Published: July 24, 2026
This is the second installment in a special edition of our series, “Miniaturization and Horizontal Specialization in the Semiconductor Industry.”
In this post, I take a closer look at advanced logic semiconductors. First, let’s start with the term “processor,” which is used in the widest range of applications.
What Is a Processor?
Types of Processors: CPUs and Microprocessors
What is an MCU?
What Is SoC?
Volume 20: Advanced Logic Semiconductors: CPUs, GPUs, MPUs, MCUs, and SoCs
According to English dictionaries, the term “processor” can be defined, for example, as “a person, device, or business that processes things.”
In the context of information technology (IT), a processor refers to hardware or software that processes information according to a set of instructions. For example, I created this document using Microsoft Word, which is a type of word processing software. A microprocessor, a type of advanced logic semiconductor, is also a type of processor. It performs logical operations on input data according to a program.
PUs, such as central processing units (CPUs), graphics processing units (GPUs), and microprocessor units (MPUs), which are classified as advanced logic semiconductors, are abbreviations for processing units.
The “PU” in terms such as CPUs, MPUs, and GPUs, which are advanced logic semiconductors, stands for “processing unit.” In an IT context, “processing unit” is essentially synonymous with “processor.” Therefore, xPUs—such as CPUs, GPUs, and MPUs—can all be regarded as types of processors. In fact, MPU is an abbreviation for microprocessors.
I will take a closer look at the different types of processors. I begin with the CPU, which is one of the most clearly defined terms.
CPU
A CPU is one of the components of a computer. It is a term from computing rather than semiconductor technology, and its definition is well established. Modern computers typically have the structure shown in the figure below. A CPU is a part of this structure that performs control and arithmetic/logic operations. It serves as a brain of the computer. Unlike CPUs in modern PCs, a CPU does not have to be integrated into a single semiconductor chip, nor does it have to be made of semiconductors. Before the invention of integrated circuits, CPUs were built from discrete components such as individual transistors. Earlier systems used vacuum tubes.
Computer configuration
Microprocessor
Another term commonly used in computing is “microprocessor.” Unlike “processor” or “CPU,” the term “microprocessor” does not have a clearly established definition.
Today, the most widely accepted definition of a microprocessor is a device that integrates the functions of a computer’s CPU—namely control and arithmetic/logic operations—onto a single semiconductor chip. This is a semiconductor-related term and has a more specific meaning than “CPU.”
The Intel 4004, introduced in 1971, is widely regarded as the world’s first microprocessor. It was originally developed as a custom chip for calculators for the Japanese company Busicom. For further details, see Supplementary Information 1 at the end of this article.
Note: In a news release dated November 15, 2021 (U.S. time), marking the 50th anniversary of the Intel® 4004 microprocessor, Intel described it as “the world’s first commercially available microprocessor.”
Although the term “microprocessor” had been used earlier with a different sense, the Intel 4004 can be considered the starting point of the modern concept of a microprocessor.
According to Masatoshi Shima, an engineer at Busicom who was involved in its development, the term “microprocessor” was coined not because the device was physically small, but because it employed “microinstructions” rather than “macroinstructions.” Today, however, the term is commonly understood not in its original sense, but simply to mean something physically small.
The term MPU is often used as an abbreviation for a microprocessor. It stands for Microprocessor Unit (or Microprocessing Unit).
Relationship Between Processors, CPUs, and Microprocessors
Based on the discussion above, the relationship between processors, CPUs, and microprocessors can be summarized in the figure above. A CPU is a functional component of a computer. A microprocessor, on the other hand, refers to a CPU implemented on a single semiconductor chip, and is therefore a semiconductor device term. Both are, of course, processors.
Therefore, semiconductor devices that serve as the “brain” of personal computers, such as those from Intel and AMD—may be referred to as CPUs, microprocessors, or simply processors. All of these terms are correct.
As shown in the figure above, the conceptual relationship is as described earlier; however, in practice, all CPUs used in modern IT systems—such as personal computers, servers, and smartphones—are implemented as semiconductor chips. As discussed in the previous section, a CPU implemented on a semiconductor chip is referred to as a microprocessor. Therefore, when considering only modern IT systems, “CPU” and “microprocessor” can be considered the same in practice, even though their meanings differ conceptually. In other words, all CPUs are microprocessors.
When Intel first developed microprocessors, a microprocessor was generally considered to have been equivalent to a CPU. Today, however, increased integration density has made it possible for semiconductor chips to integrate various circuits in addition to the CPU. As a result, a microprocessor is no longer simply a CPU, but rather a CPU together with other related circuits.
Originally, the term “CPU” referred to the central processing unit—the “brain” of a computer—as a component of the computer. However, when “CPU” is used today in the context of IT equipment or semiconductors, it is often used to refer to circuitry that implements the CPU’s functions—control and arithmetic/logic operations—on a semiconductor chip.
Changes in the relationship between CPUs and microprocessors
In addition to CPUs, there are now other processors such as GPUs and NPUs—often collectively referred to as xPUs—implemented on semiconductor chips. For this reason, they can, in a broad sense, be considered microprocessors, and are in fact sometimes referred to as such. Recent products may even integrate CPUs, GPUs, and NPUs on a single chip.
Note: GPUs are discussed in the next section. NPU stands for Neural Processing Unit, also referred to as Neural Network Processing Unit, and is a processor specialized in AI processing. “Neural network” refers to biological neural networks.
Note: According to Intel’s website, as accessed on May 13, 2025, CPUs and GPUs are described as “both are silicon-based microprocessors.”
Furthermore, since the xPUs (e.g. CPUs, GPUs and NPUs) used in modern IT systems are implemented as semiconductor chips, they are all microprocessors. Consequently, there is little need to explicitly use the term “micro” in the term, and they are often simply referred to as processors.
In fact, both Intel and AMD use the term “Processors” for their product families on their websites (accessed May 13, 2025).
Therefore, in the context of modern IT systems, the relationship among processors, xPUs (e.g., CPUs, GPUs, NPUs), and microprocessors can be roughly illustrated in the figure below.
Relationship between processors, CPUs, and microprocessors in modern IT systems
Next, let’s look at the GPU, which is probably the most well-known processor besides the CPU. Since the announcement of ChatGPT in 2023, the term has suddenly become much more visible and frequently heard, especially in AI-related contexts.
A GPU is a processor specialized in rendering images for display on computers, game consoles, and similar systems. Although dedicated graphics processing chips have existed since the 1970s, NVIDIA is credited with coining the term “GPU” when it first used the term for its own products in 1999. It has now become a widely used term. Thanks to their ability to perform large numbers of relatively simple computations in parallel, GPUs are now widely used in other fields, including AI workloads. Compared with CPUs, GPUs have a more limited range of applications.
Note: The use of GPUs for purposes beyond image processing, such as AI, by leveraging their capabilities is called GPGPU (General-Purpose Computing on Graphics Processing Units).
While a CPU can substitute for a GPU, albeit more slowly, a GPU can handle only part of the CPU’s workload and cannot serve as a replacement for the CPU. In practice, there are no personal computers without a CPU. The CPU functions as the brain and command center, whereas the GPU can be thought of as a large team of workers carrying out large volumes of simple, highly parallel tasks under the CPU’s direction. Alternatively, the CPU can be seen as outsourcing certain computationally intensive but simple tasks to the GPU.
AI-generated image of GPU racks in a data center
As miniaturization progresses and integration increases, it becomes possible to integrate nearly all of a system’s functions on a single semiconductor chip. Next, I will look at products that integrate various functions in addition to the processor.
A term closely related to MPU is MCU, which is also commonly used.
MCU stands for Microcontroller Unit, and the term is commonly used as “microcontroller” or simply “MCU.” An MCU incorporates a CPU, and as the term “controller” suggests, it is mainly used to control home appliances, industrial equipment, automobiles, and similar systems.
Note: Examples of MCU-based control include automatic bath filling, automatic rice cooking in rice cookers, automatic temperature and humidity control in air conditioners, and automatic temperature control in refrigerators.
In most cases, an MCU integrates the key functions required for control—such as a CPU, memory, and input/output (I/O) circuits—on a single chip. Because their primary role is control, high performance is generally not required, and they typically do not require the latest process nodes. Many products can be manufactured using so-called legacy processes. MCUs commonly include on‑chip memory (e.g., flash memory) for program storage, and they often incorporate analog circuits, so they are not purely logic semiconductors. Nevertheless, in industry classifications they are generally considered to be classified as logic semiconductors.
The first MCU appeared in the early 1970s, roughly at the same time as microprocessors. Calculators were also a key factor in their development. They were initially referred to as “one-chip microcomputers” or “single-chip microcomputers,” but over time the term appears to have gradually shifted to “microcontroller.” For further details on the historical background, see Supplementary Information 2 at the end of this article.
According to the websites of major MCU manufacturers such as Renesas, NXP, and STMicroelectronics (accessed May 13, 2025), they offer not only MCUs but also MPUs (or microprocessors, sometimes simply called processors). These are not the types of processors used in PCs, servers, or smartphones; rather, like MCUs, they are used in a wide variety of equipment. Nevertheless, MCUs and MPUs are treated as distinct product categories. While the boundary between MCUs and MPUs is not clearly defined, MPUs are generally considered to be used not only for simple control but also for higher-performance, more IT-oriented applications.
Finally, let’s look at SoCs—a term commonly encountered in smartphones and similar electronics. This represents a different kind of classification from those discussed earlier. SoC stands for System on a Chip (sometimes written as System on Chip) and is typically pronounced “S O C.” It is a semiconductor device that integrates logic circuits such as CPUs and other major system functions onto a single chip. Because SoCs often include a CPU, they are sometimes referred to as processors. However, “processor” refers to a function, whereas “SoC” refers to a type of integration rather than a function.
Since SoCs often include analog circuits and memory in addition to logic circuits, it may seem somewhat counterintuitive to classify them as logic semiconductors. However, since their primary function is logic processing, they can reasonably be categorized as logic semiconductors.
For compact systems such as smartphones, reducing the component count is a major advantage. SoCs can be considered a highly effective solution. Conversely, one could argue that smartphones were made possible precisely because of the existence of SoCs.
This raises the question: when did the term “SoC” come into use? It is difficult to pinpoint an exact starting point, but it is estimated to have emerged in the 1990s. A similar term, “system LSI,” was once widely used, but today “SoC” has become the dominant term, and “system LSI” is now rarely used. As miniaturization progressed and the number of elements on a chip increased exponentially under Moore’s Law, it became possible by the 1990s to integrate a system’s major functions on a single chip, which likely led naturally to the emergence of the SoC concept. That said, products known in the 1970s as “one‑chip microcomputers” (also called “single‑chip microcomputers,” essentially today’s MCUs) had already appeared. As a result, small-scale systems could, in effect, already be regarded as SoCs even at that time. However, MCUs are rarely referred to as SoCs; the term “SoC” is generally used for devices with much higher levels of integration and performance.
As mentioned above, an SoC refers to a semiconductor device that integrates logic circuits such as CPUs and other major system functions on a single chip. There does not appear to be a clear definition of how much functionality must be included for a device to be considered an SoC. In practice, as integration has advanced, many processor-related products now integrate multiple functions beyond the CPU on a single chip, and many of them can effectively be regarded as SoC-like devices. This tendency is particularly pronounced in smaller systems that incorporate such processor-related products. Products that incorporate components such as DRAM, which are difficult to integrate on the same chip, within the same package are also common. When multiple chips are contained within a single package, the system is integrated at the package level rather than on a single chip; this is referred to as an SiP (system in package, occasionally written “system in a package”), pronounced “S I P,” rather than an SoC.
Note: For the pronunciation of SoC and SiP, see Supplementary Information 3 at the end of this article.
SoCs typically include analog circuits and memory in addition to logic circuits. However, the manufacturing processes for logic, analog, and memory are not necessarily the same. For example, in the case of memory, SRAM can be fabricated using a logic process, whereas flash memory and DRAM require distinct process flows. Consequently, even if they are integrated on the same chip, the overall manufacturing process becomes highly complex, which increases lead time and cost.
Furthermore, as chip size inevitably increases, yield (the percentage of functional chips) tends to decrease. Therefore, integrating all functions into a single chip is not always a rational choice. In some cases, it is more reasonable to partition the system into multiple chips and place them in the same package—that is, an SiP (as mentioned earlier). Recently, the chiplet concept has been attracting attention in advanced logic semiconductors. From this perspective, chiplets are a rational solution. In other words, it suggests that the SoC concept itself may be becoming less viable.
I’ll wrap up here for now. In the next post, I’ll take a closer look at chiplets.
Supplementary Information
1. The Intel 4004 was originally developed as a proprietary product for Busicom, at the request of Busicom. Busicom initially requested a dedicated set of calculator ICs that would allow different calculator models to be created simply by changing the software (program). However, for a variety of reasons, Intel instead proposed a more general four-chip product family, which could likewise be customized in software. One of these chips was a microprocessor that integrated the CPU onto a single semiconductor chip. Together, the four chips functioned as a complete computer system and were called the “MCS-4 Micro Computer Set.” Although it was originally developed as a custom product for Busicom, Intel later sold it as a general-purpose product. Masatoshi Shima of Busicom, the original client, was responsible for logic design.
2. Texas Instruments (TI) is widely credited with developing the first microcontroller unit (MCU). The TMS0100 series, released in 1971, is sometimes regarded as the first. However, others consider the TMS1000 series, released in 1974, to be the first, since the TMS0100 was intended specifically for calculators, whereas the TMS1000 was more general-purpose. A datasheet available online refers to the device as a “one-chip microcomputer.”
As mentioned earlier, Intel's first microprocessor was also referred to as part of the four‑chip MCS‑4 microcomputer set. At that time, the term “microcomputer” may have been the more commonly used expression. Intel’s first microcontroller, the MCS‑48 (8048 series), was released in 1976. User manuals available online from that period describe it as a “single-chip microcomputer.” This indicates that the term “microcomputer” was also in use in this context.
Intel’s successor product, the MCS-51 (8051 series), released in 1980, is known to have been a widely adopted microcontroller, with many compatible variants. A review of datasheets available online indicates that devices initially referred to as “microcomputers” gradually came to be called “microcontrollers” by the 1990s. As the datasheet describes these devices as “optimized for control applications,” the terminology likely evolved to reflect their intended use, although the background remains unclear.
3. SoC and SiP. According to the glossary at the end of the 2022 Edition Executive Summary of the International Roadmap for Devices and Systems (IRDS), introduced earlier in Volume 3, SoC stands for System on Chip and SiP stands for System in Package. Neither term includes the article “a.” The glossary does not provide pronunciations. To confirm how they are actually pronounced by English speakers, I listened to several English-language videos on YouTube, and found that SoC is pronounced “S O C,” and SiP “S I P.” Those pronunciations are therefore used in the main text.
Click below to read this series.
Semiconductor Miniaturization:
Volume 1: Semiconductor Miniaturization: What is Moore’s Law?
Volume 2: Semiconductor Miniaturization and Manufacturing Process
Volume 3: Semiconductor Miniaturization and International Technology Roadmap
Volume 4: Semiconductor Miniaturization and Semiconductor Business
Volume 5: Semiconductor Miniaturization and Semiconductor Business (Part 2)
Volume 6: Semiconductor Miniaturization and Semiconductor Devices
Volume 7: Semiconductor Miniaturization: What is MOSFET Scaling?
Volume 8: Semiconductor Miniaturization: Limitations of MOSFET Scaling
Volume 9: Semiconductor Miniaturization and Analog Circuits
Shift to Larger Diameter Silicon Wafers:
Volume 10: Shift to Larger Diameter Silicon Wafers: How a Common Material, Silicon, Became a Main Player
Volume 11: Shift to Larger Diameter Silicon Wafers (Part 2): How Silicon Wafers Are Made
Volume 12: Shift to Larger Diameter Silicon Wafers (Part 3): Reasons and History
Horizontal Specialization in the Semiconductor Industry
Volume 13: Horizontal Specialization in the Semiconductor Industry and the Rise of Fabless Companies
Volume 14: History of Horizontal Specialization in the Semiconductor Industry–Emergence of Fabless Semiconductor Companies
Volume 15: History of Horizontal Specialization in the Semiconductor Industry–Emergence of Foundries
Volume 16: Horizontal Specialization in the Semiconductor Industry—Is a Foundry Just a Subcontractor?
Volume 17: Horizontal Specialization in the Semiconductor Industry—Horizontal Specialization by Product and Country (Part 1)
Volume 18: Horizontal Specialization in the Semiconductor Industry—Horizontal Specialization by Product and Country (Part 2: By Country)
Volume 19: Miniaturization and Horizontal Specialization in the Semiconductor Industry (Special): Roots, Definitions, and Types of Advanced Logic Semiconductors
Volume 20: Advanced Logic Semiconductors: CPUs, GPUs, MPUs, MCUs, and SoCs