Computer Architecture - 11

With the characteristics of optical discs, the price is increasingly low, considered a suitable medium for distributing software to computers. In addition, optical discs are also used for long-term storage of data to replace magnetic tapes.

V.4. TYPES OF MEMORY CARDS

Nowadays, memory cards are one of the latest technologies used as storage devices. Flash memory cards are a type of EEPROM semiconductor memory (technology used to make BIOS chips on motherboards), which is made up of rows and columns. Each intersection is a memory cell consisting of two transistors, these two transistors are separated by a thin layer of oxide. One transistor is called a floating gate and the other transistor is called a control gate. Floating gates can only be connected to the row (word line) through the control gate. When the connection is established, the bit has a value of 1. To change to a value of 0 is a process called Fowler-Nordheim tunneling. The speed, low power requirements and especially the compact size of memory cards make this type of memory widely used in storage and entertainment technology today.

Figure V.3: Illustration of two states of a memory bit in a memory card

V.5. MAGNETIC TAPE

Magnetic tapes have the same technology as magnetic disks but differ from magnetic disks in two ways:

- Access to magnetic disks is random while access to magnetic tapes is sequential. Thus, finding information on magnetic tapes takes more time than finding information on magnetic disks.

- Magnetic disks have limited capacity, while magnetic tapes consist of many reels that can be removed from the tape reader, so the capacity of magnetic tapes is very large (hundreds of GB). With low cost, magnetic tapes are still widely used in backup data storage.

Magnetic tapes, varying in width from 0.38cm to 1.27cm, are packed into rolls and contained in a protective case. The data recorded on magnetic tapes is structured as a number of parallel grooves running along the length of the tape.

There are two ways to record data onto magnetic tape:

Serial recording: with spiral recording, data is recorded sequentially on a track of the magnetic tape, when a track ends, the magnetic tape will rotate back, the magnetic head will record data on the next new track but in the opposite direction. The recording process continues until the magnetic tape is full.

Parallel recording: to increase the speed of reading and writing data on magnetic tape, the read-write head can read and write several adjacent tracks simultaneously. Data is still written vertically on the magnetic tape, but data blocks are considered to be written on adjacent tracks. The number of simultaneous recording tracks on a typical magnetic tape is 9 tracks (8 data tracks - 1 byte and 1 error checking track).

V.6. PERIPHERAL CONNECTION BUS TO PROCESSOR AND INTERNAL MEMORY

In a computer, the processor and internal memory communicate with peripherals using a bus. A bus is a system of connecting cables (about 50 to 100 separate cables) in which a group of cables is defined by different functions including: data lines, address lines, control lines, power supply. Using a bus has two advantages: low cost and easy to change peripherals. People can remove a peripheral or add a new peripheral to computers using the same bus system.

A bus system is cheap to design and implement, since many inputs and outputs share a few simple wires. However, the main disadvantage of buses is that they create bottlenecks, which limit the maximum throughput. Since computer systems for management often use peripherals, the main challenge is to have a bus system that is capable of supporting the processor in communicating with the peripherals.

One of the reasons why designing a bus system is difficult is that the maximum speed of the bus is limited by physical factors such as the length of the bus and the number of components connected to the bus.

Buses are generally of two types: the system bus that connects the processor to memory (system bus, Front Side Bus-FSB) and the peripheral bus (I/O bus) (Figure V.4). I/O buses can be long and can connect to many types of peripherals, which can have different information flows and different data formats. The processor-memory bus is short and usually very fast. In the design phase of the processor-memory bus, the designer knows in advance which components and parts he needs to connect, while the I/O bus designer must design a bus that satisfies many peripherals with very different delays and flows (see Figure V.6).

Figure V.4: Bus system in a computer

Currently, in some computer systems, the peripheral bus is divided into two sub-bus systems. In which, the high-speed bus supports the connection of high-speed devices such as SCSI, LAN, Graphic, Video, etc. and the expansion bus system is designed to connect to low-speed peripherals such as modems, serial ports, parallel ports, etc. Between the two peripheral bus systems in the hierarchical bus system organization is a buffer interface (Figure V.5).

Figure V.5: Hierarchical bus system

Figure V.6: Table showing data rates of peripherals

We can have many choices in designing a bus, as shown in Table V.3.

Bus characteristics

Maybe you are interested!

• Computer Architecture Course - 10
• Computer Architecture - 27
• Computer Structure - Hanoi Industrial College - 15
• The influence of Indian culture on Chinese Buddhist architecture and sculpture - 14
• Basic Features of Authoring a Computer Program

Table V.3 : Major options for a bus

Table V.3 contains the following concepts related to bus owners - the components that can initiate a read or write operation on the bus. For example, the processor is always a bus owner. A bus has multiple owners when it has multiple processors, or when peripherals can initiate a bus operation. If there are multiple bus owners, then there must be an arbitration mechanism to decide which owner is entitled to the bus. A multi-mastered bus can provide more bandwidth by using packets instead of using the bus for each individual operation. The technique of using packets is called task partitioning (using a packet-switched bus). A read operation is decomposed into a read request operation (which contains the address to be read), and a memory response operation (which contains the information to be read). Each operation has a label indicating the type of operation. In task partitioning technique, while the memory reads information at the specified address, the bus is reserved for other masters.

The system bus is a synchronous bus, consisting of a clock on the control lines, and an algorithm for the addresses and data to the clock. Since there is little or no logic to decide what action to take next, synchronous buses are both fast and cheap. Everything on this bus must operate at the same clock.

In contrast, I/O buses are asynchronous buses, which do not have a synchronous clock in the bus system. Instead, there are handshake protocols with specific timing rules used between the transmitter and receiver of the bus. Asynchronous buses are very adaptable to many peripherals and allow the bus to be extended without worrying about synchronization. Asynchronous buses are also easy to adapt to changes in technology.

V.7. BUS STANDARDS

The number and types of input/output devices are not predetermined in information processing systems. This allows computer users to use the input/output device that meets their requirements. Input/output is the interface over which the devices