technology. This is a strong validation for our own technology – showing that ReRAM is an effective solution for a clear market need. According to market research firm Yole Group, the embedded emerging NVM market is expected to reach $2.9B by 2027, and ReRAM is expected to comprise 33% of that.
Source: Yole Emerging Non-Volatile Memory 2022*
As we look ahead, one thing is clear – ReRAM is here. It is no longer ‘emerging’ or ‘future’. The technology is functional, industry-qualified and ready for production. Importantly, it’s able to deliver the promised advantages in terms of cost, power, endurance and reliability. It is the successor to flash, and it is here today.
We’re in discussions with potential customers and partners who are interested in getting an early mover advantage with Weebit ReRAM. If you have a design project, you too can get started with Weebit ReRAM. On the SkyWater website, you will find the preliminary specifications of Weebit ReRAM in SkyWater’s 130nm CMOS process. Contact SkyWater or Weebit to start discussing your design needs.
Above: the new Weebit Nano logo and tagline, reflecting ReRAM’s market entry
*Note: The embedded emerging NVM market size is evaluated based on assumptions of the average chip area occupied by a given memory technology