• Industry’s first 321-high NAND of 1Tb developed for supply in 1H25
• “Three Plugs” process technology leads to technological breakthrough for stacking, improves speed, power efficiency
• Company on track to becoming “Full Stack AI Memory Provider” with enhanced competitiveness in AI storage

Seoul, November 21, 2024

SK hynix Inc. (or “the company”, www.skhynix.com) announced today that it has started mass production of the world’s first triple level cell¹-based 321-high 4D NAND Flash with 1Tb capacity.

¹NAND Flash products are categorized into single-, multi-, triple-, quadruple-, and penta-level cells, depending on the number of the information in the format of bit unit is stored in a cell. A bigger number of information stored means more data can be stored in the same space.

Following its previous record as the industry’s first provider of the world’s highest 238-layer NAND since June last year, SK hynix has become the world’s first supplier of the NAND with more than 300 layers by finding a technological breakthrough for stacking. The company plans to provide the 321-high products to customers from the first half of next year.

Stacking more than 300 layers came into reality as the company successfully adopted the “3 plugs²” process technology. Known for an excellent production efficiency, the process electrically connects three plugs through an optimized follow-up process after three times of plug processes are finished. For the process, SK hynix developed a low-stress³ material, while introducing the technology that automatically corrects alignments among the plugs.

²Plug: a vertical hole through layers of substrates aimed at creating cells at once
³Low Stress: Preventing wafer warpage by changing the material into the plugs

With the adoption of the same development platform from the 238-high NAND on the 321-high product, the company could also improve the productivity by 59%, compared with the previous generation, by minimizing any impacts from a process switch.

The latest product comes with an improvement of 12% in data transfer speed and 13% in reading performance, compared with the previous generation. It also enhances data reading power efficiency by more than 10%.

SK hynix plans to steadily expand the use of the 321-high products by providing them to the nascent AI applications, which require low power and high performance.

Jungdal Choi, Head of NAND Development at SK hynix, said that the latest development brings the company a step closer to the leadership of the AI storage market represented by SSD for AI data centers and on-device AI. “SK hynix is on track to advancing to the Full Stack AI Memory Provider by adding a perfect portfolio in the ultra-high performance NAND space on top of the DRAM business led by HBM.”

About SK hynix Inc.

SK hynix Inc., headquartered in Korea, is the world’s top tier semiconductor supplier offering Dynamic Random Access Memory chips (“DRAM”), flash memory chips (“NAND flash”) and CMOS Image Sensors (“CIS”) for a wide range of distinguished customers globally. The Company’s shares are traded on the Korea Exchange, and the Global Depository shares are listed on the Luxembourg Stock Exchange. Further information about SK hynix is available at www.skhynix.com, news.skhynix.com.

Media Contact

SK hynix Inc.
Global Public Relations

Technical Leader
Kanga Kong
E-Mail: global_newsroom@skhynix.com

Technical Leader
Sooyeon Lee
E-Mail: global_newsroom@skhynix.com

The post SK hynix Starts Mass Production of World’s First 321-High NAND first appeared on SK hynix Newsroom.

Driving key technologies, intertwined with industries across the world, and an essential part of the AI era—semiconductors are integral to daily life. This fifth episode in the Semiconductor 101 series explores in detail why semiconductors are crucial for shaping modern society. The article also delves into why SK hynix is adapting its approach in the AI era and examines the sweeping influence of the semiconductor industry on the global economy.

Why does the world need semiconductor memory?

Semiconductor memory is indispensable for modern technology, enabling efficient data storage and processing across a variety of devices. To gain a better understanding of semiconductor memory’s importance, let’s picture how its absence would impact some areas of life.

A world without semiconductor memory would be like stepping back in time

Work

Without semiconductor memory, workers would find tasks that once could be finished in a day would take weeks or even months to complete. Computers would be painfully slow and inefficient —making opening and saving files a chore—and offer significantly reduced storage capacity. Multitasking would also be inefficient, with noticeable delays when switching between applications like Word documents and web browsers.

Other essential business applications such as video conferencing, cloud-based software, and messaging tools would be unreliable and costly, hindering collaboration and presenting huge obstacles to remote workers.

Digital Entertainment

People may struggle to fill their evenings and weekends as the entertainment on offer becomes severely limited. Online activities would be time-consuming or almost impossible without semiconductor memory, which enables networking equipment to manage and route data. Social media would be unrecognizable, as users would be unable to instantly share photos and videos. Streaming, gaming, and cloud services would also be unusable without fast and scalable storage solutions. Smartphones would not exist in their current form either, as high-performance and efficient memory is key to the functionality, compact size, and battery life of modern phones.

Banking & Finance

Financial transactions would look very different in a world without semiconductor memory. It would be necessary to make frequent trips to the bank, as app-based transactions would be impossible. This would affect personal finances and the broader economy.

Why is semiconductor memory crucial for the AI era?

In the AI era, semiconductor memory plays a key role in ensuring the smooth operation of AI applications. Today’s AI systems are performing advanced tasks such as image generation and problem solving which require significant data processing capabilities. Moreover, as these AI tools require ever-growing amounts of data to continue their evolution, there is a pressing need to manage this explosion of data. That’s where semiconductor memory comes in.

Semiconductor memory offers several key features which support the smooth operation of AI

High-performance memory is essential for data storage, retrieval, and processing, enabling technologies to keep pace with the rapid rate of data generation. Let’s break down some of the reasons why semiconductor memory is essential in the age of AI.

• Rapid Data Processing: AI systems must rapidly process vast amounts of data to make real-time decisions. Semiconductor memory supports this process by enabling fast data access and storage.
• AI Training Support: AI applications are trained on large datasets to improve their performance. Memory products with high capacity and bandwidth can store and manage these datasets. Moreover, the rapid read/write capabilities of memory solutions can accelerate AI training, making AI systems more efficient.
• Energy Efficiency: AI applications are resource intensive. Semiconductor memory helps reduce power consumption, making them more sustainable. Edge devices¹, such as smart sensors and mobile AI processors, depend on energy-efficient semiconductor memory to perform tasks like real-time image recognition while conserving battery life.

As AI continues to evolve and integrate into various technologies, semiconductor memory remains the key to unlocking its full potential.

Why is HBM an essential product for AI?

Featuring stacked DRAM² chips, High Bandwidth Memory (HBM) is an ultra-fast memory technology which has become crucial in the age of AI. Combining high performance with low-power consumption, HBM can efficiently handle the growing demands of AI systems.

Combining high performance with low-power consumption, HBM is an essential AI memory product

To illustrate some of the benefits of HBM for AI compared to other DRAM products, think of it as an upgraded metro system that reduces passenger wait times at stops. Traditional DRAM chips have a limited number of data paths, akin to station entrances and exits, which are prone to bottlenecks in the data-centric AI era.

Meanwhile, HBM features significantly more data paths which run vertically to support the smooth and rapid passage of data, just like passengers freely passing through a station with numerous entrances and exits. HBM’s ability to minimize delays, or latency, is key in AI systems which need real-time data access to perform tasks such as deep learning.

As touched upon in the previous question, AI systems handle massive datasets and complex computations. HBM supports these tasks by enabling swift and efficient data processing, making it ideal for AI training. For example, SK hynix’s world’s best-performing 12-layer HBM3E offers industry-leading data processing speeds of 1.18 terabytes (TB) per second.

In addition, HBM is optimized for supporting parallel processing in AI systems due to its high bandwidth. As it offers high performance with low-power consumption, HBM also enables AI systems to operate efficiently, particularly in data centers and edge computing devices.

Why does SK hynix need to expand its semiconductor production capacity?

Today, phones instantly recognize faces, autonomous cars cruise down highways, and smart fridges can place orders for food. These developments are largely thanks to technologies such as AI, 5G, and Internet of Things (IoT) devices, which all require high-performance memory products. Moreover, the rapid growth of these and other advanced technologies is further heightening memory demand.

The number of AI users is forecast to grow rapidly, heightening demand for semiconductor memory

Looking at the AI market alone, the number of AI users is projected to grow from almost 255 million in 2023 to around 730 million by 2030³. In response to this changing technological landscape and resultant rising demand for memory, SK hynix is ramping up its production capacity. The company currently operates four production facilities in its home country of South Korea and plans to build two additional domestic facilities in the future, namely:

• The Yongin semiconductor cluster which will boost the company’s production of AI memory
• A fabrication facility (fab) in Cheongju which will produce next-generation DRAM products, including HBM

While it may seem straightforward to simply build new manufacturing facilities to increase production, there are other aspects to consider. For example, there are often challenges associated with commercializing innovations through mass production.

SK hynix is adept in this area, as the company has a long history of transitioning innovations from the R&D stage through to high-volume manufacturing, particularly with DRAM and NAND flash⁴. A recent example of this occurred in March 2024, when the company successfully mass-produced its 8-layer HBM3E through close client collaboration and enhancements in product design.

Why does the semiconductor industry have so much impact on the global economy?

The value of the semiconductor market is greater than the GDP of numerous countries

Semiconductors are essential components in a multitude of electronic devices and are driving technological advancements, making the industry a cornerstone of the global economy. In 2023, the semiconductor market was valued at around 611 billion USD, greater than the GDP of Singapore. However, the full extent of the semiconductor sector’s impact on the global economy can only be understood when looking at some of the industries fueled by these chips.

The consumer electronics market, which is valued at around 950 billion USD, relies heavily on semiconductors. The market’s size is emphasized by the fact that global smartphone subscriptions have surpassed 6.9 billion.

The telecommunications sector, including the expected trillion-dollar 5G market, also depends on semiconductors along with the automotive and healthcare industries. As previously mentioned, semiconductors are also key to advanced technologies such as AI, 5G, and IoT, highlighting their ubiquity in a wide range of high-value technology markets.

While it has been established that the semiconductor sector is a multi-billion-dollar industry which creates immense direct revenue, it also contributes to the economy in other forms. In particular, the sector employs people across the world in roles such as R&D, supply chain management, and manufacturing. By 2030, an estimated 3 million employees will be needed in the sector, highlighting its impact on jobs markets worldwide.

Exploring semiconductor manufacturing in more detail, production is concentrated in East Asian countries such as Taiwan and South Korea. Slowdowns in production which contributed to global chip shortages throughout the years have further highlighted the sector’s strategic importance to numerous industries across the world.

Looking ahead, the semiconductor industry’s impact on the global economy will only continue to grow as the sector itself, and others which rely on chips, expand in the future.

¹Edge: The boundary of a network where data is generated and processed locally on devices or sensors, rather than being transmitted to a centralized data center.
²Dynamic Random-Access Memory (DRAM): A type of short-term computer memory that stores data temporarily while a device is on, losing information when the power is turned off.
³Statista, “Generative artificial intelligence (AI)” report (2024)
⁴NAND Flash: A long-term storage technology that retains data even when the power is off, commonly used in USB drives and memory cards.

The final episode of the Semiconductor 101 series will explore “how” semiconductors are made and provide insights on launching and advancing a career in this dynamic industry.

<Other articles from this series>

[Semiconductor 101] SK hynix’s Guide to Who’s Who in the Semiconductor Industry

[Semiconductor 101] SK hynix Explains “What’s What” in the Semiconductor World

[Semiconductor 101] When Semiconductors & SK hynix Made Their Mark on the World

[Semiconductor 101] “Where” in the World Are Semiconductors Made and Applied? SK hynix Reveals All

The post [Semiconductor 101] “Why” Modern Tech Needs Semiconductors & SK hynix’s Key Contributions first appeared on SK hynix Newsroom.

When the earliest semiconductor devices emerged in the 19th century, a revolution was set in motion that would reshape the technological landscape. One scientific breakthrough after another, these milestones mark the dawn of an era in which innovation and progress became the new normal. Ultimately, these discoveries changed the way people live, work, and connect. In this Semiconductor 101 episode, the milestones and key turning points of the semiconductor industry, and SK hynix, will be examined to better understand how semiconductor technology reached today’s unprecedented levels.

When was the first semiconductor made?

As shown in the first episode, the development of semiconductor technology throughout history cannot be attributed to one person or a single moment. Instead, a long and intertwined web of pioneers and innovators contributed to the field’s evolution.

Starting with the first documented observation of a semiconductor effect by English scientist Michael Faraday in 1833, progress in the semiconductor world began to gather pace with several key discoveries. In 1874, German physicist Karl Ferdinand Braun invented what is widely considered to be the first-ever semiconductor diode¹. The mid-20th century then witnessed a number of key inventions including the transistor² in 1947, the integrated circuit³ in 1958, and the MOSFET⁴, in 1960.

The origins of semiconductor memory can be traced back to the 1960s

A number of these innovative technologies would go on to be applied in the memory field, leading to the first semiconductor memory products. Developed by engineers at major semiconductor companies, some of these early semiconductor memory solutions include: DRAM⁵, SRAM⁶, and NAND flash.

DRAM
In 1966, IBM’s Robert Heath Dennard invented DRAM, a system that would hold one bit in a single transistor. The technology was put into popular use in 1970 when Intel developed a 1-kilobit DRAM chip using a three-transistor cell design. Called the 1103, the first-ever commercial DRAM replaced magnetic core memories as the new standard technology for computer memory.

SRAM
Robert Norman patented a semiconductor SRAM design at Fairchild Semiconductor in 1963 to create a faster and more reliable form of memory capable of enhancing computing performance. Two years later, IBM commercialized the first SRAM chip for use in its computer. Unlike DRAM, SRAM can retain data without the need for constant refreshing. This capability, coupled with its high speed and reliability, ensured that SRAM became an essential component in modern computing.

NAND Flash
As the most widely-used flash memory today, NAND flash has numerous applications including smartphones, solid-state drives (SSDs), and data centers. This electronic, non-volatile computer memory storage that can be erased and reprogrammed was first unveiled by Fujio Masuoka of Toshiba Memory Corporation (now known as KIOXIA) in 1987. The company would go on to commercialize the world’s first NAND flash memory product in 1991.

When did semiconductor memory become important for our daily lives?

Semiconductor memory has progressed and grown in importance in line with the rapid explosion of data and technological advancements. By tracing the history of technological progress over the last century, it is clear to see the key role that semiconductor memory plays in these technologies and daily life.

Demand for semiconductor memory has grown rapidly as data generation has increased over the years

Age of Giant Computers: Before 1980s
Although unimaginable today, the first line of computers such as the ENIAC⁷ was so large that they would take up a whole room. These massive devices were almost exclusively used by businesses and governments for bulk data processing and complex calculations. In terms of data storage, there was a landmark change in the 1970s when semiconductor memory replaced magnetic-based memory to become the dominant storage technology.

Emergence of PCs: 1980s–2000s
Many people’s memories of their first family computer will be of a bulky, gray PC sitting atop an office desk. While they are primitive by modern standards, these early popularized PCs of the 1980s brought computing—and data—into homes across the world for the first time. One of the most prevalent PCs of this era was Commodore 64 (C64), which came with 64 kilobytes (KB) of RAM⁸ when it was released in 1982.

Smartphones & Cloud Computing: 2000s to Present
Although it was not the first smartphone, the original iPhone kicked off the smartphone boom upon its release in 2007. Able to take photos, send emails, access the internet, and more, smartphones generate huge amounts of data which is stored in semiconductor memory. As the pace of data generation continues to quicken, demand is also growing for cloud storage which relies on server DRAM, NAND flash, and SSDs.

The AI Era & Beyond
As AI continues to permeate more aspects of daily life, semiconductor memory is driving the technology’s development. High-performance memory such as SK hynix’s HBM⁹ is optimized for AI applications as it can handle and access vast amounts of data. As the pace of data usage continues to accelerate over the course of the AI era and beyond, semiconductor memory will only play a bigger role in AI and other advanced technologies.

When did SK hynix start manufacturing semiconductors?

Founded as Hyundai Electronics in 1983, SK hynix has been a hub of innovation in the semiconductor industry over the course of its 40-plus-year history. In 1984, SK hynix’s pilot production of South Korea’s first-ever 16 kilobit (Kb) SRAM kickstarted the company’s journey as a semiconductor manufacturer. Just a year later, the company mass-produced its first DRAM product—a 64 Kb DRAM. SK hynix then completed construction of a semiconductor assembly plant two months later, enabling it to ramp up production.

As the years ticked by, the company enhanced its product lineup and fully established itself as a leader in the semiconductor sector. Some of its key milestones included mass-producing the 256 Mb SDRAM for mobile phones in 2003, developing the 512 Mb NAND flash in 2004, and the world’s first TSV¹⁰-based HBM in 2013.

Fast-forward to today, SK hynix has a broad portfolio of ultra-high-performance solutions including DRAM, NAND flash, SSDs, and CMOS image sensors (CIS)¹¹. The company is continually pushing technological limits to develop groundbreaking products such as the world’s first 321-layer NAND flash and industry-leading HBM3E. Taking the latest 12-layer, 36 GB HBM3E as an example, this groundbreaking DRAM for AI systems boasts around 4.5 million times¹² more data storage than the 64 Kb DRAM of 1985. Having come so far over the years, the company is set to continue to set new standards in the semiconductor sector with its next-generation products.

SK hynix’s semiconductor memory products have evolved immensely since the company’s early days

When did SK hynix expand into the global market?

SK hynix has constantly advanced its product lineup, formed strategic global partnerships, and innovated across its operations to become a leading light in the semiconductor memory industry. Looking at the company’s flagship products, SK hynix held a 31.8% share of the DRAM market and a 21.6% share of the NAND flash market in the fourth quarter of 2023. The company is also the HBM market leader, dominating the sector with its industry-leading products.

SK hynix commands significant shares of the global DRAM and NAND flash markets

While there wasn’t a single moment when SK hynix expanded in the global market, there are several instances over the years where the company has made its mark worldwide:

1983 – Began overseas expansion with establishment of Hyundai Electronics America—later becoming SK hynix America

1989 – Ranked 20th in the global semiconductor industry

1995 – Established a non-memory U.S. corporation called SYMBIOS and constructed a semiconductor plant in the U.S. state of Oregon

1996 – Began its initial public offering (IPO) process

2004 – Sold non-memory business to shift focus to semiconductor memory

2006 – Completed construction of DRAM chip plant in Wuxi, China

2010 – Named to Dow Jones’ Sustainability World Index, started joint development with Hewlett-Packard for next-generation memory product, and completed second memory chip plant in China capable of 100 million units of 1 GB DRAM chips per month

2012 – Joined SK Group and officially rebranded as SK hynix

2018 – Posted record operating profits

2020 – Invested in AI company Gauss Labs located in Silicon Valley

2021 – Completed first phase of acquiring Intel’s NAND business to enhance AI memory capabilities

2023 – Developed and mass-produced HBM3E with the world’s best specifications for AI systems and applications

2024 – Announced plan to build advanced chip packaging plant and R&D facility in Indiana, U.S.

¹Semiconductor diode: A simple electronic component that conducts electricity in one direction.
²Transistor: A semiconductor device that regulates current or voltage flow and acts as a switch or gate for electronic signals.
³Integrated circuit (IC): A small electronic device comprising of components such as transistors, resistors, and capacitors on semiconductor material.
⁴Metal-oxide-semiconductor field-effect transistor (MOSFET): An active semiconductor device in which a conducting channel is induced in the region between two electrodes by a voltage applied to an insulated electrode on the surface of the region.
⁵Dynamic random-access memory (DRAM): Volatile memory that needs to be refreshed periodically to maintain stored data. Like SRAM, it loses stored data when the power supply is removed.
⁶Static random-access memory (SRAM): Volatile memory that holds memory permanently as long as power is supplied. Unlike DRAM, it does not have to be refreshed periodically to keep storing data.
⁷Electronic Numerical Integrator and Computer (ENIAC): Built in 1946, the ENIAC was the world’s first general purpose electronic computer.
⁸Random access memory (RAM): A computer’s main memory in which data can be rapidly accessed directly by the central processing unit regardless of the sequence it was recorded.
⁹High Bandwidth Memory (HBM): A high-value, high-performance product that possesses much higher data processing speeds compared to existing DRAMs by vertically connecting multiple DRAMs with through-silicon via (TSV).
¹⁰Through-silicon via (TSV): A type of vertical interconnect access (via) that completely passes through a silicon die or wafer to enable the stacking of silicon dice.
¹¹CMOS image sensor (CIS): Acting as the “eyes” of electronic devices, a CIS converts light into electrical energy to create images.
¹²Calculation based on the decimal standard.

Having looked at when semiconductor technology and SK hynix began to make their mark on the world, the next episode will cover where semiconductors are developed and more.

<Other articles from this series>

[Semiconductor 101] SK hynix’s Guide to Who’s Who in the Semiconductor Industry

[Semiconductor 101] SK hynix Explains “What’s What” in the Semiconductor World

The post [Semiconductor 101] When Semiconductors & SK hynix Made Their Mark on the World first appeared on SK hynix Newsroom.

FMS 2024 features a broader focus on different memory types compared to previous years

SK hynix has returned to Santa Clara, California to present its full array of groundbreaking AI memory technologies at FMS: the Future of Memory and Storage (FMS) 2024 from August 6–8. Previously known as Flash Memory Summit, the conference changed its name to reflect its broader focus on all types of memory and storage products amid growing interest in AI. Bringing together industry leaders, customers, and IT professionals, FMS 2024 covers the latest trends and innovations shaping the memory industry.

Participating in the event under the slogan “Memory, The Power of AI,” SK hynix is showcasing its outstanding memory capabilities through a keynote presentation, multiple technology sessions, and product exhibits.

Keynote Presentation: Envisioning the Future of AI With Leading Memory & Storage Solutions

Keynote presentations have always been a highlight at FMS. The talks act as an important forum for attendees to learn about emerging memory and storage technologies from industry leaders. Due to its leadership in the AI memory field, SK hynix was selected to give a keynote presentation titled “AI Memory and Storage Solution Leadership and Vision for the AI Era” at the newly expanded event.

Held on the first day of FMS 2024, the keynote was delivered by Vice President Unoh Kwon, head of HBM Process Integration (PI) and Vice President Chunsung Kim, head of SSD Program Management Office (PMO). The pair provided insights into the company’s DRAM and NAND flash solutions which aim to solve the pain points of generative AI and promote continued development of AI. These pain points include maximizing the efficiency of AI training and inference¹ while minimizing floor space and power utilization for storing data.

¹AI inference: The process of running live data through a trained AI model to make a prediction or solve a task.

Kwon and Kim delivering their keynote on SK hynix’s leading AI memory products

Each speaker covered a particular type of memory. Kwon touched on the company’s DRAM memory products that are optimized for AI systems such as HBM², CXL®³, and LPDDR5T⁴. Meanwhile, Kim introduced the company’s best-in-class NAND flash storage devices including its SSD and UFS⁵ solutions, which will continue to be crucial for AI applications in the future.

²High Bandwidth Memory (HBM): A high-value, high-performance product that revolutionizes data processing speeds by connecting multiple DRAM chips with through-silicon via (TSV).
³Compute Express Link^® (CXL^®): A PCIe-based next-generation interconnect protocol on which high-performance computing systems are based.
⁴Low Power Double Data Rate 5 Turbo (LPDDR5T): Low-power DRAM for mobile devices aimed at minimizing power consumption and featuring low voltage operation. LPDDR5T is an upgraded product of the LPDDR5X and will be succeeded by LPDDR6.
⁵Universal Flash Storage (UFS): A high-performance interface for computing and mobile systems which require low power consumption.

In addition to its keynote, SK hynix also held five sessions that took a deeper look into its next-generation products set to solidify the company’s AI technology leadership. These sessions covered a variety of topics, including the company’s DRAM, SSD, and CXL solutions.

Product Booth: Presenting the Industry’s Best AI Memory

SK hynix’s booth at FMS 2024 consists of four sections showcasing many of the products which featured in the company’s keynote and session talks. One of the booth’s highlights is the samples of the 12-layer HBM3E, the next-generation AI memory solution which is expected to be mass-produced in the third quarter of 2024. The company is also holding demonstrations of select products with its partners’ systems, highlighting its strong collaboration with various major tech companies.

The AI Memory and Storage section includes industry-leading solutions such as SK hynix’s HBM3E and eSSD portfolio

• AI Memory and Storage: Features SK hynix’s flagship AI memory products such as samples of its 12-layer HBM3E, which is set to be the same height as the previous 8-layer version under JEDEC⁶ standards. The section also includes GDDR6-AiM⁷, a product suitable for machine learning due to its computational capabilities and rapid processing speeds, and the ultra-low power LPDDR5T optimized for on-device AI. Storage solutions on display include the PCIe Gen5-based eSSD PS1010, which is ideal for AI, big data, and machine learning due to its rapid sequential read speed.

⁶Joint Electron Device Engineering Council (JEDEC): A U.S.-based standardization body that is the global leader in developing open standards and publications for the microelectronics industry.
⁷Accelerator in Memory (AiM): A special-purpose hardware made using processing and computation chips.

At the NAND Tech, Mobile, and Automotive section, attendees can see products such as the company’s 321-layer NAND and ZUFS 4.0

• NAND Tech, Mobile, and Automotive: This section includes solutions such as the world’s highest 321-layer wafer technology as well as triple-level cell (TLC) and QLC NAND. Mobile technologies are also showcased including ZUFS⁸ 4.0, an industry-best NAND product optimized for on-device AI⁹ which boosts a smartphone’s operating system speed compared to standard UFS.

⁸Zoned Universal Flash Storage (ZUFS): A NAND flash product that improves efficiency of data management. It optimizes data transfer between an operating system and storage devices by storing data with similar characteristics in the same zone of the UFS.
⁸On-device AI: A technology that implements AI functions on the device itself, instead of going through computation by a physically separated server.

The AI PC and CMS 2.0 section features a demonstration of the company’s industry-leading SSD, PCB01

• AI PC and CMS 2.0: Attendees can see a system demonstration of the industry-best SSD PCB01, which is able to efficiently process large AI computing tasks when applied to on-device AI PCs. In addition, CMS¹⁰ 2.0, a next-generation memory solution that boasts equivalent data processing capabilities as a CPU, is applied to a vector database¹¹.

¹⁰Computational Memory Solution (CMS): A product that integrates computational functions into CXL memory.
¹¹Vector Database: A collection of data stored as mathematical representations, or vectors. As similar vectors are grouped together, vector databases can make low-latency inquiries, making them ideal for AI.

The OCS, Niagara, CMM section displays innovative solutions such as Niagara 2.0 and HMSDK

• OCS, Niagara, and CMM: This section features a demonstration of OCS¹² technology which enhances the data analysis and pooled memory solution Niagara 2.0¹³. It also includes a demonstration of CMM¹⁴– DDR5¹⁵, which expands system bandwidth by 50% and capacity by up to 100% compared to systems only equipped with DDR5.

¹²Object-based Computational Storage (OCS): A new computational storage platform for data analytics in high-performance computing. OCS has not only high scalability but also data-aware characteristics that enable it to perform analytics independently without help from compute nodes.
¹³Niagara 2.0: A solution that connects multiple CXL memories together to allow numerous hosts such as CPUs and GPUs to optimally share their capacity. This eliminates idle memory usage while reducing power consumption.
¹⁴CXL Memory Module (CMM): A new standardized interface that has an advantage in scalability and helps increase the efficiency of CPUs, GPUs, accelerators, and memory.
¹⁵Double Data Rate 5 (DDR5): A server DRAM that effectively handles the increasing demands of larger and more complex data workloads by offering enhanced bandwidth and power efficiency compared to the previous generation, DDR4.

Super Women Conference: Cherishing Diversity in the Memory & Storage Industry

Haesoon Oh delivers the keynote at the FMS Super Women Conference

SK hynix also co-sponsored the FMS Super Women Conference, an event held on the sidelines of FMS 2024 which celebrates the achievements of female leaders and promotes diversity in the memory industry. Head of NAND Advanced PI Haesoon Oh, the company’s first female executive-level research fellow, delivered a keynote address on the company’s next-generation innovations and the importance of understanding diversity.

Paving the Way Forward in the AI Era

At FMS 2024, SK hynix underlined its commitment to lead the industry by providing integrated AI memory solutions and expanding its expertise in the sector. Collaborating with other leading partners, the company will strive to provide customers with the best possible solutions that match their rapidly changing needs.

The post SK hynix Presents Extensive AI Memory Lineup at Expanded FMS 2024 first appeared on SK hynix Newsroom.

• Development of PCB01, 5th generation of 8-channel PCIe, to be followed by mass production within this year
• PCB01, industry’s best SSD for PCs, optimized for on-device AI
• Advancement in NAND solution to extend success story of HBM, solidify leadership in AI memory space

Seoul, June 28, 2024

SK hynix Inc. (or “the company”, www.skhynix.com) announced today that it developed PCB01, an SSD product with the industry’s best specifications, for on-device AI¹ PCs.

The product marks the first case where the industry adopts the fifth generation of the 8-channel² PCIe³ technology and brings innovation to performance including the data processing speed.

¹On-device AI: a technology that implements AI functions on the device itself, instead of going through computation by a physically separated server. A smartphone’s direct collection and computation of information allows fast reactions of the AI performance, while promising an improved customized AI service.
²Channel: a route for the input/output of data between a NAND flash and a controller on the SSD. An increase in the number of channels leads to advancement of the PCIe to the next generations and an improvement in the data processing speed. A 4-channel SSD is typically adopted for the conventional PCs, while an 8-channel SSD is for high-performance PC.
³Peripheral Component Interconnect Express (PCle): a serial-structured, high-speed input/output interface used on the motherboard of digital devices

The company expects the latest advancement in the NAND solution space to add to its success stories in the high-performance DRAM area led by HBM, enhancing its leadership in the overall AI memory space.

With a validation process with a global PC customer underway, SK hynix plans to mass produce and start shipping the products to both corporate customers and general consumers within this year.

PCB01 comes with the capabilities of sequential read and write speeds of 14GB and 12GB per second, respectively, bringing the performance of an SSD to the level unseen before. The speeds allow the operation of a large language model⁴, or LLM, for AI training and inference, in a second.

⁴Large language model (LLM): a language model trained on vast amounts of data, which is essential for performance of generative AI tasks such as creating, summarizing, and translating texts.

The product also comes with an improvement in power efficiency of more than 30% compared with the previous generation, enhancing the stability of the large-scale AI computing tasks.

SK hynix also adopted the SLC⁵ caching for the production of PCB01. With the adoption of the technology that places the single-level cell, or SLC, in some parts of the NAND cell to accelerate performance, a PC user can experience a faster performance for both AI services and conventional computing.

The product is also equipped with the capability aimed at protecting personal data. SK hynix engineers built the root of trust⁶, or ROT, a security solution, in the PCB01 to prevent external cybersecurity attacks and forging and falsification of information, while protecting a user’s password.

⁵Single-level cell (SLC): a type of memory cell used in NAND flash that stores one bit of data in a single cell. As the amount of data stored increases, the memory cell becomes a multi-level cell (MLC), a triple-level cell (TLC), and a quad-level cell (QLC). An increase in the data storage means more data can be stored in the same area, but the speed and stability decreases. An SLC enables faster processing of selected data.
⁶Root of Trust (RoT): an area of the hardware that can always be trusted in terms of security, while enabling prevention of forging and falsification of information.

The company plans to launch PCB01 in three capacities – 512GB, 1TB, and 2TB.

Ahn Hyun, Head of the N-S Committee at SK hynix, said that numerous global providers of CPU for on-device AI PCs are requesting collaboration for compatibility validation process. “We will work towards enhancing our leadership as the global top AI memory provider also in the NAND solution space by successfully completing the customer validation and mass production of PCB01, which will be in the limelight.”

About SK hynix Inc.

SK hynix Inc., headquartered in Korea, is the world’s top tier semiconductor supplier offering Dynamic Random Access Memory chips (“DRAM”), flash memory chips (“NAND flash”) and CMOS Image Sensors (“CIS”) for a wide range of distinguished customers globally. The Company’s shares are traded on the Korea Exchange, and the Global Depository shares are listed on the Luxembourg Stock Exchange. Further information about SK hynix is available at www.skhynix.com, news.skhynix.com.

Media Contact

SK hynix Inc.
Global Public Relations

Technical Leader
Kanga Kong
E-Mail: global_newsroom@skhynix.com

Technical Leader
Sooyeon Lee
E-Mail: global_newsroom@skhynix.com

The post SK hynix Develops PCB01 for Artificial Intelligence PCs first appeared on SK hynix Newsroom.

SK hynix’s executive team just got younger. As part of efforts to foster young leadership, Lee Donghun was appointed to the N-S Committee¹ at the end of 2023 to become SK hynix’s youngest-ever executive.

¹N-S Committee: SK hynix’s strategic unit formed at the beginning of 2024 that fortifies its NAND flash and solution businesses.

Born in 1983, Lee is a young talent who finished his master’s and doctoral studies after participating in SK hynix’s scholarship program in 2006. He joined the company in 2011 and played a key role in the development and evolution of 4D NAND. Lee led the Technological Strategy team for the 128- and 176-layer NAND flash before working as the head of the Performance and Reliability (PnR) team for the 238-layer NAND flash, contributing significantly to establishing SK hynix’s 4D NAND as the industry standard.

As part of the newsroom’s interview series with newly appointed executives, Lee spoke about the future of NAND flash development and his aspirations as a leader.

Flexibility is Crucial in Times of Change

Lee anticipates reaching new heights in NAND flash technology by adopting a flexible approach

Lee emphasized the importance of rapid and flexible responses to the rapidly changing technological landscape following the advent of the AI era.

“Advances in technology are rapidly changing people’s lives, so responding to these changes will be critical for companies’ future,” he said. “A good comparison is the emergence and subsequent impact of the internet and smartphones.”

The N-S Committee is an example of SK hynix’s response to lead this changed business environment. As generative AI and other applications need to store ever-growing amounts of data, NAND flash requires assistance from NAND solutions. Accordingly, the N-S Committee will act as a control tower to optimize the NAND flash and NAND solution businesses simultaneously and strengthen the company’s competitiveness by increasing development efficiency and customer satisfaction.

“In order to respond flexibly when new technologies emerge, all related divisions need to constantly share information and collaborate,” Lee said. “Consequently, collaboration between the NAND flash and NAND solutions developments is not optional but necessary if we want to speed up the process of developing products desired by our customers. As a new executive, I plan to make such collaborations even more efficient.”

Beyond 4D NAND: The NAND Flash Revolution in the AI Era

Lee believes that further development in NAND flash technology holds the key to leading the AI era

Lee expressed his excitement about working on the world’s highest 321-layer 4D NAND to ensure the product’s performance, reliability, and quality.

“The 321-layer 4D NAND we’re developing is set to be a new milestone in the industry with its superior performance, and that’s why our role is so important,” Lee said. “For this product, it’s not just about performance, it’s also about ensuring reliability. If we focus only on rapidly increasing demand, there are potential risks to quality and reliability. Our short-term goal is to finish development and deliver products as quickly as possible while minimizing risks.”

Lee stressed that, in the long run, the company needs to continue innovating by taking on new challenges.

“Until now, the key to NAND flash development has been to improve cost-effectiveness,” he said. “That’s why we’ve seen 2D and 3D NAND in the past, followed by 4D NAND. Now that we’re in the midst of a transformation, NAND needs to innovate in multiple directions.”

Lee also emphasized the need to focus on the data explosion expected across various fields.

“As the number of sectors utilizing AI technology expands, there will also be an increase in data generation,” Lee explained. “In the automotive space alone, data related to roads and traffic for autonomous driving is rapidly growing. Depending on the type of device or environment which generates data, the standards and conditions required from NAND flash products may evolve drastically. It will be my job to anticipate these changes in the industry and lead proactive innovations, ensuring SK hynix continues its technological leadership.”

Building Empathy With Employees Pivotal in Overcoming Challenges

Lee sees team effort as the answer to overcoming unexpected challenges

Lee also sees tailwinds blowing in the semiconductor industry today as he predicts that NAND flash, like DRAM, will go on an upward direction in 2024. However, he predicted that there will be vast innovations in various technological fields that will create more challenges for SK hynix employees.

“In the midst of so many challenges, it is important that our members understand why we need to overcome these obstacles while finding the right motivation to continue progressing,” he said. “At the end of the day, the company and its people need to be on the same page to achieve a common goal.”

Even in his New Year’s greeting to his team members, Lee emphasized the need to pay attention to pivotal changes in the technological space and flexibly respond to situations.

“The semiconductor memory market is on the upswing in 2024, and I think we can greet this upturn with a smile as we faced a lot of adversity. However, there are still challenges ahead,” Lee explained. “Especially as we are expecting the launch of next-generation NAND flash products this year, I will do my best to help the company navigate this period of drastic transformation. We hope to accomplish various goals in 2024.”

<Other articles from this series>

New Leadership Spotlight: SK hynix’s First Female Research Fellow, Vice President Oh Haesoon, on Advancing NAND Flash

New Leadership Spotlight: Vice President Kitae Kim, Head of HBM Sales & Marketing, on Future-Proofing HBM’s Market Leadership

New Leadership Spotlight: Vice President Hoyoung Son Targets SK hynix’s Evolution Into Total AI Memory Provider Through Advanced Packaging Tech

Leadership Spotlight: Deoksin Kil, Head of Material Development, On Achieving Tech Innovation Through Advanced Materials

New Leadership Spotlight: Head of HBM PI Unoh Kwon Aims to Finish SK hynix’s HBM Roadmap & Lead in the AI Era

New Leadership Spotlight: Global RTC Head Jaeyun Yi Aims to Present a New Paradigm for the Future of Semiconductors

New Leadership Spotlight: SK hynix Execs Join Roundtable to Discuss Company’s AI Memory Leadership & Future Market Trends

The post New Leadership Spotlight: SK hynix’s Youngest Exec, Lee Donghun, Discusses Spearheading NAND Flash Development first appeared on SK hynix Newsroom.

Oh Haesoon broke new ground at the end of 2023 when she was appointed SK hynix’s first female research fellow¹. In her new role, Oh leads the newly formed N-S Committee, a strategic unit which aims to strengthen SK hynix’s NAND flash and solutions businesses.

¹Research fellow: SK hynix executives with outstanding technical capabilities who focus on researching innovative technologies.

Her appointment follows her lengthy tenure in the company’s NAND business. After working at the R&D center and DRAM development division, Oh has dedicated her career to developing next-generation NAND flash platforms since 2007. During this time, she made significant technological contributions including the successful development of the company’s first 3D NAND technology and QLC² products, and the mass production of 4D NAND.

²Quad-level cell (QLC): A form of NAND flash memory that can store up to 4 bits of data per memory cell.

In an interview which kicks off a series of talks with newly appointed executives, we spoke with Oh about her ambitions as the first female research fellow and her vision for developing the NAND flash business.

Driving Innovation Through Diversity

Oh claims diversity among researchers leads to technological breakthroughs

Oh believes that embracing diversity is key to fostering technological innovation. She claims that diverse perspectives enable researchers to transcend biases and challenge existing paradigms, paving the way for groundbreaking solutions.

“In the high-tech world of semiconductor research, technological capability is above everything else,” she said. “As a research fellow, I do feel a tremendous sense of responsibility to demonstrate my technical leadership. To that end, I will strive to foster innovation by promoting diversity in our research culture.”

Oh also expressed her excitement at becoming a role model for her female team members, helping them to grow and redefining what it means to be a research fellow.

“I don’t think men and women have different research capabilities,” she said. “But, as a female leader, I am sure that I will play a distinct role in this field, where the integration of diverse perspectives is essential for progress. Despite the pressures associated with being the first female research fellow, I am committed to fulfilling my responsibilities and blazing this path for many more women in the future.”

Propelling the NAND Flash Upturn Through Research Leadership Covering Mass Production

Oh cites high-quality research as a crucial factor in semiconductor development

Oh currently leads the Advanced Process Integration³ (PI) group, which focuses on the development of next-generation, high-value NAND flash products. Her research emphasizes achieving successful mass production by minimizing trial and error in the entire process from product development to mass production itself. These points are especially crucial in the NAND flash sector as product competitiveness is closely tied to the competitiveness in mass production.

³Process integration (PI): A holistic approach to semiconductor development which considers the entire manufacturing process from core development to increasing yield after mass production. It involves performing tasks such as determining cell schemes and deriving optimal design rules for early-phase and next-generation product development.

“Working across different phases of NAND flash development has provided me with a deeper understanding of the needs and objectives of each stage and a more comprehensive perspective of the overall process,” she said.

In 2022, leveraging her expertise in both product development and mass production, Oh pioneered the adoption of the On-Die Electrical Parameter Monitoring⁴ (ODE) system for NAND flash. This system manages production defects from the development stage and is actively being used for quality specification management and product defect control.

¹On-Die Electrical Parameter Monitoring (ODE): A series of die tests which measure transistor characteristics such as resistance and capacitance to predict the electrical characteristics and overall performance of a chip.

When asked about the outlook for NAND flash, Oh predicts a full-scale rebound in 2024 based on the company’s long-standing technological strengths.

“After DRAM’s successful upturn in 2023, NAND flash is poised for a similar shift this year. Our key mission for 2024 is to secure the underlying technology to overcome stacking limitations and at the same time, develop the next generation of high-value products in a timely manner,” she said. “We will continue research aimed at reducing trial and error from the development stage to improve profitability.”

Forging Synergies Between the NAND Flash and Solutions Businesses

Oh believes cross-department collaboration plays a key role in strengthening NAND flash development

Oh counts the successful development of SK hynix’s first 3D NAND platform in 2014 as her most memorable achievement. Building a thriving platform from the ground up during a transformative period, where both structure and characteristics underwent complete changes, is a source of great pride and confidence for her. Since then, Oh has also spearheaded several other high-profile projects, such as the development of the company’s first QLC product and the mass production of 4D NAND. Her experience has underscored the importance of synergy and collaboration as essential elements for success.

“Technology will continue to advance and so will the barriers that need to be overcome,” she said. “However, there’s no need to be discouraged by the technical difficulty. By harnessing the ideas of the team and working collaboratively to create solutions, we can accomplish our goals.”

Oh believes that the collaboration with the solutions division is vital for boosting the competitiveness of the NAND flash business. The company’s NAND flash, which needs to efficiently store a large amount of data at low cost, is more effective when equipped with tailored storage products from the solutions division. Oh emphasized the importance of the synergy between the two divisions, especially as the N-S Committee took its first steps in 2024 as the control tower for both businesses.

While acknowledging the challenges ahead, Oh shared her vision for the NAND flash business in 2024.

“This year, the NAND flash and solutions divisions will actively communicate and collaborate with one another to create synergies within the N-S Committee, and take the initiative to propel the NAND flash business forward in 2024,” she said.

Rising Up to Meet New Challenges in 2024

Wrapping up the interview, Oh turned her attention to her own goals for 2024 as she looks to continue her professional development at SK hynix.

“2024 will be a year of great challenge and innovation for me,” she said. “Above all, I will strive to be a leader who grows alongside our team and makes a real contribution to the company.”

<Other articles from this series>

New Leadership Spotlight: SK hynix’s Youngest Exec, Lee Donghun, Discusses Spearheading NAND Flash Development

New Leadership Spotlight: Vice President Kitae Kim, Head of HBM Sales & Marketing, on Future-Proofing HBM’s Market Leadership

New Leadership Spotlight: Vice President Hoyoung Son Targets SK hynix’s Evolution Into Total AI Memory Provider Through Advanced Packaging Tech

Leadership Spotlight: Deoksin Kil, Head of Material Development, On Achieving Tech Innovation Through Advanced Materials

New Leadership Spotlight: Head of HBM PI Unoh Kwon Aims to Finish SK hynix’s HBM Roadmap & Lead in the AI Era

New Leadership Spotlight: Global RTC Head Jaeyun Yi Aims to Present a New Paradigm for the Future of Semiconductors

New Leadership Spotlight: SK hynix Execs Join Roundtable to Discuss Company’s AI Memory Leadership & Future Market Trends

The post New Leadership Spotlight: SK hynix’s First Female Research Fellow, Vice President Oh Haesoon, on Advancing NAND Flash first appeared on SK hynix Newsroom.

This final episode in the Tech Pathfinder series will introduce SK hynix’s advanced 4D NAND technologies. These include its 4D^1.0 technologies, which specialize in stacking and performance improvement such as the Cost-Effective 3-Plug formation, Sideway Source, All Peri.¹ Under Cell (PUC), and Advanced Charge Trap Flash (CTF). It will also cover the 4D^2.0 NAND technologies which overcome the limitations of stacking, such as Multi-Site Cell (MSC).

¹Peripheral circuit (peri.): A circuit that controls the cell.

The Basics of NAND Flash Memory

For a better understanding of 4D NAND technology, it is prudent to review NAND’s basic concepts and related terminology.

Figure 1. An overview of different types of NAND flash memory

A cell is the smallest unit in which information is stored. In NAND flash memory, the cells consist of a control gate and a floating gate. When voltage is applied to the control gate, electrons traveling through the pathway are stored in the floating gate. NAND flash stores data by categorizing cells as either 0 or 1 using electrons stored on the floating gate. This state is characterized by the number of electrons in a cell. For example, a cell with few electrons is read as 0, while a cell with a high number of electrons is interpreted as 1.

NAND flash memory is categorized into different types depending on how much information (bits) is stored in a single cell. These include single-level cell (SLC, 1 bit), multi-level cell (MLC, 2 bits), triple-level cell (TLC, 3 bits), quad-level cell (QLC, 4 bits), and penta-level cell (PLC, 5 bits). As for the units used to measure NAND flash memory capacity, these include references to giga (a billion) and tera (a trillion). In other words, a TLC NAND flash product with a capacity of 1 Tb has about 330 billion cells that store 3 bits each.

4D^1.0 Technology: Reducing Chip Size Through Cell Stacking

There are four main 4D^1.0 NAND technologies SK hynix has employed to develop high-capacity NAND flash solutions.

Figure 2. An overview of the Cost-Effective 3-Plug formation and Sideway Source

Cost-Effective 3-Plug Formation

One of the key goals of developing semiconductor technology is improving cost efficiency. This is achieved by stacking more cells to reduce the chip size and producing as many chips as possible on a single wafer. Stacking substrates layer-by-layer and repeating the cell formation process for each layer would be inefficient and increase manufacturing costs. Therefore, multiple layers of substrate are first stacked, then vertical holes called plugs are drilled through the layers before cells are formed next to the holes.

As the number of layers increases, the more challenging it becomes to form plugs to the bottom layer as existing etching equipment can only etch around 100 layers at a time. Therefore, to develop a NAND flash product with more than 300 layers, it is necessary to stack 100 layers and perform the plug etching process three times. This is where SK hynix’s Cost-Effective 3-Plug formation is used as all the processes, including cell formation, can be performed simultaneously on all layers.

With this, SK hynix was able to conduct a single process to simultaneously fabricate the key structures—word lines² and word line staircases³—that apply voltage and the passageways for electrons. This enabled the company to unveil a 321-layer 4D NAND of the highest density in August 2023 while minimizing costs.

²Word lines: The structure that binds the control gate of each layer of NAND cells.
³Word line staircases: A staircase-like structure for exposing the word line of each layer to the top surface.

Sideway Source

Semiconductor plugs provide a pathway for electrons to travel. Inside a plug, this pathway is covered by CTF film⁴. Therefore, the CTF film needs to be removed at the connection point where the plug and the bottom of the NAND flash layer meet to connect two pathways. Sideway source connects the plug to the bottom of the NAND flash layer (channel and source line⁵). Previously, etching gas was injected from the top of the plug to vertically remove the CTF film at the bottom of the plug. However, when stacking two or more plugs, the centers of the plugs were not aligned. This prevented the etching gas from reaching the bottom, damaging the CTF film on the side of the plug that serves as a cell.

⁴CTF film: A composite of oxide and nitride films that replaces the floating gate.
⁵Source line: Located at the bottom of a NAND layer, the source line is part of a channel inside the plug. Electrons from the source line travel up the channel to the top of the NAND layer and are stored in their respective floating gates.

SK hynix solved this issue by replacing the vertical connection with a horizontal one. The etching gas is injected into a separate pathway to reach the bottom of the NAND layer and remove the CTF film on both sides of the plug.

With Sideway Source technology, the etching gas is not directly injected into the plug. Therefore, even if the plugs are misaligned, the interior remains undamaged. As a result, SK hynix has significantly reduced its defect rate, increased productivity, and addressed the problem of increased costs associated with multiple stacking.

Since SK hynix introduced the industry’s first 4D NAND in 2018, it has enhanced its expertise to produce precise horizontal pathway connections which leave no voids at the bottom of the NAND layer. Based on this advancement, the company improved production efficiency by 34% for its 238-layer NAND flash memory compared to the 176-layer product and further solidified its market leadership with its 321-layer NAND.

Figure 3. An overview of All Peri. Under Cell (PUC)

All Peri. Under Cell (PUC)

PUC reduces the chip size and increases the number of stacks by placing the peripheral circuit (peri.) under the cell. SK hynix used PUC to develop a new NAND flash structure, the world’s first 4D NAND, and then began product development. The company has further developed upon PUC with its All PUC technology, which miniaturizes the peri. so it becomes the same size as the cell or smaller to accommodate the reduced cell size. To advance the technology, SK hynix is further miniaturizing the peri. by reducing the size and number of transistors and fully placing the peri. in the empty space under the cell.

In particular, this technology was used for the first time to great effect in SK hynix’s 238-layer 512 Gb TLC NAND. For this solution, the company reduced the size of the chip and peri. by more than 30% compared to the previous generation, thus improving production efficiency and cost competitiveness. SK hynix will continue to enhance its expertise and perfect the technology so it can be applied to future products that require a smaller peri. and chips.

Figure 4. An overview of Advanced Charge Trap Flash (CTF)

Advanced Charge Trap Flash (CTF)

Advanced CTF minimizes data degradation by retaining more electrons than conventional CTF. In CTF, electrons are stored in nonconductors rather than in conductors such as a floating gate. CTF was therefore developed in part to address inter-cell interference⁶ in conductors by changing the electron storage space to nonconductors. However, electrons often escape from nonconductors as they are stored in the voids of the CTF material (nitrogen-silicon compound) which has unstable areas. When electrons are stored in these unstable areas, the bonds quickly break and the electrons are ejected, resulting in data loss.

⁶Inter-cell interference: Electrons in a cell are affected by electrons in adjacent cells due to device miniaturization, resulting in data corruption.

For its Advanced CTF, SK hynix fills the unstable areas with hydrogen to prevent electrons from entering, and increases the number of binding agents to store more electrons. Furthermore,  Advanced CTF also increases the number of electrons stored in CTF by minimizing the risk of escaped electrons. This improves the ability to determine electron counts, reduces read errors, and significantly shortens latency.

Some types of NAND flash have difficulties distinguishing data when there are a low number of electrons, resulting in errors. For example, if SLC flash memory distinguishes data using ten electrons, data with one to five electrons is 0, and data with six to ten electrons is 1. However, if five electrons escape, the data previously processed as 1 is distorted and an error occurs. This problem worsens as a cell is segmented to the MLC level and higher.

TLC differentiates between eight states from 000 to 111. If there are 10 electrons to distinguish, each state is assigned either one or two electrons. This is a significant difference from SLC, which allocates five electrons per state. Consequently, even if only a few electrons escape, it can lead to data corruption.

In contrast, consider a situation in which Advanced CTF was used to distinguish data with 100 electrons. If the number of electrons is between 0 and 50, the data is read as 0, while if it is between 51 and 100, it is 1. Even if some electrons escape, the large number of electrons overall greatly reduces the chance of misreading the data. Since there are few errors, the latency is shortened, and the read speed increases.

SK hynix first applied Advanced CTF to its 176-layer NAND solution, resulting in a 25% improvement in the ability to determine electron counts. As Advanced CTF-based memory solutions offer lower latency, they are particularly suited for the gaming and automotive markets which require rapid data processing.

4D^2.0 Technology: Increasing Horizontal Cell Density & Stacking for Enhanced Performance & Density

When developing semiconductor memory, manufacturing costs continue to rise with each additional layer. Taking into account the additional cost of increasing the number of bits beyond the TLC level, there comes a point where it is no longer possible to reduce costs. In response, SK hynix is developing 4D^2.0 technology which increases the number of layers and horizontal density of cells to improve storage capacity relative to cost. Multi-Site Cell (MSC) is a 4D^2.0 technology that structurally improves the horizontal density, thereby significantly increasing the number of bits.

Figure 5. An overview of Multi-Site Cell (MSC)

Multi-Site Cell (MSC)

There are two primary methods for horizontally expanding cell density. The first is multi-level cell (MLC) technology, which subdivides electron counts to accommodate more data (bits) in a single cell. This is the case with NAND flash types ranging from SLC to QLC. The second is MSC technology, which structurally increases the sites where electrons are stored in a cell, enabling it to hold more data (bits).

MLC technology has been commercialized in 4-bit QLC products, but it is challenging to maintain performance and reliability in 5-bit PLC and beyond. This is due to the previously mentioned limitations in determining electron counts.

For example, if you build a 6-bit hexa-level cell (HLC) with MLC, you need to store data in 64 different states ranging from 000000 to 111111. This is prone to errors and time-consuming because there are not enough electrons to distinguish each state. Compared to the 4-bit QLC, the ability to determine the number of electrons is four times poorer.

On the other hand, when developing an HLC with MSC, eight states from 000 to 111 are created in two spaces and multiplied to realize 64 states to store data. Compared to the 4-bit QLC, the ability to distinguish electron counts doubles. In other words, it has the capacity of an HLC but the speed of a TLC. SK hynix has confirmed a 20-fold improvement⁷ in read and write speeds when utilizing MSC. Due to MSC’s high capacity, rapid speed, and reliability, SK hynix’s NAND flash is the leading solution for future multimodal AI⁸.

⁷Comparison between a 5-bit regular cell and a 2.5-bit × 2.5-bit MSC
⁸Multimodal AI: AI that can simultaneously process text, speech, images, etc.

Solving Industry Problems With An Eye on the Future

In this final Tech Pathfinder episode, SK hynix’s 4D NAND technologies were shown to solve the industry issues of today and tomorrow. The company’s 4D^1.0 technologies improve the cost-effectiveness and performance of its NAND flash, while its 4D^2.0 technologies will overcome stacking limitations set to arise in the future.

<Other articles from this series>

[Tech Pathfinder] HKMG Opens the Door to Leading Mobile DRAM LPDDR5X & LPDDR5T

[Tech Pathfinder] Small Size, Big Impact: Unveiling the Latest Advances in Semiconductor Packaging and Miniaturization

The post [Tech Pathfinder] How SK hynix’s Advanced 4D NAND Technologies Are Overcoming Stacking Limitations first appeared on SK hynix Newsroom.

As part of its development of next-generation 4D NAND flash products, SK hynix recently developed products with 238 layers, the world’s highest. This could be thought of as difficult as cramming 4 billion 238-story buildings onto a dime-sized area.

These miniature skyscrapers are what allow us to store all of our life’s memories and heavy workloads on PC storage devices and servers, as well as many other devices around us.

The newsroom looks back at some of the NAND breakthroughs SK hynix has produced to push the boundaries of technology while obtaining a market advantage and being acknowledged as an industry standard.

Increasing demand for high-performance NAND

An ever-growing demand for NAND has been the force driving behind SK hynix’s efforts to meet customer needs. Price is important as it always has been, but performance has become increasingly crucial in recent years.

Take digital cameras: they now have a huge number of pixels, and a memory capable of storing a large amount of data is required for just one image. Consumer demand for fast read and write speeds is also increasing as SSDs become more popular.

Until early 2010, the most pressing issue in memory technology was the scaling of 2D NAND, that is, making things smaller. However, due to limitations in density and fine patterning, 3D NAND became the focus of all manufacturers, including SK hynix.

3D NAND development history at SK hynix

3D NAND literally refers to the stacking of memory cells vertically. The bit density improves as more layers are added, increasing the storage capacity.

In November 2014, SK hynix was able to develop 3D NAND chips featuring 24 layers. Then, in August 2015, it developed 36-layer 128Gb 3D NAND based on its TLC arrays technology, followed by 48-layer 256Gb TLC in November 2016.

Just 5 months later, the company introduced the industry’s first and highest productivity 72-Layer 256Gb TLC 3D NAND, with 30% higher productivity and 20% higher performance than 48 layers, thereby establishing its business competency in 3D NAND memory solutions.

This was also a significant accomplishment given the importance of NAND memory in this new era of AI, Big Data, and Cloud storage.

Breakthrough from 3D to 4D NAND

Figure 1. Concept diagram representing the evolution from 2D to 4D NAND

At SK hynix, it was also important to innovate 3D NAND and think outside the box. This included the development of PUC (Peri Under Cell), a technology that puts peripheral circuits, generally placed on the side, under cell circuits to reduce chip space, giving rise to 4D NAND. Two other core competencies that are applied to its 4D NAND products include Sideway Source, a technology that horizontally connects the source, and Advanced CTF, that stores electric charges in a CTF (Charge Trap Flash).

In November 2018, SK hynix achieved a major milestone in the company’s NAND flash business with the successful development of the industry’s first CTF-based 96-layer 512Gb TLC 4D NAND.

SK hynix then introduced the world’s first 128-layer 1Tb TLC 4D NAND, dramatically enhancing profitability with 40% greater productivity and 60% better investment efficiency, targeting the high-capacity mobile storage solutions and enterprise SSD markets.

In December 2020, the company developed the industry’s highest 176-layer 512Gb 4D NAND. In a recent quality evaluation survey of leading mobile device and SSD manufacturers, SK hynix’s 128-layer and 176-layer NAND products ranked first.

Figure 2. SK hynix 238-layer 512Gb TLC 4D NAND

In August of 2022, SK hynix developed the world’s first 238-layer 512Gb TLC 4D NAND, which was introduced at the Flash Memory Summit 2022 in Santa Clara. Besides increasing productivity and data transfer rate, energy consumption for data reading is reduced by 21%, contributing to the fulfillment of ESG management goals.

Thanks to the dedication and tireless work of its developers, SK hynix’s current 4D NAND technology is now acknowledged as an industry standard.

NAND 4D^2.0: Setting a new standard

Current 4D NAND reduces chip sizes thanks to PUC. While this allows the implementation of high density within a fixed area, the disadvantage is that the stacking technology may reach its limit in the future.

Figure 3. Jungdal Choi, Head of NAND Development of SK hynix (right), delivers his keynote speech with Sanjay Talreja of Solidigm (left) at the Flash Memory Summit 2022

SK hynix plans to overcome this barrier by using Multi-Site Cells (MSCs) as the core, once again demonstrating its technological prowess ahead of others. These store data by dividing existing cells into two smaller cells with micro fabrication, reducing the number of cell stacks while horizontally expanding cell density, one of the core elements of the technological concept known as 4D^2.0. Such disruptive technology is projected to become the industry’s next-generation standard.

Bold investments and foresight key to future success

As the NAND market has grown, so has SK hynix’s success in developing and mass-producing NAND solutions that are significantly different from conventional NAND, giving it a competitive advantage.

SK hynix has concentrated not just on stacking, but also on technological innovation and new ways of thinking, all part of the company’s long-term investment in R&D. By optimizing the arrangement of cells, SK hynix has been able to increase performance enough to meet the changing needs of customers.

Limitations will no doubt surface as NAND technology advances, but SK hynix is committed to delivering the industry’s best products while continuing to think outside the box and deliver new, timely, and differentiated technologies.

The post NAND Technology Development at SK hynix: Reaching New Heights first appeared on SK hynix Newsroom.