Contact Us
Contact Us
Technology

Frequently Asked Questions

ReRAM/RRAM Basics

ReRAM (resistive RAM, also called RRAM) is a non-volatile, extremely fast, low-power and cost-effective non-volatile memory (NVM) technology that can endure a significantly higher number of Program/Erase cycles than flash memory. The Weebit oxide-based ReRAM (OxRAM) cell is comprised of a thin oxide switching layer between two electrodes. Resistance can be programmed using electric voltage. Read more about ReRAM in our Guide to ReRAM.

Weebit ReRAM (RRAM) is more environmentally friendly than other types of NVM. It doesn’t require any rare earth materials, doesn’t pose any contamination risk, and doesn’t require dedicated clean room space. It also requires fewer masks and fewer steps in the fab (both vs. MRAM and embedded flash), using less resources altogether. Because it consumes less power and requires fewer resources to be manufactured, Weebit ReRAM also has a smaller GHG (greenhouse gas) emission footprint. Read more here: Weebit ReRAM: NVM that’s better for the planet.

Resistive RAM (ReRAM or RRAM) is a type of Non-Volatile Memory (NVM) that is designed to be the successor to flash memory. As an embedded technology, designers integrate it as IP into their systems on chips (SoCs). ReRAM is highly scalable, enabling it to be used as an embedded technology at manufacturing processes below 28nm, where embedded flash can’t scale. As a back-end-of-line (BEOL) technology, ReRAM has an advantage over flash which is a front-end-of line (FEOL) technology. With flash, designers must often make compromises with analog components and other devices integrated in the front end. ReRAM doesn’t have this problem, and you can also adapt ReRAM once for a technology node and it works for all its variants. Weebit created its ReRAM based on the most common materials and equipment used in fabs today. This ensures that embedded Weebit ReRAM IP is low cost and easy to integrate into any standard CMOS flow, ultimately enabling manufacturers to quickly reach good yield with minimal investment.

The Weebit oxide-based ReRAM (OxRAM) cell is comprised of a thin metal oxide switching layer between two electrodes. Through an initial forming step, a conductive filament is created between the electrodes. Applying successive positive and negative voltages to the electrodes causes the memory cell to switch from a Low Resistive State (LRS) state to a High Resistive State (HRS), encoding binary information through either a 1 (LRS) or a 0 (HRS) data bit stored in the memory cell. The filament can be broken then reformed and broken again during successive cycles of erasing and programming. ReRAM’s ability to store data as resistance in this way enables it to scale to more advanced geometries than those that store it as an electrical charge (like flash). You can read more about how Weebit ReRAM (RRAM) works on our Technology page.

ReRAM (RRAM) is a non-volatile memory (NVM) technology that stores data by changing the resistance of a memory cell. All types of ReRAM are known for their high performance, low power consumption and ability to scale to advanced processes. The differences between different types of ReRAM are generally in terms of their switching mechanisms and materials, which can translate to differences in complexity and cost. Weebit ReRAM is an Oxide-based ReRAM (OxRAM) that uses a thin metal oxide film to form a conductive filament between two electrodes. It is based on the materials commonly used in semiconductor manufacturing today – making it easy and inexpensive to adopt and integrate. Conductive bridging RAM (CBRAM) is a different type of ReRAM that uses metal ions in solid electrolytes to form/dissolve conductive filaments. Some of the materials commonly used in CBRAM, like silver, can be expensive and require special handling.

RRAM and ReRAM refer to the same technology. TSMC calls it ‘RRAM’. Most of the industry says ‘ReRAM’. There’s no technical difference.

There are two types of ReRAM, which differ in how they store and switch data:

  • OxRAM (oxide-based) uses oxygen vacancies to form or break resistance paths. Weebit ReRAM and ReRAM from TSMC are both OxRAM technologies.
  • CBRAM (conductive-bridging) relies on metal ions to create a conductive bridge.

The key differences between various types of RRAM/ReRAM lie in the specific memory device used – the details of the stack, materials, thickness, tools, masks, and specific process steps. This is what eventually differentiates one ReRAM technology from another.

RRAM stands for Resistive Random Access Memory.

Read more in our Guide to ReRAM and RRAM.

The ReRAM Difference

It depends on your application, planned process and current NVM.

If you’re using embedded flash and targeting an advanced process (28nm or below), embedded flash just isn’t an option. Even in more mature nodes, you can gain significant advantage from using ReRAM (RRAM). Some considerations about ReRAM:

  • ReRAM is less expensive, only requiring 2 additional masks, vs. ~10 such masks for flash
  • ReRAM is ~100X lower power than flash
  • Programming ReRAM is faster by >10X, helping reduce latencies and leading to more efficient system architectures
  • ReRAM is naturally tolerant to radiation (flash isn’t)
  • ReRAM is bit and byte addressable, helping you architect smarter solutions
  • ReRAM is a BEOL technology, so unlike flash which is FEOL, there is no interference with analog components, making ReRAM ideal for high-voltage and BCD processes

If you’re considering MRAM, here are some facts:

  • MRAM is based on many non-standard, rare earth, materials, and requires a separate, dedicated, manufacturing line with special tooling – a big initial and ongoing investment
  • MRAM requires many very thin layers, adding more than 30% to the cost of a silicon wafer, whereas our ReRAM adds less than 10%
  • MRAM is based on magnetic technology, so memory could fail in the field due to magnetic interference
  • Using magnetic fields, MRAM can be more easily hacked than ReRAM
  • The only real advantage MRAM has is endurance, so it makes sense in some niche domains

Flash stores data as an electrical charge, and its memory cells have gotten nearly as small as they can get. It is simply not commercially viable below 28nm. ReRAM (RRAM) and some other emerging NVM technologies store bits as resistance and can therefore scale to more advanced geometries.

Weebit ReRAM demonstrates significant security-related advantages, enabling it to better protect its content from hacking attacks, and making it more difficult to reverse engineer. ReRAM (RRAM) is inherently able to withstand attacks via electron beams, optical lasers, magnetism, power analysis and other methods.

MRAM, for example, is much easier to hack due to its magnetic nature. 

Like ReRAM (RRAM), Magnetic random-access memory (MRAM) is a type of non-volatile memory (NVM) that stores bits as resistance, but the two technologies use a different technique to reversibly change the resistance of a material. MRAM works by changing the magnetic spin of electrons rather than directly storing charge. Its fabrication requires the use of non-standard magnetic materials, and it requires special tools – a big initial and ongoing investment for a fab. MRAM is also very expensive to manufacture since it requires many very thin layers, adding significantly to the cost of a silicon wafer. Its use of magnetic technology is also a challenge given widespread use of wireless chargers, magnetic doors, airport security, MRI machines, and even EVs. Unlike MRAM, Weebit created its ReRAM based on the most common materials and equipment used in fabs today – making it low cost and easy to integrate into any standard CMOS flow, ultimately enabling manufacturers to quickly reach good yield with minimal investment.

Like ReRAM (RRAM), Phase Change Memory (PCM) is a type of non-volatile memory (NVM) that stores bits as resistance, but the two technologies use a different technique to reversibly change the resistance of a material. PCM is based on the ability to change the physical state of a material from an amorphous solid to a crystalline solid and back again by applying heat through an electric current. This type of NVM has unique materials requirements, making it expensive to manufacture. Unlike PCM, Weebit created its ReRAM based on the most common materials and equipment used in fabs today – making it low cost and easy to integrate into any standard CMOS flow, ultimately enabling manufacturers to quickly reach good yield with minimal investment.

eFlash (embedded Flash) is a type of non-volatile memory commonly used in microcontrollers and SoCs for code storage. It’s typically implemented at mature nodes (e.g. 40nm, 65nm) where process constraints still allow it.

eFlash relies on floating gate cells and is a FEOL (front-end-of-line) technology, requiring modifications to the core CMOS process — such as special oxides, high-voltage transistors, and dual-gate stacks. These additions increase mask count, fab complexity, and wafer cost. Moreover, floating gate technology does not scale well below 28nm due to these requirements, making it incompatible with advanced nodes.

In contrast, ReRAM (RRAM) is CMOS-compatible, integrates in the BEOL (back-end-of-line), and scales more easily to advanced geometries. It also supports lower-power, faster operation, and can reduce cost in embedded applications.

ReRAM (RRAM) is non-volatile memory, meaning that it retains data even when power is turned off. DRAM is volatile, meaning that it loses data when the power is off. DRAM is slightly faster in active use, but it constantly refreshes, which burns energy. ReRAM holds data without refreshing, making it more efficient for power-sensitive or persistent applications.

DRAM is usually used in larger systems, such as PCs, GPUs and datacenters, as well as mobile phones, whereas ReRAM broadly targets the entire embedded MCU and SoC market as a replacement for flash, and can be used for secure storage, edge AI and neuromorphic computing.

DRAM has a significantly higher cost per bit than ReRAM, and prices fluctuate with supply-demand cycles.

ReRAM (RRAM) is non-volatile memory, meaning that it retains data even when power is turned off. SRAM is volatile, meaning that it loses data when the power is off.

SRAM is faster than DRAM and NVMs and has low latency, so it is often used in real-time applications as level-1 (L1) tightly coupled cache memory. However, SRAM is expensive, low density and power-hungry. A typical SRAM cell requires six transistors, while a ReRAM cell requires only a single transistor. For that reason, the overall memory density of ReRAM is typically 3-4 times higher than that of SRAM.

While ReRAM is slightly slower, its lower cost, higher density, ability to retain data when powered down and its lower standby energy consumption make ReRAM more suitable for embedded, edge, and low-power use cases.

ReRAM improves on flash across the board.

  • Faster write and erase
  • Lower power consumption
  • Higher endurance
  • Better scaling with process nodes

Flash still dominates for mass storage, but when performance, efficiency, or endurance matter, ReRAM (RRAM) is a better fit and over time is expected to supersede flash.

Because flash has limits.
It’s slower to write and erase, uses more power, and wears out sooner. ReRAM handles more write cycles, operates at lower voltages, and supports faster switching.

ReRAM (RRAM) is also lower cost to manufacture. Weebit ReRAM requires only two additional masks, compared to more than 10 added masks for flash. As a result, Weebit ReRAM only adds ~5% to the wafer cost, whereas flash adds ~20-30%, and even 40% in some process nodes.
As applications push for speed and efficiency, ReRAM is built to keep up.

Weebit ReRAM (RRAM) is extremely low-power, and it has excellent endurance and retention – even at high temperatures and in harsh conditions, and it is scalable to advanced process nodes. It beats other NVMs on key metrics including cost, power consumption, endurance, access time, and more.

Visit The Weebit ReRAM Advantage to learn more.

Applications of ReRAM

ReRAM (RRAM) has inherent physical characteristics that make it ideal for security mechanisms like physical unclonable functions (PUFs), true random number generators (TRNGs) and other technologies integrated during manufacturing. These same traits help keep memory content safe from hacking when ReRAM is embedded as a traditional NVM.

AI inference requires huge amounts of memory, and a significant part of it contains coefficients and other data which is mostly read. It needs fast access to this memory. The huge computation requirements push it to use the smallest geometries where flash is not available. ReRAM (RRAM) is a very good option, as it can be embedded in the AI chip, eliminating large discrete NVM dies, and even replacing some of the SRAM. Since a ReRAM cell operates very similarly to a synapse in your brain, it is also ideal for neuromorphic computing and In Memory Compute (IMC), and many research institutes today use Weebit ReRAM as the basis for their research in this domain.

ReRAM (RRAM) has many advantages for automotive SoCs including high-temp reliability, immunity to EMI, endurance, fast switching speed, longevity, security, and the ability to effectively scale to the most advanced process nodes. Our module in SkyWater S130 has been fully qualified for 10K cycles and 10 years @ 125°C, and we have demonstrated the same module at 100K cycles and for 150°C operation, according to specific mission profiles and based on AEC-Q100.

As a low-cost, high-performance, highly scalable technology, Weebit ReRAM (RRAM) enables key requirements for embedded NVM in analog ICs like PMICs. Importantly, ReRAM is integrated in the Back-end-of-line (BEOL) of the manufacturing process, allowing full optimization of analog components integrated in the Front-end-of-line (FEOL), so there is no impact on design rules. 

Weebit ReRAM (RRAM) consumes 10x less programming power than flash, extending critical battery life for data logging. It is also much lower cost to manufacture than flash and other emerging NVMs, requiring fewer additional masks and using standard tools and materials. ReRAM is also inherently secure, enabling data to be saved securely and making it more difficult to read or change the content. In applications such as industrial IoT, where the last moment of data collection before power loss is most important, a fast embedded NVM like Weebit ReRAM provides a key benefit over standalone flash since there is no need to communicate to an external flash device where critical data could be lost.

Adopting ReRAM/RRAM

Weebit use standard materials and equipment which the fabs are very familiar with, so it is straightforward for fabs to integrate Weebit ReRAM (RRAM) with their existing manufacturing processes. Technologies like MRAM, FRAM, and PCM require fabs to make large capital investments for special equipment and often exotic materials.

Weebit ReRAM (RRAM) IP is highly scalable and customizable by storage capacity, foundry and process node.

No—ReRAM (RRAM) is designed for affordable integration. Like any emerging technology, initial development costs exist, but ReRAM is built to scale cost-effectively. It’s compatible with standard CMOS processes, which makes it easier to manufacture at scale. It only uses two additional masks. Companies like TSMC are already using it, which tells you what you need to know.