Multipurpose Machines: Early computers able to calculate output using fixed instructions. 1940s=
Neumann and Turing proposed stored program concept.
Stored Program Concept: Program loaded to main memory to be executed by processor, instructions
fetched one at time, decoded and executed sequentially by processor. Sequence of instructions can only
be changed by condition/unconditional jump instructions. One version of this concept is Von Neumann
Architecture.
John von Neumann and Von Neumann Architecture: most common of this concept is Von Neumann
architecture. Instructions and data stored in main memory and transferred using one shared bus. Almost
all general purpose computers based on these principles. Simplifies CU design. Data from memory and
devices accessed in same way.
Harvard Architecture: Alternative model, separates data and instructions into separate memories with
different buses. Program instructions and data no longer compete for the same bus. Varying memory
and word lengths can be used for data and instructions. Used with specialist embedded systems and
Digital Signal Processing where speed takes priority over complexities of design.
Contemporary processor architectures: Modern CPU chips use aspects of both of the above
architectures. In desktop computers, one main memory for data and instructions but cache memory is
divided into instruction cache and data cache; data and instructions are retrieved using Harvard
architecture.
Complex Instruction Set Computers (CISC): Large instruction set used to accomplish tasks in as few lines
of assembly language as possible. Combines load/store instruction with instruction that carries out
actual calculation. Quicker to code, compiler has little work to translate high-level into machine code,
code relatively short so little RAM required to store instructions. WAY TO REMEMBER: Complex will have
more available options to code in so will be able to do the same code but in lesser lines as the code will
be more complex.
Reduced Instruction Set Computers (RISC): Minimum number of simple instructions, each taking one
clock cycle, accomplish all required operations in multiple Gen Purpose Registers. Hardware simpler to
build needing fewer circuits for carrying out complex instructions. Each instruction takes same amount of
time. RAM is cheaper, RISC use of RAM and software allows better performance processors at less cost.
Multi-core and parallel systems: Multi-Core able to distribute workload across many cores, achieving
high performance due to parallel processing(only happens if software written to allow it to happen).
They are therefore called: parallel systems. Many PCs and mobile devices are multi-core, meaning they
have many cores. Supercomputers= thousands of cores.
Co-Processor systems: Extra processor to supplement functions of the CPU (AKA Primary processor).
GPU: Graphics Processing Unit. Efficient at manipulating computer graphics and image processing.
Thousands of small cores designed for parallel processing, process large blocks of visual data
simultaneously. Can work with CPU to accelerate scientific, engineering and other applications. Used
numerous devices such as: phones and tablets, cars, drones and robots.
Neumann and Turing proposed stored program concept.
Stored Program Concept: Program loaded to main memory to be executed by processor, instructions
fetched one at time, decoded and executed sequentially by processor. Sequence of instructions can only
be changed by condition/unconditional jump instructions. One version of this concept is Von Neumann
Architecture.
John von Neumann and Von Neumann Architecture: most common of this concept is Von Neumann
architecture. Instructions and data stored in main memory and transferred using one shared bus. Almost
all general purpose computers based on these principles. Simplifies CU design. Data from memory and
devices accessed in same way.
Harvard Architecture: Alternative model, separates data and instructions into separate memories with
different buses. Program instructions and data no longer compete for the same bus. Varying memory
and word lengths can be used for data and instructions. Used with specialist embedded systems and
Digital Signal Processing where speed takes priority over complexities of design.
Contemporary processor architectures: Modern CPU chips use aspects of both of the above
architectures. In desktop computers, one main memory for data and instructions but cache memory is
divided into instruction cache and data cache; data and instructions are retrieved using Harvard
architecture.
Complex Instruction Set Computers (CISC): Large instruction set used to accomplish tasks in as few lines
of assembly language as possible. Combines load/store instruction with instruction that carries out
actual calculation. Quicker to code, compiler has little work to translate high-level into machine code,
code relatively short so little RAM required to store instructions. WAY TO REMEMBER: Complex will have
more available options to code in so will be able to do the same code but in lesser lines as the code will
be more complex.
Reduced Instruction Set Computers (RISC): Minimum number of simple instructions, each taking one
clock cycle, accomplish all required operations in multiple Gen Purpose Registers. Hardware simpler to
build needing fewer circuits for carrying out complex instructions. Each instruction takes same amount of
time. RAM is cheaper, RISC use of RAM and software allows better performance processors at less cost.
Multi-core and parallel systems: Multi-Core able to distribute workload across many cores, achieving
high performance due to parallel processing(only happens if software written to allow it to happen).
They are therefore called: parallel systems. Many PCs and mobile devices are multi-core, meaning they
have many cores. Supercomputers= thousands of cores.
Co-Processor systems: Extra processor to supplement functions of the CPU (AKA Primary processor).
GPU: Graphics Processing Unit. Efficient at manipulating computer graphics and image processing.
Thousands of small cores designed for parallel processing, process large blocks of visual data
simultaneously. Can work with CPU to accelerate scientific, engineering and other applications. Used
numerous devices such as: phones and tablets, cars, drones and robots.
