Emerging memory technologies
Emerging memories refers to new and innovative types of computer memory technologies that are in the early stages of development or deployment, aiming to overcome the limitations of existing memory technologies.
These emerging memory technologies often offer improvements in terms of speed, capacity, energy efficiency, and durability compared to traditional memory technologies like DRAM (Dynamic Random-Access Memory) or NAND flash.
Follow us on Linkedin for everything around Semiconductors & AI
What is computer memory?
Computer memory is just like the human brain. It is used to store data/information and instructions. It is a data storage unit or a data storage device where data is to be processed and instructions required for processing are stored.
When you open a program, it is loaded from secondary memory (SSD/HDD) into primary memory (RAM). The primary memory is quickly accessible and the CPU accesses it, does computation on the retrieved data and stores the results back in primary memory. The data is then updated into the secondary memory for permanent storage.
Read More: Explained: What the hell is memory?
Conventional Memory: DRAM and Flash
Dynamic Random Access Memory (DRAM):
DRAM, the swift and volatile workhorse, provides temporary storage for active applications, ensuring rapid data access for the CPU. Its successor, Double Data Rate (DDR) RAM, has been the stalwart in facilitating seamless multitasking and efficient data processing.
Flash Memory:
In the realm of non-volatile memory, Flash has become ubiquitous, powering everything from BIOS chips to storage solutions. Its ability to retain data even when the power is turned off makes it a crucial component in modern computing. This offers a bridge between volatile and non-volatile memory.
Read More: Why can’t we Scale memory chips?
Why do we need Emerging memory ?
Emerging memories are being developed to address various challenges associated with existing or traditional memory technologies. Some of the key challenges of existing memories that emerging memories aim to solve include:
Volatility:
- Challenge: Many existing memories, such as Dynamic Random Access Memory (DRAM), are volatile, meaning they lose data when power is turned off. This necessitates frequent data refresh cycles.
- Emerging Memory Solution: ReRAM provide persistent storage capabilities, eliminating the need for constant power supply to retain data.
Endurance and Write Cycling:
- Challenge: Existing memories, particularly NAND Flash, have limited endurance, meaning they can withstand only a certain number of program/erase cycles before degradation.
- Some emerging memories, including certain types of ReRAM and Phase Change Memory (PCM), offer higher endurance and reliability. This makes them suitable for applications with frequent write cycles.
Speed and Latency:
- Challenge: While traditional memories like DRAM offer fast read and write speeds, latency can still be a limiting factor, especially in applications requiring quick data access.
- Emerging Memory Solution: Various emerging memory technologies, such as Memristors and STT-RAM, aim to provide comparable or improved speed and lower latency, enhancing overall system performance.
Energy Efficiency:
- Challenge: Traditional memory technologies, especially when dealing with large-scale data centers and mobile devices, can consume significant amounts of energy.
- Emerging Memory Solution: Emerging memories often target lower power consumption, making them more energy-efficient. This is particularly beneficial for battery-powered devices and environments where power efficiency is crucial.
Scalability and Density:
- Challenge: As data storage requirements continue to grow, existing memory technologies may face challenges in terms of scalability and achieving higher storage density.
- Emerging Memory Solution: Emerging memories, including technologies like 3D XPoint, aim to provide higher storage density and improved scalability, addressing the demands of modern data storage applications.
New Computing Paradigms:
- Challenge: Traditional memory architectures may not be well-suited for emerging computing paradigms, such as neuromorphic computing or in-memory processing.
- Emerging Memory Solution: Certain emerging memory technologies, like Memristors, are designed to support new computing architectures, enabling advancements in areas like artificial intelligence and brain-inspired computing.
Technological Limits of Scaling:
- Challenge: Traditional memory technologies face challenges associated with physical scaling limits, which can impact further miniaturization and integration.
- Emerging Memory Solution: Some emerging memory technologies are exploring alternative materials and structures that allow for continued miniaturization and improved performance beyond the limits of traditional scaling.
Read More: China Makes World’s Most Advanced 3D NAND memory chip Despite US Sanctions
Types of Emerging Memory Technologies:
1. 3D Integrated Memory:
Three-dimensional integrated memory (3D-IC memory) is a revolutionary memory technology that stacks multiple layers of memory cells vertically, interconnected through tiny vertical vias called through-silicon vias (TSVs). This vertical stacking approach offers several advantages:
- Higher Capacity: Achieves significantly higher memory capacity within the same physical footprint.
- Reduced Interconnect Lengths: The vertical stacking of memory cells in 3D-IC memory significantly shortens the interconnects
- Wider Bandwidth: 3D-IC memory allows for the creation of wider data buses, enabling faster data transfer between the memory and other components on the chip.
- Lower Power Consumption: The reduced interconnect lengths and lower signal resistance reduce the power consumption
3D-IC memory is still in its early stages of development, but it has the potential to revolutionize the semiconductor industry and enable a new generation of high-performance computing.
2. Phase-Change RAM (PRAM):
PRAM is a non-volatile ram which harnesses the unique properties of materials that can switch between amorphous and crystalline states.
The crystalline state has lower electrical resistance, while the amorphous state has higher electrical resistance. The state of the alloy represents the stored data.
PRAM is considered a storage-class memory that fits between DRAM and NAND in the memory hierarchy. The most prominent memory in this class is 3D XPoint, developed by Intel and Micron, which is expected to be used in DIMMs (Dual In-line Memory Module) for servers
3. Magneto resistive RAM (MRAM):
Unlike conventional memories which store information in the form of electric charges, MRAM leverages the intrinsic spin of electrons to store and retrieve data.
Offering a compelling blend of speed, non-volatility, this technology is making waves in various applications. Its inherent advantages include faster write speeds and better scalability compared to traditional RAM.
MRAM is used in IoT devices for edge computing. As it is non-volatile it acts as a storage for real time data generated by sensors. It also has wide applications in Artificial neural networks for storing weights.
4. Resistive RAM (RRAM):
ReRAM is a non-volatile memory that operates by changing the resistance across a dielectric solid-state material, often referred to as a memristor. This change in resistance is used to represent binary data.
One major advantage of RRAM over other non-volatile memory technologies is its ability to scale below 10nm. It finds it applications in PMICs, edge AI and In-memory computing architectures.
| Memory Type | Common Applications | Example Emerging Memory Technology |
|---|---|---|
| Non-volatile | Data storage, SSDs | 3D XPoint, ReRAM, Phase Change Memory (PCM) |
| Low Power | Mobile devices, IoT | Spin-Transfer Torque RAM (STT-RAM), FeRAM |
| High Density/Scalability | Cloud computing, Big Data | 3D XPoint, ReRAM |
| Endurance/Reliability | Industrial applications, Data logging | ReRAM, MRAM (Magnetoresistive RAM) |
| New Computing Paradigms | Neuromorphic computing | Memristors, PCM |
Read More: Who invented memory..Intel or Toshiba?
Advantages of Emerging Memory technologies
We need emerging memory technologies for several reasons, as they offer improvements and advancements over traditional memory technologies like DRAM and NAND Flash. Here are some key reasons, illustrated with an example:
| Memory Type | Existing Memory (Representative: DRAM) | Common Applications | Emerging Memory Technologies |
|---|---|---|---|
| Volatile/Non-volatile | Volatile (loses data when power is off) | Databases, System Memory | Emerging memories often non-volatile (retain data even without power) |
| Speed | Fast read and write speeds | Main system memory, cache | Varies; some emerging memories aim for comparable or improved speeds |
| Energy Efficiency | Relatively high power consumption | Laptops, desktops, servers | Many emerging memories aim for lower power consumption |
| Density/Scalability | High density, scalable to an extent | Data storage, cloud computing | Emerging memories often target higher density and improved scalability |
| Endurance/Reliability | Limited endurance, frequent refresh | System memory, temporary storage | Varies; some emerging memories offer higher endurance and reliability |
| Applications | Real-time data processing, caching | Gaming, graphics processing | Diverse, including storage, in-memory computing, specialized tasks |
| New Computing Paradigms | Not designed for neuromorphic computing | General-purpose computing | Some emerging memories enable neuromorphic computing and new architectures |
| Example Technology | DRAM | In-memory databases, real-time systems | Memristors (for neuromorphic computing), 3D XPoint (Optane) for high-speed storage, ReRAM for low-power storage, STT-RAM for fast, non-volatile memory, etc. |
Conclusion: Embracing the Future
The semiconductor industry is undergoing a transformative era. We stand at the crossroads of conventional and emerging memory technologies. The shift from traditional memories to emerging technologies is underway. It represents a quest for better performance in computing. More than just performance, it signifies a profound reimagining of computing foundations.
