diff options
Diffstat (limited to 'notes/hardware.tex')
| -rw-r--r-- | notes/hardware.tex | 239 |
1 files changed, 239 insertions, 0 deletions
diff --git a/notes/hardware.tex b/notes/hardware.tex new file mode 100644 index 0000000..d9cebed --- /dev/null +++ b/notes/hardware.tex @@ -0,0 +1,239 @@ +\chapter{Hardware} + +\begin{myquote} +Hardware is the wiring and other physical components of a computer. +\end{myquote} + +Modern computers have several typical components. There is a +motherboard which has wiring printed on it in order to connect all of +the other components. The central processing unit (CPU) is the main +component that performs the instructions of computer programs. +Graphics and sound cards provide the user with graphics on screen and +sound through headphones or speakers. Input/output ports (I/O ports) +transfer data between input devices such as a mouse or keyboard and +output devices such as a printer or speakers. + +A popular I/O connector is the Universal Serial Bus (USB). A bus is a +collection of parallel wires that transmit data as electrical signals. + +\section{von Neumann architecture} + +\footnote{Prof David Murray, University of Oxford, `A3 Computer +Architecture', +\url{https://www.robots.ox.ac.uk/~dwm/Courses/3CO_2000/3CO-L2.pdf}} +In the 1940s von Neumann's (pronounced Noy-Man, not New-Man) created a +design for general purpose computer, reliant on a CPU alongside +\textbf{a single memory unit for both instructions and data}. +Instructions are executed in series in the `fetch-decode-execute' +cycle. It is very successful due to its need for only one memory unit +and the ability to have general purpose rather than specific purpose +hardware, which early computers were and had to be rewired to change. + +The design however does lead to the \textbf{`von Neumann bottleneck'} as data +and instructions must be fetched from the same memory unit over the +same bus. + +\section{Boolean logic} + +George Boole was able to devise a mathematical representation of +arithmetic using only 1 (True) and 0 (False). Computers use circuits +that can perform this logic by using on and off circuitry to represent +1s and 0s. We can represent these circuits as `gates' and describe the +logic in truth tables to show all possibilities. +Typically we consider: +\begin{itemize} + \item `And' gates, where the result is true if and only if both of + the two inputs are true. + \item 'Or' gates, where the result is true if either of the two + inputs are true. + \item 'Not' gates where the result is the inverse of the one input, + so a 1 becomes a 0 and a 0 becomes a 1. +\end{itemize} + +\section{Central processing unit and performance} + +The central processing unit (CPU) performs the instructions that we +ask of a computer. It controls data and instructions through the +\textit{fetch-decode-execute} cycle, where an instruction is fetched +from main memory, decoded by the CPU so that it can perform the +correct logic (such as adding two numbers), and then the instruction +is executed giving the output of the instruction (say the result of +the addition). + +There are some critical components of the CPU that we typically consider: + +\begin{itemize} + \item The \textbf{control unit} `oversees' the operations of the CPU and + organises the operations performed and the movement of data + and instructions. + \item The \textbf{arithmetic and logic unit} (ALU) carries out the boolean + logic and mathematics via circuits that are designed to do so. + \item \textit{Registers} are high speed, small memory locations which hold + important data. + \begin{itemize} + \item The \textit{program counter} is a specific register + which stores the location in memory of the next + instruction to be executed. + \end{itemize} +\end{itemize} + +The CPU executes instructions at a rate dictated by the system +\textit{clock}. The \textit{clock speed} determines how many +\textit{clock cycles} (fetch-decode-execute cycles) occur per second. +This is measured in Hertz (Hz), which means times per second. A CPU +in a computer today cycles with a clock speed around 4 GHz or +around 4 billion clock cycles per second +\footnote{7-Zip LZMA Benchmark, Intel Ice Lake: +\url{https://www.7-cpu.com/cpu/Ice_Lake.html}}. + +The performance of the CPU can be bettered by: + +\begin{itemize} + \item Increasing the clock speed, so that more clock cycles occur + per second (overclocking). This does, however, increase power + consumption and CPU heat production. + \item Increasing the number of CPU cores. A \textit{multicore} CPU + has the ability to do multiple cycles at the same time, + depending on the number of cores and, for programs that have + the functionality, can perform parallel processing where a + large algorithm or calculation is distributed between cores, + thus improving performance. + \item Increasing cache size, to allow faster access to data and + instructions. + \item Increasing cache/memory speed, to allow faster access to data and +\end{itemize} + +\section{Memory} + +Volatile memory is erased when electrical power to the system is +turned off or interrupted, meaning that all data is lost. Random +access memory (RAM) is reasonably fast and is volatile. RAM holds +current application and operating system data, as well as instructions +for programs that need to be executed. + +Read only memory (ROM) is not volatile and cannot be changed by the +system, as the name suggests. Modern computers use ROM to contain the +Basic Input/Output System (BIOS) and other instructions that are +required in the \textit{booting process} of a computer. This performs +tasks such as initialising hardware and initialising the operating +system. + +Virtual memory might be necessary when there is insufficient RAM for +data and instructions. This stores data that would be stored in RAM in +secondary storage. \textit{Paging} is the process of swapping chunks +of data between RAM and storage. Because secondary storage is +significantly slower than RAM and due to the paging process, when +virtual memory is necessary, the computer is significantly slowed +down. + +\section{Cache} + +Cache is memory that is much smaller and more expensive than RAM, but +is significantly faster also. Cache is used to store current +instructions and data in order to increase CPU performance. There are +three levels of cache, each with varying size and speed +\footnote{7-Zip LZMA Benchmark, Apple M1: +\url{https://www.7-cpu.com/cpu/Apple_M1.html}; Intel Ice Lake: +\url{https://www.7-cpu.com/cpu/Ice_Lake.html}}: + +\begin{itemize} + \item Level 1 (L1) cache is very fast and small. It is located + directly on CPU cores. It is in the range of 64kB. + \item Level 2 (L2) cache is larger and slower than L1. It is + shared between CPU cores but is still very close to them, It + is in the range of 512kB. + \item Level 3 (L3) cache is significantly slower and larger than + L2, but is still much faster than normal RAM. It is in the + range of 8MB. +\end{itemize} + +\section{Storage media} + +--intro + +\subsection{Magnetic storage} + +Magnetic storage relies on rotating \textit{platters} coated with +magnetic material. A \textit{read/write head} can move across the +platter. To write data the head contains electromagnets that can give +the magnetic particles a positive or negative charge, representing one +or zero in binary. A platter is divided into \textit{tracks} which are +concentric circles, and segments which are radial divisions. A single +segment on a single track is called a \textit{block}. Data is +generally organised by blocks. + +Magnetic storage is high capacity with the highest available capacity +in 2021 being 20TB +\footnote{\url{https://www.tomshardware.com/news/seagate-ships-hamr-hdds-increases-dual-actuator-shipments}}. +Magnetic storage is reasonably cheap and fast. However, the read/write +head of a modern hard drive can be only 5 nm from the platter, which +can spin at up to 15,000 rpm +\footnote{\url{https://en.wikipedia.org/wiki/Hard_disk_drive}; +\url{https://en.wikipedia.org/wiki/Flying_height}}. This can create a +lot of noise, power. The moving parts involved can fail and the +read/write head can \textit{crash} into the platter causing +significant data loss, particularly in the case the drive is impacted +when running, such as if a laptop is dropped. + +\subsection{Solid state storage} + +\textit{Flash} memory chips store data with transistors that can reain +their one or zero sate without power (non-volatile). As there are no +moving parts, they are durable, silent, and portable. They are, +however, very expensive per unit data and risk corruption of data in a +power interruption event. + +\subsection{Optical storage} + +Optical media (such as a DVD) contains an inwardly spiralling track +with pits and lands (small changes in height) to store data. A LASER +is aimed at the media, and the reflection is captured by a sensor. The +change from a pit to land can represent a zero or a one. This relies +on the laser to be moved very precisely by motors to align with the +track, and for the disc to be rotated at a precise speed (much like +magnetic storage) in order to read data at the correct rate. + +Optical storage is very portable and can be read by many devices, but +requires that they have a compatible drive. However, large quantities +of data are expensive to store and discs can very easily be scratched +or damaged, although error correcting methods are generally used +\footnote{3Blue1Brown, `How to send a self-correcting message': +\url{https://www.youtube.com/watch?v=X8jsijhllIA}}. Reading data from +an optical disc is slower than solid state or magnetic storage. + +\subsection{Cloud storage} + +Businesses and other organisations (such as schools) traditionally use +servers for data storage of all data and for running services such as +email and websites. This however, depends on skilled IT staff and +expensive equipment and maintenance. Cloud data storage involves the +use of offsite servers around the world, provided by large +organisations providing storage for a monthly cost. This can be +cheaper than an in house operation and provides the benefit of the +server model to individuals such as the ability to access data from +many devices and reliable backups. However, a cloud storage solution +requires a reliable internet connection to access data, can become +expensive for large amounts of data and due to the repeat cost, and +raises philosophical concerns such as those surrounding the ownership +of data. + +\section{Embedded systems} + +An embedded system +\footnote{\url{https://en.wikipedia.org/wiki/Embedded_system}} is +\textbf{custom built hardware} designed to perform a +\textbf{specific task as part of a wider system}, as opposed to a +general purpose computer. Examples of embedded systems are visible +all around us, such as in common electricals appliances such as +printers, televisions, and cameras; in household appliances such as +dishwashers and microwaves; and in advanced situations such as engine +management in cars, guidance systems in aeroplanes, and medical +equipment such as an MRI scanner. + +Embedded systems generally have no RAM and only have ROM which +contains the \textit{firmware} for the specific program the system +needs. They have \textbf{no operating system}. The have +\textit{sensors} to gather informations about their surroundings (such +as a thermometer or camera) and \textit{actuators} to act on or change +their surroundings (such as the shutter in a camera or the motor in a +washing machine). |