What is computer architecture?

Computers are a key part of our everyday lives, from  the machines we use for work to the smartphones and smartwatches we rely on. 

All computers, no matter their size, are based around a set of rules stating how software and hardware join together and interact to make them work. This is what is known as computer architecture. In this article we’re going to delve into what computer architecture actually is.

Computer architecture is the organisation of the components which make up a computer system and the meaning of the operations which guide its function. It defines what is seen on the machine interface, which is targeted by programming languages and their compilers. 

The three categories of computer architecture

There are three categories of computer architecture, and all work together to make a machine function.

System design

System design includes all hardware parts of a computer, including data processors, multiprocessors, memory controllers, and direct memory access. It also includes the graphics processing unit (GPU). This part is the physical computer system.

Instruction set architecture (ISA)

This includes the functions and capabilities of the central processing unit (CPU). It is the embedded programming language and defines what programming it can perform or process. This part is the software that makes the computer run, such as operating systems like Windows on a PC or iOS on an Apple iPhone, and includes data formats and the programmed instruction set.

Microarchitecture

Microarchitecture is also known as computer organisation and defines the data processing and storage element and how they should be implemented into the ISA. It is the hardware implementation of how an ISA is implemented in a particular processor.

The evolution of processors

Complex Instruction Set Computer (CISC) and Reduced Instruction Set Computer (RISC) are the two major approaches to processor architecture. 

CISC processors have a single processing unit, external memory, and a small register set with hundreds of different instructions. These processors have a single instruction to perform a task, and have the advantage of making the job of the programmer easier, as fewer lines of code are needed to get the job done. This approach uses less memory, but can take longer to complete instructions. 

The RISC architecture was the result of a rethink, which has led to the development of high-performance processors. The hardware is kept as simple and fast as possible, and complex instructions can be performed with simpler instructions.

Microprocessors are digital systems which read and execute machine language instructions. Instructions are represented in a symbolic format called an assembly language. These are processors which are implemented on a single, integrated circuit. Common microprocessors used today are the Intel Pentium series, IBM PowerPC, and the Sun SPARC, among others. Nearly all modern processors are microprocessors, which are often available as standard on von Neumann machines.

The von Neumann architecture

Mathematician John von Neumann and his colleagues proposed the von Neumann architecture in 1945, which stated that a computer consists of: a processor with an arithmetic and logic unit (ALU) and a control unit; a memory unit that can communicate directly with the processor using connections called buses; connections for input/output devices; and a secondary storage for saving and backing up data.

The central computation concept of this architecture is that instructions and data are both loaded into the same memory unit, which is the main memory of the computer and consists of a set of addressable locations. The processor can then access the instructions and data required for the execution of a computer program using dedicated connections called buses – an address bus which is used to identify the addressed location and a data bus which is used to transfer the contents to and from a location.

The pros and cons of the von Neumann architecture

Computers as physical objects have changed dramatically in the 76 years since the von Neumann architecture was proposed.  Supercomputers in the 1940s took up a whole room but had very basic functionality, compared to a modern smartwatch which is small in size but has dramatically higher performance. However, at their core, computers have changed very little and almost all of those created between then and now have been run on virtually the same von Neumann architecture.

There are a number of reasons why the von Neumann architecture has proven to be so successful. It is relatively easy to implement in hardware, and von Neumann machines are deterministic and introspectable. They can be described mathematically and every step of their computing process is understood. You can also rely on them to always generate the same output on one set of inputs.

The biggest challenge with von Neumann machines is that they can be difficult to code. This has led to the growth of computer programming, which takes real-world problems and explains them to von Neumann machines. 

When a software program is being written, an algorithm is reduced to the formal instructions that a von Neumann machine can follow. However, the challenge is that not all algorithms and problems are easy to reduce, leaving unsolved problems.

Harvard architecture

Another popular computer architecture, though less so than the von Neumann architecture, is Harvard architecture. 

The Harvard architecture keeps instructions and data in separate memories, and the processor accesses these memories using separate buses. The processor is connected to the ‘instructions memory’ using a dedicated set of address and data buses, and is connected to the ‘data memory’ using a different set of address and data buses. 

This architecture is used extensively in embedded computing systems such as digital signal processing (DSP) systems, and many microcontroller devices use a Harvard-like architecture. 

Take your learning on computer architecture to the next level

At the University of Sunderland we offer a 100% online learning MSc Computer Science course, designed for individuals who aren’t from a computer science background and want to change their career path, or for those who want to incorporate computer science knowledge into their current field for career progression.

Study part-time and entirely online and grow your global network as you connect with peers from all over the world. Apply today and start within weeks, we have six start dates a year.