What is a Computer Chip?

A computer chip is a tiny piece of semiconducting material with an embedded electronic circuit. This chip contains millions of microscopic electronic components called transistors that control electrical signals. Initially, computers were large and room-sized, connecting transistors and other components by wiring, and computing was only something done in national labs, universities, or large companies. Breakthrough development reduced a computer's main circuitry to a small chip, or integrated circuit board. As technology advanced, it became possible to create smaller computers that could be used at home. Further innovation in computer chip technology resulted in the development of high-performance processors that power all types of advanced analytics, graphics, and machine learning applications. Today, the computer chip is omnipresent in modern technology—in everything from our microwaves to our toothbrushes.

Computer chips are typically manufactured in factories called fabrication plants or semiconductor fabs. Here's a summary of the computer chip manufacturing process and the discrete components that combine to form a single chip.

Electronic properties and chemical building blocks

Chips are made from silicon, a common chemical element found in sand. Silicon is a semiconductor material, which means its electrical conductivity falls somewhere between metals like copper and insulators like glass.

Silicon extraction and shaping

Single-crystal silicon ingots are extracted by melting and refining sand. The ingots are nearly 100% pure. They are cut into wafer-thin chips that are cleaned, polished, and coated with a layer of silicon dioxide. An additional coating of a chemical known as photoresist is added on top of the silicon wafers to increase photosensitivity. Strict precautions are taken to ensure there is no contamination of dust or other foreign substances during this process. Once the basic silicon wafer chips are ready, electronic circuits are etched onto them.

Circuit etching

The silicon wafer is covered by a circuit-patterned plate, called a mask, and exposed to ultraviolet light. The light hardens the exposed photoresist material in the circuit pattern. Hot gases then melt the exposed material to reveal the silicon dioxide beneath. A 3D landscape is left behind, replicating the mask's circuit design pattern.

In the computer chip manufacturing process, etching is the chemical removal of layers from the wafer, and doping is the introduction of impurities to modify the wafer. The process of etching and doping may be repeated hundreds of times on a single chip to create more complex integrated circuitry. Multiple chips are cut from a single wafer.

Computer chips work by transmitting electrical signals through the circuit elements.

Analog integrated circuits

Analog circuits transmit continuous, varying signals in a given time period. The output signal is a linear function of the input with directly proportional voltages. This type of integrated circuit is used for electronic devices' functionality like timers, comparators, voltage regulators, and operational amplifiers. Analog chips are used in sweep generators, oscillators, audio amplifiers, and filters.

Digital integrated circuits

Digital circuits transmit discontinuous or binary signals. The output voltage may be high or low. The high voltage represents the boolean value 1, and the low voltage represents the boolean value 0.

Digital circuits are designed to perform different logical operations like AND, OR, and NAND. For example, logic gate OR operations correspond to Boolean addition and are the basis of computer addition operations.

Digital integrated electronic circuits form the basis of all computing operations. They are crucial for all programmable devices, logic boards, microcontrollers, and memory.

Mixed-signal integrated circuits

Mixed chips combine elements from both analog and digital chips. Chip designers use this hybrid approach to make it possible to have chips that act as digital-to-analog and analog-to-digital converters. These advanced integrated circuits are another core component of modern computing.

Quantum circuits

Quantum circuits are the next stage of computing evolution. A quantum circuit is a computing routine that defines a series of logical quantum operations on underlying qubits, or quantum bits. Qubits are represented by quantum particles, which are different from Boolean digital signals. Boolean signals are 1 or 0, but qubits can be placed in a superposition of states. Chips with quantum circuits form the basis of quantum computing, which is an emerging technology.

Energy efficiency and computing power

Modern computer chips combine computing power with considerations for power consumption. With more complex integrated circuitry, circuit designers manage how signals move through the semiconductor material to ensure reliable data processing. The electrical properties of these materials directly impact the signal speed and efficiency of a chip.

Advances in technology mean that it is now possible to fabricate billions of transistors on a single integrated circuit, continuously refining the thin wafers. The improvement gains result in better performance per watt, which can make a significant difference in large-scale environments, such as data centers. A data center may have millions of computer chip devices operating simultaneously, so advances in chip performance can result in major reductions in power consumption within the center.

Computer chips are classified into four broad categories according to functionality.

Memory computer chips

Memory chips store programs and data on computers and storage devices. Random Access Memory (RAM) chips provide temporary storage, whereas flash drives and solid state drives (SSDs) can hold information permanently. Flash memory units can store data even when the electrical current is switched off.

Logic chips

Logic or processor chips process data to complete tasks. They are the brains of modern electronic devices. A Central Processing Unit (CPU) is the main type of logic chip found in the microprocessors of servers and other computing hardware. However, logic chips can also be designed for specific functions. Here are some examples:

• Graphical processing units (GPUs) are designed to optimize visual displays or perform machine learning tasks
• Neural processing units are modern chips designed for deep learning and machine learning applications

ASICs

Application-specific integrated circuits (ASICs) are designed to perform repetitive processing routines for a specific application. These chips are produced in batches for single-purpose appliances, such as consumer electronic devices. Another example is bitcoin mining, where ASICs perform the complex mathematical routines required for producing new bitcoins.

SoCs

The system on a chip (SoC) is a newer type of computer chip. All electronic components needed for an entire system are built into one chip. SoC capabilities are more extensive than microcontrollers. A microcontroller generally combines the Central Processing Unit with memory and I/O processing. However, the SoC may integrate graphics, audio, camera, and video processing.

Amazon Web Services (AWS) has invested decades in designing custom computer chips optimized for the cloud. As a result, Amazon EC2 offers instances powered by AWS-designed processors that are optimized for a range of compute, memory, and storage-intensive workloads. Amazon EC2 also offers purpose-built machine learning (ML) chips for ML training and inference.

Here are some other AWS innovations:

• The AWS Nitro System is the foundation for our next generation of EC2 instances that enables AWS to innovate faster, further reduce cost for our customers, and deliver added benefits like increased security and new instance types.
• AWS Graviton processors are a family of processors designed to deliver the best price performance for your cloud workloads running in Amazon EC2.
• AWS Inferentia chips are designed by AWS to deliver high performance at the lowest cost in Amazon EC2 for your deep learning (DL) and generative AI inference applications.
• AWS Trainium is a family of purpose-built AI accelerators—Trainium1, Trainium2, and Trainium3—designed to deliver scalable performance and cost efficiency for training and inference across a broad range of generative AI workloads.

Get started with AWS instances powered by custom computer chips by creating a free account today.