Acronym · Memory & Storage
ReRAM (Resistive Random Access Memory)
Non-volatile memory that switches a metal oxide between high and low resistance by forming and rupturing a conductive filament. It is simple and crossbar friendly, but cell-to-cell variability has limited adoption.
- Also known as
- RRAM
- Type
- Acronym
- Domain
- Memory & Storage, Emerging memory
Notes
A favourite substrate for analogue in-memory compute because its resistance can be tuned to many intermediate levels.
Overview
ReRAM (resistive random access memory, also written RRAM) is a non-volatile memory that stores each bit as the resistance of a thin insulating layer between two electrodes, switched between a high and a low state by applying voltage. Its simple cell sits in the metal layers above the transistors, which makes it a candidate to replace embedded flash in microcontrollers and an attractive device for research on computing inside memory arrays. It is not one technology but a group, with oxide-based and metal-bridge types the best known.
How it works
In the oxide type, often called OxRAM, a metal oxide is sandwiched between a top and a bottom electrode. A voltage of one polarity forms a conductive filament of oxygen vacancies through the oxide, giving a low resistance; a voltage of the opposite polarity breaks it again, restoring a high resistance. In the conductive-bridge type (CBRAM), the filament is made of copper or silver atoms driven into the switching layer. Both are read by sensing the resistance with a small current.
Because the cell is built between two metal layers in the back end of line, adding it to a logic process needs few extra masks. Weebit Nano, which licenses ReRAM, says its version needs only two, and describes endurance of 100,000 to a million write cycles. ReRAM does not store charge, so it is not disturbed by magnetic fields and resists some kinds of physical attack, which suits smart cards and security uses.
The main difficulty is variability. A filament is controlled by a small number of vacancies or atoms moving at random, so the resistance can differ from one write to the next almost as much as it differs between cells. Higher switching currents reduce the spread but cost power, and self-heating at small sizes narrows the gap between the two states. Write times are also much slower than SRAM.
Examples
- TSMC and Infineon's jointly developed RRAM, which Infineon has used in chip cards and controllers and is bringing to automotive microcontrollers.
- Weebit Nano ReRAM qualified in SkyWater Technology's 130 nm CMOS process and offered as embedded memory IP.
- IBM research using hafnium oxide ReRAM for analogue deep learning hardware.
- The NeuRRAM research chip from UC San Diego and Stanford, which used ReRAM for neuromorphic computing.
Why it matters
ReRAM concerns embedded memory and device engineers at foundries, microcontroller and secure chip designers looking for on-chip non-volatile memory beyond the limits of embedded flash, and researchers building analogue in-memory computing for AI, where the resistance of each cell can stand for a stored weight.
Related terms
Part of the Memory & Storage glossary, with every term in the topic and the lifecycles, tools and standards that use them.
Where it appears on the Hub
Used on 4 pages of the Hub. In each group, the pages that mention it most come first.
Suggest an improvement
Suggest an improvement to this entry
A clearer definition, an example, a correction or a source an editor can check. Editors review every suggestion before the entry changes.
Last reviewed . Written by The Semiconductor Hub editorial team. The full glossary