The motherboard is a large printed circuit board (PCB) to which are fixed the Central Processor, the data bus, memory and various support chips, such as those that control speed and timing (chipset), the keyboard, etc. Printed circuits use copper traces instead of vast amounts of wires, and when there is no space left, another layer is added with more on, so the board gets thicker. In the early days, two-layer boards were common, but six or eight layer ones are what you get from reputable manufacturers now, because the layers also keep the circuit traces separate; you would have one with them going one way and another at 90 degrees, both reducing crosstalk and making the board sturdier –
A good example is the FIC SD-11 for the Athlon, which has six layers. As a result, some of the circuitry resembles twisted pair or coax cabling, which are both well-known methods of reducing radiation and interference in networks, so you might find a signal cable sandwiched between two ground layers, or a multi-layer board could have signals on the outer layers and ground planes in the inner ones, so there is a relatively wide separation.
On the left is a simplified picture of a typical PC-compatible motherboard. The Central Processor does all the thinking, and is told what to do by instructions contained in memory, so there is a direct two-way bus connection between them-the bus width (multiplied by its frequency and number of times data is transferred per clock cycle) determines how much data can be read or written in one go. Extra circuitry in the form of expansion cards is placed into expansion slots on the data bus, so the computer’s basic setup can be changed easily (for example, you can connect more disk drives or a screen here). To save you typing in the same old instructions every time, you buy software prepared earlier and copy it over the data bus into memory via the processor.
One problem is that the parts of the computer that do the most work, such as the video card and hard disk, live here, which is at once the slowest and busiest part of the machine – the ISA bus runs at only 8 MHz (think of it as miles per hour), and the PCI bus at 33 MHz. Of course, the PCI bus is also 4 times wider, but the point is that it is still very much slower than the rest of the machine. Sometimes a math co-processor is fitted to work alongside the main processor, which is specially built to cope with floating point arithmetic (e.g. decimal points). Later CPUs (i.e. some versions of the 486 and upwards) have it integrated, so it’s more correctly called a floating point unit, or FPU.
The main processor has to convert decimals and fractions to whole numbers before calculating on them, and then convert them back again, and the size of the number it can cope with depends on the register width. A coprocessor won’t be used automatically – your software must be aware of it, otherwise you won’t get any benefit. If you’re only doing addition, multiplication, subtraction and division, you won’t find much difference in performance. Oddly enough, a copro in a 286 is slower than one in an XT, due to the connections. Over time, more functionality has been added, and, after the Pentium, our simplified picture now looks something like:
Essentially, the functionality of the chipset is combined on two main components, the Northand South- Bridge chips, which live at each end of the PCI bus. The CPU, Memory and AGP talk to the Northbridge and the Southbridge handles all the I/O, including the ISA bus.
NVidia calls them the IGP and MCP, respectively, standing for Integrated Graphics Processor and Media and Communication Processor (their chipset also uses AMD’s Hypertransport technology, which is mentioned below). The link between the CPU and Northbridge is called Front Side Bus (FSB), which is (usually) the same speed as the Memory Bus, and can be varied, as when overclocking. The Backside bus connects the CPU with L2 cache. But even this is no better, because the essential problems that plagued the original PC still haven’t gone away, in that some parts of the machine simply run too slowly. The PCI bus, at 33 MHz, is 10 times slower than even a 333 MHz CPU. If that weren’t bad enough, all the I/O, including USB and Firewire, goes through the Southbridge and hence the PCI bus, losing all that speed advantage when subject to arbitration, etc.
The latest architecture increases efficiency by handling multiple data streams better. To improve communications, Intel and AMD have removed the PCI bus connection between the North and South Bridges. Intel’s solution is found in the 800 series chipsets, and works like a network system, which is why individual components are called hubs. For example, the North Bridge is replaced by the Memory Controller Hub (MCH), in the middle, and the South Bridge by the I/O Controller Hub (ICH), which now has the PCI bus as a subcomponent. The BIOS becomes a Firmware Hub (FWH), which handles the ISA bus, if present. The MCH, unfortunately, is only designed to work with the very expensive Rambus. All are tied together with a 266 Mbps interface.
AMD uses a high speed, high performance, unidirectional point to-point link between components, called HyperTransport, which can be up to 32 bits wide. It can be likened to something like Firewire just for the motherboard, but faster. In fact, it is about 50 times faster than PCI, 6 times faster than PCI-X and 5 times faster than InfiniBand 4 channels. Asymmetric buses can be used where different up- and downstream requirements exist.
I hope that you understand my post. if you like this post then share my post, and comment on my post if you have any suggestion or Queries related to Motherboard. Thank you!