What is a computer bus? What are the different bus types in a computer?

[Pyrotechnic]

Definitions

The various components of a computer are interconnected by wiring systems ensure the transmission of signals necessary for its operation. As it is hardly possible to connect each unit to every other computer, it uses lines operated jointly by all devices attached to it, called the bus cable assemblies by analogy with transport city of the same name.

A bus can be used by all units are connected, but never by more than two of them at the same time it raises problems of expectation and arbitration requests of use. A special circuit, called a bus controller, is appointed for the purpose of organizing all access and avoids conflicts that could occur within the computer.

A bus is capable of carrying signals corresponding to three fundamental types of information: addresses (which identify the location of a device or a data in memory), data and commands (such as for example the clock signals synchronization).

A bus is finally characterized by the number and arrangement of its lines and its operating frequency in megahertz (MHz): A bus has 16 lines for example is a parallel bus 16 bits, which means that each line is able to carry 1 bit at a time. The operating frequency of this bus is used to calculate bandwidth, ie the flow of information it can convey in the same period. If our 16-bit bus operates at a frequency of 8 MHz (recall: 1 Hz = 1 pulse / s), we can estimate the theoretical bandwidth maximum: 8 x 16 = 128 Mbit / s or 128 / 8 = 16MB / s.

[Pyrotechnic]

The different bus types found in a computer

The Processor Bus: It is the communication channel between the CPU, the chip further associated with it (usually called chipset) and the external cache memory. The role of this bus is to transfer signals to and from the CPU to maximum speed, so it is much swifter than the other bus system.


The Memory Bus: It ensures data transfer between processor and main memory (RAM). The rate of transfer of information traveling through the memory bus is less than the information conveyed by the processor bus, which requires the establishment of a memory controller to audit the interface between the processor bus fastest and the slower RAM. Note that the interface of the main memory bus is always the same as the processor bus: for example, a system with a 64-bit processor (Pentium), the memory bus is 64 bits, this number denotes the size of a memory block. If you use a Pentium 64-bit, each memory block added to the RAM in any transaction must necessarily be of 64 bits. Hence certain constraints when you install of the RAM in your computer according to their characteristics: 32 bit SIMM, DIMM 64-bit, etc..).


The Address Bus and the Bus Command:They are actually subsets of the processor bus and memory bus. These consist of various dedicated lines, some of them responsible for conveying addresses, other data or commands, the address lines are the address bus, command lines determine the bus command (the bus of a Pentium processor is for example made of 32 address lines, 64 data lines and some lines of controls).

The address bus indicates the memory address or an address of a bus used during a data transfer within the computer. The width of the address bus also determines the maximum amount of RAM that the processor may send (an address bus of a width of 16 bits allows the processor to send 216 = 65,536 memory addresses).

The bus command vehicle all signals used to synchronize the various activities taking place in the functional units of computer clock signals, signals the read / write, interrupt signals, etc..


The Data Bus (or bus I / O): It ensures the data transmission process between the various elements of the computer as such, it is like the address bus and controls a subset of the processor bus and memory (remember the 64 lines data bus of the Pentium processor). The bus I / O allows not only the components contained on the motherboard to communicate data between them, but also add additional devices with expansion cards for this purpose, the bus of E / S is characterized by a number of slots (or slots).

Extending the capacity of a computer with the addition of devices is crucial for the basic systems are unable to meet all the expectations of users: the essential components can then be attached to the connector bus I / O such as graphics cards, sound cards, network interface cards or SCSI adapters.

[Pyrotechnic]

Different types of I/O buses

The ISA bus (Industry Standard Architecture)
The ISA bus is behind the 8-bit bus which powered the first IBM PC in 1981 and has subsequently been extended to 16 bits for the IBM PC / AT 1984. The first version of the ISA bus running at 4.77 MHz, but the extension to 16 bits was followed by an increase in operating frequency of the ISA bus 6 and 8 MHz. Subsequently, the industry as a whole has chosen a standard speed of 8.33 MHz for all versions of ISA bus (8 and 16 bit).

The ISA bus has long formed the basis of micro-computer and is the first architecture used in most PC systems until the late 1990s, when the ISA bus has gradually been replaced by standard PCI bus. In 1993, Intel and Microsoft have begun to change the specifications of the ISA bus to reach ISA Plug & Play with this new version, the operating system is able to automatically configure the newly connected devices without requiring the user to manually adjust these settings using jumpers (jumpers) and switches on the device concerned.

It may seem strange that an old system also has been used for over 20 years, but this is explained by factors of reliability, availability and compatibility of the ISA bus. While it is undeniable that it is slow (8 MB / s max.), The ISA bus is still faster than many modern devices such as sound cards or modems, however, it is now completely obsolete when it comes to connecting high speed devices such as 2D/3D graphics adapters or disk drives.


The MCA bus (Micro Channel Architecture)
The advent of 32-bit processors (the 386 with Intel in 1985) made the ISA bus unfit to manage the power of this new generation of chips: when a 386DX was able to transfer 32 bits of data simultaneously, the ISA bus slowed considerably as operations limited to 16 bits. Therefore in 1987, IBM has tried to impose a new standard: the MCA bus, superior in every respect to the ISA bus.

The MCA bus is a 32 bit bus operating at 10 MHz, which offers ease of configuration of connected devices, without jumpers or DIP switches. Better still, it supports bus mastering, ie the takeover of the bus by any device that is connected to transfer information to another component without requiring the intervention of the central processor, which discharge all the work and he leaves the 'entertainment' to take care of something else.

As counterpart, the MCA bus is completely incompatible with standard ISA so many expansion cards designed for the ISA bus does not work on systems MCA. Finally, IBM demanded royalties from users of the new model, which resulted in the rejection of MCA bus by the entire industry and the development of competing EISA bus.


The EISA Bus (Extended Industry Standard Architecture)
The EISA bus was introduced in September 1988 as a response to the introduction of IBM's MCA bus and how he intended to use it. Originally EISA bus manufacturers found nine competitors IBM: AST Research, Compaq, Epson, Hewlett-Packard, NEC, Olivetti, Tandy, WYSE, and Zenith Data System.

The EISA bus is a 32 bit bus as the IBM MCA, operating at 8.33 MHz and is compatible with the standard ISA: it adds 90 new connections to a conventional ISA connector without changing their appearance and in these conditions, ISA cards can still be used with EISA expansion slots. This new format also uses a software system similar to the MCA bus to automate the configuration of expansion cards without using sets of jumpers or switches. Developments, the EISA bus integrates for the first time the possibility of sharing an IRQ with multiple devices.

[Pyrotechnic]

Local Buses

The bus I / O presented so far have one thing in common: they are relatively slow. The restricted speed is a legacy of the first PC in which bus and processor bus I / O share the same clock frequency of 8 MHz. The initial problems have actually occurred in the early 1990s with the success of GUIs such as Windows: they require the treatment of so many video data bus I / O has become the bottleneck Throttling the most important computer systems - What interest have a microprocessor clocked at 66 MHz if the rate of data transfer on the bus I / O cannot exceed 8 MHz?

The solution to this dilemma was found by moving some expansion slots bus I / O to a place where they could enjoy the high speed processor bus, a bit like the external cache memory. This is the architecture which is called local bus (local bus).

The VESA Local Bus (VESA Local Bus or VL Bus )
To improve video performance computer systems, the VESA Local Bus was developed in August 1992 by the Committee VESA (Video Electronics Standards Association), an association nonprofit corporation created by the NEC. The basic idea of the VL-Bus was to connect directly pin CPU in vogue - the Intel 486 - and plug them into a media expansion slot, the result is a cheap architecture no chip or chipset additional control is needed.

The VESA bus is a 32 bit bus can theoretically change the maximum frequency of 66 MHz. In reality, the electrical characteristics of the connector limits the execution frequency to 40 or 50 MHz and, in practice it was found that the use of a VL-Bus at speeds greater than 33 MHz is capable of 'cause many problems. Therefore, 33 MHz has become the acceptable speed for this bus.

Moreover, the structure of VL-bus as it was originally designed makes it inseparable from the processor bus 486. Although the VL-Bus can be adapted to other processors (including the 386 or Pentium), compatibility remains elusive and is in conjunction with a 486 that the VL-Bus offers the best performance.

Physically, the VESA bus is an extension of the slots used for the base system computer: a configuration based on ISA, for example, the VL-Bus is found as an extension of the ISA slots 16 existing bits.


The PCI (Peripheral Component Interconnect)
The PCI bus was developed by Intel and several other leaders of the computer industry in early 1992. In reality, the PCI bus is not a true local bus: it occupies an intermediate bus located between the basic premises (processor bus and memory bus) and bus I / O standard, the PCI bus add somewhat a new layer to the standard configuration of the system.

The width of the PCI bus is set to 32 bits and its operating frequency to 33 MHz, although the designation for a PCI bus of 64 bits and 66 MHz already exists (the new specifications will be introduced will be justified when the cost of the bus changes PCI to increase its bandwidth).

[Pyrotechnic]

Compared with the VESA bus, the PCI bus has a lot of advantages:

1) While the frequency of VL-Bus is intimately linked with the CPU, the PCI bus is independent;

2) When the VL-Bus is active, it overrides the local buses and forces the processor to stop its activities, it is not the case of PCI bus which works in parallel with the processor bus: when the processor sends data to a destination device, they are stored in the buffer controller PCI bus, the processor can then devote to other activities while the controller is responsible for routing information to the device concerned. Conversely, the devices can continue to route data via the PCI bus even when the processor is active;

3) Like the VL-Bus, the bus supports PCI bus mastering;

4) Another major feature of the PCI bus is that it is the model specifications Plug & Play (PnP) Intel: this means that the expansion cards in PCI format or does riders or switches and they are automatically configured by software. Since 1995, most PC systems include a PnP BIOS automatically ensures that the device configuration conformesd this standard;

5) The PCI bus can operate with a voltage of 3.3 V in laptops (cons to 5V bus VESA), which is greatly in favor of energy conservation in such computers.

Note that unlike all the formats described so far bus, PCI bus interface is not exclusively for PC: it is also found as standard in some models of Macintosh (iMac, G3, G4, etc..).

The PCI slots are easily recognizable by their small size and white standard.


The AGP (Accelerated Graphics Port)
As the ISA bus was eventually found unfit to manage large flows of data related to the development of graphical interfaces, the PCI bus has now reached its limits since the onset of a particular application domain resource-intensive system: 3D. The PMA has been developed and introduced by Intel in 1997 in order to overcome the shortcomings of the PCI bus with respect to the video.

This brings the first observation following the PMA is not a 'bus' in the true sense of the term, because a bus must be capable of supporting multiple peripheral I / O, which is not the case of the PMA that do Supports that in reality it is a 'port', i.e. an independent and direct connection between the CPU, RAM and graphics card system computer.

[Pyrotechnic]

What are the differences and advantages of AGP over PCI bus?

1) The concept of AGP is based on the specifications PCI 2.1 64-bit extension: it has a width of 32 bits and operates at the maximum frequency of 66 MHz, which gives a theoretical bandwidth double that of buses PCI 264 MB / s. Furthermore, the PMA has a mode '2 X 'which allows him to double the bandwidth to 528 Mb / s (1056 Mb / s with AGP 4X).

At each cycle, the clock signal timing transits from 0 (no signal) to the value 1 (presence of signal), which corresponds to the rising edge of the signal, before returning the value 1 to value 0, which corresponds to the falling edge signal. While the PCI bus is able to convey information during the rising phase of synchronization signal, the PGA can do on both fronts.

2) The PCI bus share its bandwidth of 132 Mb / s with all devices attached to it, which slows down the exchange of data when a graphics adapter is present and accounts for the majority resource bus. The advent of the PGA allows one hand to the video subsystem to provide a dedicated circuit performance, and also relieves the PCI bus of the intense traffic related to the display;

3) The AGP also has the ability to directly access the main memory system (without CPU intervention, it is therefore a DMA - Direct Memory Access) to store a surplus of data when the memory circuit graph is saturated. Better yet, the PMA has the technology DIME (Direct Memory Execution), which can perform calculations textures directly into the main memory before loading into memory video card.

Note : however that the solution DIME is not faster than the execution of operations in the memory of the video graphics adapter: first because the special bus which connects the processor to video memory on the graphics card is much more Swift as the PMA and then for the central RAM of the computer is usually 4 to 6 more slowly than the video memory, which slows down the processing of textures that are made. DIME technology is especially useful for configurations with a graphics circuit with little video memory;

4) The course runs AGP bus mastering, and another nice feature, can transmit multiple requests by following the technique of pipelining. While the PCI protocol requires that each request is followed by a receipt in order to send the following query, the PMA is not required to wait for a response from the processor to send multiple requests simultaneously;

5) To end this non-exhaustive list capabilities of AGP, note that it incorporates the technology SBS (Side Band Signaling, or Side-band signaling), which refers to the use of a set of circuits dedicated s 'adding the PMA: in addition to the 32-bit bus, an sideband additional 8-bit enables the AGP card to send new requests while receiving data transmitted by the motherboard response to requests previously issued. The SBS also carries a marker that maps synchronization modes of operation (1X, 2X, 4X) of those with AGP graphics card.
The AGP slot is easily recognizable on the motherboard of the same size around the PCI slots and brown, set slightly back from them.

[Pyrotechnic]

The bus schedules

The PC Card bus
This interface was designed in 1989 by a consortium of more than 300 computer manufacturers: the Personal Computer Memory Card International Association (PCMCIA) to provide computers protable the same expansion capabilities as desktops. The PCMCIA has undertaken to standardize the expansion cards for portable machines with credit card format, with the following dimensions: 54 x 85 mm for a 68-pin connector and three standard thicknesses have also been adopted.

Note: There is a card type IV 15.5 mm thick, but is not recognized by the consortium PCMCIA.

Several modifications to the standard PC Card has been created between 1990 and 1994, which remained during this time a 16-bit interface relatively slow, limited bandwidth (2 Mbps), unable to endure the bus mastering and with 26 lines of address bus not allowed to handle only 64 MB of RAM maximum, the main advantage of the PC Card bus is above all the small connectors and expansion cards and the possibility of connection and disconnection Hot devices.

In 1995 a new version of the standard PC Card has been introduced under the name CardBus: it is a 32-bit interface much more efficient, operating at 33 MHz, supporting bus mastering and whose blood was reduced to 3.3 V (instead of 5V).


The USB (Universal Serial Bus)
The USB specification was published in 1996 by a consortium of Compaq, Digital, IBM, Intel, Microsoft, NEC and Northern Telecom. The USB bus is essentially an interface to theoretically connect to your computer up to 127 devices in a chain, with a transfer rate of 12 Mbit / s (about 1.5 Mb / s). The advantages of USB is self-identification device by the system and can connect and disconnect them warm without having to remove power from the computer before. Connecting serial devices can also save the connection ports, which is not the case with parallel and serial interfaces treaty requiring that each device is associated with an individual connector.

With a bandwidth of 12 Mbps shared between devices connected, only components that do not require high transfer rates are at present concerned with USB keyboards, mice, printers, digital cameras, joysticks etc.. A USB version 2.0 is yet under development and will increase the performance of USB by 40, while maintaining compatibility with current USB 1.1 standard.

Since 1996, all computers are equipped with standard USB ports and this standard is supported from OSR 2.1 of Windows 95.


The IEEE 1394 (FireWire)
The IEEE 1394 was developed by Apple in 1995 in collaboration with the IEEE (Institute of Electrical & Electronics Engineers Inc.). The other name by which this is known bus - FireWire - is the trade name filed by Apple IEEE 1394.

The IEEE 1394 standard describes a digital bus at high speed (up to 400 Mbit / s) to connect to the computer up to 63 devices chained. As the USB which it closely resembles, IEEE 1394 supports Plug & Play and Hot-Plug (hot-plug devices). Its high bandwidth intended especially for the IEEE 1394 devices that generate large data transfers with the computer, including multimedia components: digital cameras and camcorders, audio synthesizers, disk drives high-speed scanners and printers of Professional Imaging, etc..

Ultimately, it is largely possible to replace all the serial interfaces (RS-232), parallel (Centronix) and even SCSI standard unified standard IEEE 1394.