By
04.23.2023
0
//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>
While the world is opening up again as COVID restrictions are eased, some of the changes brought about by the pandemic remain. We have become increasingly reliant on various technologies which keep us connected to the world wherever we are. This shift is driving the generation of an unprecedented amount of data. Today, data-heavy technologies such as ChatGPT and the metaverse have become part of our daily lives. The backbone of these impactful technologies is the semiconductor, posing a challenge to semiconductor companies to advance their products to meet increasing technological demands. However, it also creates opportunities for semiconductor technologies to become faster and more efficient in terms of power, performance, area, and cost (PPAC).
Memory innovation is recognized as one of the key solutions to address the challenges of the data explosion era. Not only is it important that memory technologies offer standard specifications such as high performance, lower power consumption and cost, and higher capacity, but they must also offer smarter solutions to effectively eliminate issues inherent to the memory wall1. In addition, the explosion of data and technology scaling challenges open up opportunities toward more memory-centric computing and distributed system architectures. This article will look at how semiconductor companies such as SK hynix are developing emerging memory solutions for the advanced technologies of today.
1Memory wall: As processor speed improves at a faster rate than memory speed, the difference in performance between the two components is a key cause of system bottlenecks.
Evolution of Emerging Memory from PCM to SOM
Memory innovation for next-generation computing is a journey which takes several steps. It begins with developing newly emerging memory technologies to support new applications, and continues on the path to ultimately break down the memory and computation boundaries. The introduction of new interfaces such as Compute Express Link (CXL)2 can offer many opportunities for emerging storage memories. The starting point consists of several research options including chalcogenide-based memories which offer better performance and process simplification, going beyond previous industry solutions such as 3D XPoint phase-change memory (PCM)3.
2 Compute Express Link (CXL): A PCIe-based next-generation interconnect protocol on which high-performance computing systems are based.
3 Phase-change memory (PCM): A technology which enables nonvolatile electrical data storage at the nanometer scale. A PCM device consists of a small active volume of phase-change material placed between two electrodes.
Industry 3DXP products were implemented in several system solutions. However, PCM suffers from slow write speed and endurance as a result of the fundamental device characteristics. These limitations pose several challenges in system applications, although significant progress has been made to resolve these issues in recent years. Furthermore, the high aspect ratio of cross-point phase-change RAM cells has been one of the key challenges in integration and technology scalability.
SK hynix’s Revolutionary Technology Center (RTC) has explored chalcogenide-based memory solutions such as selector-only-memory (SOM) to improve performance and simplify processes. Unlike PCM, this new SOM has a dual function material which acts as both memory and selector in bi-directional operations. However, SOM does not suffer from the issues which have prevented PCM media from being widely adopted. Moreover, SOM takes advantage of an already existing chalcogenide manufacturing ecosystem, therefore lifting significant roadblocks for new materials. As shown in Figure 3 below, SOM has demonstrated a write speed as low as 20 nanoseconds (ns) and up to 10 million write cycles at statistically meaningful distributions. Although the potential of SOM is promising, there are some technical challenges to address related to CXL memory solutions. These are related to bi-directional operations and further improvement of endurance.
VSOM: The Next Stage in SOM’s Development
SK hynix aims to continue developing chalcogenide-based memory scalability options beyond current SOM architectures. The company believes that chalcogenide-based CXL memory architectures can be extended further with vertical SOM (VSOM). As shown in Figure 4, VSOM essentially takes advantage of 3D NAND-like structures with SOM materials to develop ultra-high density memory solutions. At IEEE International Memory Workshop (IMW) 2022, SK hynix presented an early feasibility study of VSOM including reasonable memory windows. However, VSOM is still at a very early research stage since it requires significant material innovations such as a robust chalcogenide atomic layer deposition (ALD) process. In order to make significant progress in this area, SK hynix is targeting collaborations with material solution partners in the coming years.
Although emerging memories offer numerous opportunities and advantages, they are not ideal for all applications due to the differing physical characteristics of their respective materials. To illustrate this point, the table in Figure 5 shows the comparison between various emerging memories. Moreover, cost, endurance and latency need to be reviewed carefully for target applications in a similar way to the trade-off of PPAC in logic technology.
How ACIM Can Realize the “Beyond Memory” Era
Finally, emerging memory solutions are set to be critical to realize the “Beyond Memory” era by breaking the boundaries between computation and memory. In light of this, analog-compute in memory (ACIM) has been of great interest both in academia and industry as a path to energy-efficient AI accelerators for next-generation computing. ACIM has the potential for simultaneous computation and storage due to its non-volatile memory characteristics, which has brought it under the spotlight in recent years.
SK hynix’s RTC is evaluating the potential for ACIM as its cells have many commonalities with known memory cells while it also offers unique linear optimization. RTC has successfully demonstrated 16 levels of a resistive RAM-based synapse cells platform with good set/reset characteristics which are embedded in a CMOS technology. The results of research into these cells is expected to be published in the future.
New Memory R&D Ecosystem Vital for Future of Emerging Memory
Although many opportunities exist in emerging memory technology, it should be stressed that the introduction of new memory solutions requires a whole new memory ecosystem. The introduction of emerging memory is a testcase of system technology co-optimization (STCO)4 whether exploring CXL memory or “Beyond Memory” solutions. To realize emerging memory, building a new memory R&D ecosystem and working together across the ecosystem will play a critical role to move beyond memory-wall issues in current Von-Neumann computing architectures as we move toward next-generation computing.
4 System technology co-optimization (STCO): Process of combining memory, processors, mixed-signal IP and sensors into single packages.
Therefore, the journey of memory innovation is only possible if all companies and academic institutions which are part of the memory ecosystem collaborate to address the various issues in computing. As the old proverb on the importance for teamwork says: “It takes a village to raise a child.” Now the time has come for those in the memory sector to come together to realize the future of emerging memory.
