Unit 1.1.2 Types of Processor
RISC and CISC Architectures
Reduced Instruction Set Computing (RISC) and Complex Instruction Set Computing (CISC) are different design strategies for CPUs.
The instructions in RISC are simpler such as LOAD and STORE. This results in higher performance when combined with a microprocessor architecture capable of executing the instructions using fewer cycles per instruction. These instructions perform simple tasks, as a result of this more complex actions require the combination of numerous instructions. This means that a more complex task may take many more FDE cycles to complete. Its simple design means it requires less power and generates less heat, and as such is used in portable devices however the disadvantage to RISC is that it can take many clock cycles to carry out a complex task that could easily be done by a cisc processor.
CISC is a design where a more complex instruction set is permitted. The design is such that single instructions can execute several low level operations (e.g. load from memory, store to memory or an arithmetic operation), or are capable of multi step operations or addressing modes within single instructions. These instructions may take multiple FDE cycles and may have a variable format.The advantage to this is that instructions will take up significantly less space in memory as few are required to carry complex tasks. Assembly code for Cisc can also take less time to write for a programmer.
However, the disadvantage of this more complex design is that it results in complicated hardware to decode the larger instruction set. As a consequence, CISC architecture tends to generate more heat due to extra power requirements, making it less portable. They also tend to be more expensive, and can cause programs to run slower due to their extensive circuitry. However, they make the life of a programmer easier as fewer instructions have to be written to accomplish the same task.
RISC vs CISC
|Emphasis on Software (ARM)||Emphasis on Hardware (Intel/AMD)|
|All instructions take a single cycle||Includes more complex, multi-cycle instructions|
|Register to Register "LOAD" and "STORE" are independent instructions||Memory to Memory "LOAD" and "STORE" are incorporated into instructions|
|Low cycles per second, larger code sizes||Higher cycles per second, smaller code sizes|
|More transistors for memory registers||Transistors used to store complex instructions|
|lower energy consumption,can "Sleep"||Greater energy consumption|
|Physically smaller - less silicon, less expensive||Physically larger so more silicon thus expensive|
|Used in smartphones, tablets and some embedded systems such as digital signal processing||Used in laptops and desktop computers|
Total: 4 marksMaximum of 4 marks, 1 for each of the following points. Maximum of 2 marks for RISC
and 2 marks for CISC
• RISC has a smaller/limited range of instructions
• RISC requires less power/generates less heat to run.
• RISC is generally cheaper to produce.
• Each instruction in RISC will generally take one clock cycle to complete.
• CISC has a larger range of instructions.
• CISC programs require less memory.
• One instruction in CISC may take multiple clock cycles to complete.
Hints:This is a comparison question, so make sure you keep your points balanced — the same number for RISC and for CISC. It does not have a limit on the number of points in the questions, so look at the number of marks available. 4 marks mean 4 points, 2 for
RISC and 2 for CISC.
Graphics Processing Units (GPU's) are specialist processors designed for graphics rendering, often in 3D. They tend to run in SIMD (Single Instruction Multiple Data), as much of the processing requires the same action to be performed to multiple pieces of data. For example, rotating an object may require performing the same mathematical function to all of its vertices. GPU's are increasingly being used in other fields, such as physical system modelling, audio processing, password cracking and machine learning.
- CPUs are general purpose processors (1)
- whereas GPUs are designed specifically for graphics (1).
- And so likely to have built in circuitry/instructions for common graphics operations (1).
- GPUs are able to perform an instruction on multiple pieces of data at one time (1)
- often we want to do this when processing graphics (e.g. transforming points in a polygon or shading pixels) (1)
- which means it can perform transformations to onscreen graphics quicker than a CPU (1)
Types of Processor
There are several different types of processor available to use in various different systems:
This is an additional processor working alongside the main CPU. It is designed, and indeed optimized, to be used for a specific task, and improves the processing speed by executing concurrently with the main CPU. An example of this could be a maths co-processor, or a floating point accelerator.
A Scalar CPU is one which performs computations on a single number or set of data at one time. Most modern-day computers have scalar CPU's. They are known as Single Instruction Stream Single Data Stream processors, or SISD for short.
An Array CPU (Also known as a Vector CPU) is one which allows a single instruction to simultaneously operate on multiple data locations. Applying the same calculation to several pieces of data is very quick. It is known as a SIMD processor, Single Instruction Multiple Data. Examples of their use include weather forecasting and airflow simulations.
A parallel processor uses independent processors to work on the same program. The process in question is split into tasks, each of which may be processed by any of the processors. The processors are all coordinated by a complex operating system. Programs must be specifically written for parallel processing, otherwise some processes cannot be completed until others have finished, as they depend on the outcome of the current process.
|Faster than one processor when handling lots of data, with each component of data requiring the same processing||Only certain types of problem are suitable - a problem where its component parts depend on each other cannot be parallel processed|
|Not limited by bus transfer rate||More costly as more processing blocks are needed|
|Can make maximum use of CPU despite bottlenecks||Software is harder to write and a complex OS is required to communicate with multiple processors|
- Simultaneous use of... (do not accept: apparently)
- more than one processor...
- to carry out large number of calculations...
- because the calculations are simple/similar/repetitive...
- carried out in much shorter time (compared with single processor)
- calculations are interdependent with results of one group feeding into next calculations.
(1 per -, max 3 per dotty, max 4) 
(b) Need for complex software/O.S.