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

 

The journey of semiconductors starts with a grain of sand and ends with a groundbreaking technology that impacts lives around the world. Just as these complex microchips have various components, this second episode in the Semiconductor 101 series will break down the key aspects of semiconductors. From the types, functions, and specifications of semiconductors to the challenges and future trends in the industry, learn more about the foundations of modern technologies.

What are the different types of semiconductors?

Semiconductors can be classified according to various criteria such as material composition and the purity of these materials. However, one of the most common classifications is based on their functionality. On this basis, there are three main types of semiconductor chips: memory, logic, and a broader group comprising discrete, analog, and other (DAO) chips.

• Memory: As their name suggests, memory chips are optimized for data storage. They ensure that systems can retain data either permanently or temporarily and rapidly access stored data. These chips can be categorized into volatile and non-volatile memory, which will be explored further in the following question.
• Logic: These chips are known as the “brains” of electronics as they can process information and perform calculations to execute various tasks. The main logic chip in a computer is the CPU¹, but GPUs² have grown in importance as they evolved to be applicable to more areas including AI. This is due to GPUs’ parallel processing capability, which allows them to process vast amounts of data simultaneously.
• DAO: Generally simpler than their memory and logic counterparts, DAO chips have various applications. Used for a single specific task, discrete chips are elementary devices which can function independently from a larger circuit. Meanwhile, analog chips convert analog information such as audio into binary code. Finally, the “other” category includes optoelectronic chips, which translate light into digital signals, and various sensors which are used to detect environmental changes such as heat and pressure variations.

These different types of chips will often be used together in a single device, combing to ensure the smooth operation of a system.

An overview of the main types of semiconductors based on function: memory, logic, and DAO

 

What does semiconductor memory actually do?

Semiconductor memory’s primary role is to store data in devices such as computers, smartphones, and servers. In terms of storage, semiconductor memory is divided into two main types depending on their data retention when power is lost.

• Volatile memory: Temporary storage that requires a continuous power supply to maintain its stored information. Offering rapid read/write speeds, it is used for active data and program instructions while a system is running. RAM³ is the most common volatile memory, and is further divided into DRAM⁴ and SRAM⁵.
• Non-volatile memory: Storage that permanently retains data even when power is lost. The most common type is ROM⁶, which is designed specifically for reading data. Flash, which includes NAND flash, is a type of non-volatile memory which can read and write data. Due to these capabilities, NAND flash is applied to USB drives, memory cards, and solid-state drives (SSDs).

In addition to storage, semiconductor memory is also evolving to be used for computation. Traditionally, memory chips have only assisted the CPU or GPU in computational tasks such as the performance of complex calculations. However, solutions such as PIM⁷ have emerged which has its own computational capabilities to share the workload.

Other roles of semiconductor memory include rapid data access, which is related to the process of reading and writing data to the memory cells. Some semiconductor memory products also offer error checking and correction functions to ensure data integrity and increase reliability.

In addition to storage, semiconductor memory has a variety of roles

 

What are the indicators of high performance in semiconductor memory?

What does it really mean when a semiconductor memory product is considered as “high performance”? The following are some of the key specifications which indicate the performance level of a product.

• Speed: Speed is a crucial indicator of memory performance. Read/write speed, measuring how fast memory can access and save data, respectively, is a key speed metric. Other common measures include data transfer rate, which specifies how fast information moves between the memory and other devices, and data processing speed—the rate at which stored data can be processed. Offering ultra-fast data processing speeds, SK hynix’s HBM3E⁸ is leading the memory field in terms of speed.
• Capacity & density: Generally measured in units of bytes, capacity refers to the maximum amount of data that can be stored in a device. Meanwhile, density is the amount of data that be stored in a given physical area of a storage device. For SK hynix, the company has continued to push the limits of product density, developing samples of the world’s first 321-layer NAND flash in 2023.
• Power efficiency: This refers to the effectiveness with which a memory product uses power to perform its operations. Typically measured in performance per watt⁹, power efficiency is a key consideration for semiconductor companies including SK hynix as it looks to optimize performance and enhance its sustainability. An example of the company’s power-efficient products is LPDDR5T¹⁰, the world’s fastest mobile DRAM renowned for its low-power and low-voltage characteristics.
• Reliability: This indicates the probability a memory product can perform to the required standard without failures (errors during product use) over a set period. One of the common metrics for reliability is early failure rate (EFR), which estimates the number of device failures to occur within a year in the user environment. To ensure product reliability, companies must ensure they meet industry standards and conduct various tests.

Key indicators such as speed and capacity signify memory performance

 

What are some of the major challenges in semiconductor manufacturing?

From striving to continue scaling through to dealing with supply chain disruptions, semiconductor companies face various challenges in their quest for advancement. Below is a list of some of the main issues that impact the semiconductor manufacturing process.

Continuing Scalability

As technology advances, there is a growing demand for smaller and more powerful semiconductors. However, semiconductor scaling—the process of miniaturizing semiconductor devices while improving performance—is a significant challenge due to physical and technological limitations. To continue scaling, manufacturers must continue innovation and investment in design, materials, and manufacturing.

Increasing Costs

This large-scale investment is another challenge for semiconductor companies, as equipment such as lithography machines are particularly costly. On a broader level, developing next-generation semiconductor technologies involves substantial R&D costs for developing new materials and manufacturing processes. Furthermore, fabrication facilities, or fabs, that produce ever-increasing quantities of semiconductor products are now also multi-billion-dollar investments.

Improving Sustainability

The rising production of semiconductor products contributes to another issue for manufacturers—managing their environmental impact. The industry is coming together to improve its sustainability, including cutting carbon emissions and reducing waste. However, implementing these measures while maintaining production efficiency and meeting regulatory requirements is a continuous challenge for semiconductor companies.

Supply Chain Disruptions

Global challenges such as the COVID-19 pandemic have highlighted the fragility of global supply chains. Semiconductor manufacturing relies on a complex network of suppliers for materials, equipment, and expertise. Disruptions in any part of this chain can lead to shortages and price fluctuations. To strengthen their supply chain resilience, companies are introducing measures such as diversifying their suppliers, locally sourcing materials, and enhancing inventory management.

Semiconductor companies must overcome key manufacturing challenges to continue making progress

 

What are the latest advancements and future trends in semiconductor technology?

Driven by the need for greater performance, efficiency, and scalability, semiconductors are undergoing rapid advancements and propelling the development of new technologies.

Today, AI is making headlines around the world and semiconductor memory is set to be crucial in ensuring AI’s future progress. Semiconductor companies are making high-performance chips specifically for AI and machine learning applications to enable more efficient data processing. In particular, SK hynix’s industry-leading HBM3E is suited for AI training as it can rapidly handle and access data. The company’s HBM products are set to continue fueling the evolution of AI, with the planned mass-production of its next-generation HBM4 set for 2025.

While some consider Moore’s Law¹¹ to be part of the past, SK hynix’s next-generation packaging technologies are pushing the limits of scalability. Innovative packaging methods like MR-MUF¹² have propelled the company to its leadership position in the HBM market, while emerging technologies such as chiplet¹³ and hybrid bonding¹⁴ are expected to contribute to new product development.

Semiconductor technology will also play a key role in the evolution of quantum computing, which is set to tackle problems currently beyond the capabilities of even the most powerful traditional computers. Semiconductor materials have been used in trials involving this emerging technology, enabling researchers to utilize quantum computers at room temperature. This will allow quantum computing to leap forward and bring about a potential technological revolution.

Semiconductor technology is driving advancement of various technologies including AI

 

¹Central processing unit (CPU): A hardware component which is the core computational unit in a device.
²Graphics processing unit (GPU): A computer chip that renders computer graphics and images by performing mathematical calculations.
³Random 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.
⁴Dynamic random access memory (DRAM): A type of RAM that serves as the main memory in computers. While DRAM is more cost-effective and offers greater capacity than SRAM, it needs to be periodically refreshed to maintain stored data.
⁵Static random access memory (SRAM): A type of RAM which is often used for a computer’s cache memory. Unlike DRAM, it does need to be refreshed to maintain stored data and therefore offers improved performance and lower power usage.
⁶Read-only memory (ROM): A type of computer storage containing permanent data that generally can only be read, not written to.
⁷Processing-In-Memory (PIM): A type of intelligent memory that embeds the computational functions of a processor in memory.
⁸HBM3E: The fifth-generation and latest High Bandwidth Memory (HBM) product. HBM is a high-value, high-performance product that revolutionizes data processing speeds by connecting multiple DRAM chips with through-silicon via (TSV).
⁹Performance per watt: An indicator of how much computation is performed per watt of power consumed.
¹⁰Low Power Double Data Rate 5 Turbo (LPDDR5T): Low-power DRAM for mobile devices, including smartphones and tablets, aimed at minimizing power consumption.
¹¹Moore’s Law: Proposed by Intel co-founder Gordon Moore, it states the number of transistors on a microchip doubles approximately every two years.
¹²Mass reflow-molded underfill (MR-MUF): Mass reflow is a technology that connects chips together by melting the bumps between stacked chips. Molded underfill fills the gaps between stacked chips with protective material to increase durability and heat dissipation.
¹³Chiplet: A technology that breaks up chips into functions and connects these separated pieces on a single substrate to enable heterogeneous bonding and integration.
¹⁴Hybrid bonding: A technology that connects chips together directly without bumps to enable higher bandwidth and capacity.

 

The next episode will explore “when” semiconductors and SK hynix started to change peoples’ lives.

 

<Other articles from this series>

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

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