Knowledge Center
Navigation
Knowledge Center

Processors

An integrated circuit or part of an IC that does logic and math processing.
popularity

Description

Processor is a generic term referring to logic circuitry on an integrated circuit that processes instructions and data in a computer. The processor takes in instructions and data and performs operations on the data as defined by the instructions, calculating an output that is used to either control the computer or device or complete a specific task. In the early days of computing, the processor was made up of vacuum tubes, relays and room-sized equipment, followed by transistors on circuit boards and early integrated circuits. When the microprocessor—an integrated circuit (IC) on a silicon substrate, colloquially called a chip—came into being in the 1970s, the processor on a dedicated IC was called a microprocessor. A processor can also be in the form of an intellectual property (IP) core, added to an IC.

Also often in the past, thanks to the computer industry, the word “processor” was used interchangeably with “CPU,” or “central processing unit,” but many types of processors have developed over the years for specific purposes. Central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), data processing units (DPUs), tensor processing units (TPUs) are all different processor types.

A processor is a dedicated IC or is a part of an IC that has an arithmetic logic unit (ALU) to run logic and math operations. Some processors may have a math co-processor, which runs floating point math. ALUs do four basic math operations: add, subtract, multiply and divide. The logic operations have three conditions: Equal-to, less-than, or greater than. The conditions can also include less than or equal to, greater than or equal to, and not equal. The logic unit basically compares the numbers. (More info: “How Computers Work: The CPU and Memory.”) A control unit tells the parts of the processor what to do with the data. The set of control instructions and what they do defines the instruction set architecture (ISA) for the processor. A set of those instructions that performs a defined function is called a program, or often referred to as software.

Microcontrollers are processors that include memory and some peripherals integrated into the chip alongside the CPU, whereas a microprocessor only has a CPU or other basic processor architecture. The microcontroller has everything needed for embedded systems design. Sometimes the microcontroller is also considered custom designed for a purpose, whereas the microprocessor is for a generic purpose.

The relationship between the processor and memory generally conforms to one of two architectures: the Von Neumann and Harvard architectures. With the von Neumann architecture, instructions and data are mixed within a single contiguous memory space. The Harvard architecture has different memory spaces for instructions and data.

In recent times, especially when a chip contains multiple processors that share a single contiguous memory space, the performance of the processors becomes limited by the bandwidth of the connection to memory. This makes it advantageous to bring as much memory as possible onto the chip where latency is minimized. This however reduces the space available for processors. Many techniques, such as memory caching, have been used to try and make it appear as if more memory is closer to the processors, and new fabrication techniques, such as 2.5D ICs, are being developed that enable memory to be contained within the package of a chip even if not on the same piece of silicon as the processor itself. The on-chip versus off-chip debate continues with AI and machine learning becoming more prevalent, especially separating the memory from the processing. Now with AI and machine learning chips, however, the processing units (CPUs, GPUs, etc.) can be a block on a larger chip, mixed in with memory, networks on chips, and many multiply/accumulate blocks.

Custom processors
The custom processor is now in a renaissance. A generation ago, every major semiconductor company had in-house processors: SuperH, PowerPC, V800, Alpha, MEP, Trimedia, etc., with some specializing more than others for particular domains. Industry consolidation and the enormous expense of maintaining proprietary architectures caused many of these to fade away, and the industry entered a long period of ‘Standard Architecture,’ while custom processors filled niches for applications like audio processing. Now, spurred by the latest cyclical development boom, the semiconductor industry is entering a new golden era of custom processors, but this time ‘custom processor’ means something different.

A custom processor has grown to mean a processor optimized for a particular class of tasks. The microarchitecture and instruction set are optimized for the software that will ultimately run on it. Today, demand is high for customization tools that can realize proprietary instruction extensions of a standard ISA. A custom processor also used to mean a CPU designed from scratch, but the definition has evolved and expanded over time. (Read more here.)