Abstract
Today’s and emerging computing tasks are extremely demanding in terms of storage, energy efficiency, and computing efficiency; data-intensive/big-data applications and Internet-of-Things are couple of examples. In addition, today’s computer architectures and device technologies are facing major challenges making them incapable to deliver the required functionalities and features. Computers are facing the three well-known walls: the memory wall, the instruction level parallelism wall, and the power wall. Similarly, nanoscale CMOS technology is facing three walls: the reliability wall, the leakage wall, and the cost wall. In order for computing systems to continue to deliver sustainable benefits for the foreseeable future society, alternative computing architectures have to be explored in the light of emerging new device technologies. Using memristive device technology to enable new computing paradigms such as computation-in-memory architecture is one of the emerging alternatives that could provide a huge potential in terms of energy and computing efficiency.
Introduction: The Crisis of Traditional Computing
Modern computing is currently besieged by two sets of fundamental "walls" that threaten to stall technological progress. On the architectural side, systems are hitting the memory wall (the bottleneck between the processor and RAM), the power wall, and the instruction-level parallelism wall. Simultaneously, the underlying nanoscale CMOS technology is reaching its physical limits, encountering the reliability wall, the leakage wall, and the cost wall.
To meet the massive data demands of the Internet of Things (IoT) and Big Data applications, a paradigm shift is required. We must move away from the traditional "von Neumann" architecture—where data is constantly shuffled between a central processing unit and separate memory—and toward paradigms like Computation-in-Memory (CiM). Memristive devices, essentially resistors with memory, are a leading candidate to bridge this gap by allowing data storage and calculation to occur in the same physical location.
A Roadmap for Memristive Innovation
The December 2018 special issue of IEEE Transactions on Very Large Scale Integration (VLSI) Systems serves as a curated roadmap for this field. From over 100 submissions, 24 high-impact papers were selected (a 24% acceptance rate), covering six key domains:
- Memory Design: Research focuses on improving the reliability and energy efficiency of non-volatile memories such as STT-RAM and RRAM.
- Logic and Circuit Design: Innovations include using memristors to build 3D-integrated circuits and "self-healing" processors capable of resisting manufacturing variations.
- Hardware Security: By exploiting the inherent randomness of memristive switching, researchers are creating True Random Number Generators (TRNGs) and lightweight authentication tools.
- Architectures and Accelerators: This domain involves designing specialized hardware for matrix multiplication, image compression, and pattern matching.
- Neuromorphic and Deep Learning: Brain-inspired hardware, such as hybrid synapses and retinal networks, aims to mimic biological efficiency.
- Design Automation: The development of synthesis algorithms is crucial for helping engineers design complex memristor-based logic circuits.
Key Insights: Breaking the Bottleneck
The most significant advancement presented in this research is the ability to move computation to the data, rather than the data to the computation. This shift drastically reduces the energy overhead associated with data movement.
Furthermore, researchers are exploring logic styles that go beyond binary systems. "Memristor Ratioed Logic" and "Stateful Logic" allow the resistance state of the device itself to represent a logic value, enabling much denser circuitry than traditional transistors. Interestingly, researchers are also "embracing imperfection" in the security domain. The cycle-to-cycle variations in memristors—typically viewed as a reliability flaw—are utilized here to create robust, unhackable hardware security keys.
Results and Performance Gains
The quantitative results highlighted in the special issue are promising. One featured paper describes a memristor-based approximate matrix multiplier that, when used alongside a general-purpose processor, achieves an 86× performance boost for specific tasks. Other research into "Via-Switch FPGAs" and "2M1M Crossbars" points toward highly dense, reconfigurable architectures that occupy significantly less chip area than current CMOS-only designs.
Challenges and Future Outlook
Despite the "huge" potential, several hurdles remain before widespread adoption:
- Variability and Reliability: Manufacturing memristors with perfect consistency remains a challenge, necessitating "variation-tolerant" designs.
- Integration: While memristors can be fabricated on the "Back-End-of-Line" layers of standard CMOS, perfecting this hybrid integration is an ongoing research effort.
- EDA Tooling: There is a critical need for better electronic design automation (EDA) tools to allow circuit designers to work with memristors as seamlessly as they do with standard transistors.
Conclusion
Memristive technology has transitioned from a theoretical curiosity to a burgeoning field providing practical solutions for next-generation computing. By addressing the physical and architectural limitations of current systems, these devices pave the way for a more energy-efficient and powerful digital future.
Summary updated