|It's been a while since there have been any major changes to the chipsets motherboards have been built around over the past few years. Even non-Intel chipsets seem to follow the same general mold for their foundation, you generally have two parts to a chipset, the North Bridge and the South Bridge, the former controlling the memory/graphics and the latter controlling the Disk/IO parts of a motherboard. Ever since the advent of the PCI bus in the first Pentium mainboards, Intel has pushed the same standard North/South Bridge chipset configuration with all of their chipsets, and therefore the competition has refrained from making the bold move of not conforming to the standards set by Intel.|
|The time for change is upon us, and Intel's outdated North/South Bridge chipset configuration system is on its way out, the replacement? What is known as the Accelerated Hub Architecture (AHA), the basis for Intel's most recent entry into the chipset arena, the i810 aka Whitney chipset.|
The problem with the more conventional way of how a chipset functions is that transfers from the peripherals, disk drives, and video sub-systems are made via the PCI bus rather than directly to the CPU/memory. You can imagine the PCI bus as being a tunnel of a set width (32-bits) able to transfer only a certain amount of traffic at a time (32-bits * 33.3MHz operating speed of the PCI bus = 1066.6Mbits/s = 133MB/s transfer rate). The reason for making the move to the PCI bus was because the former "king of the hill," the ISA bus was slowly but surely becoming a bottleneck for operations where data was transferred to/from the CPU. The ISA bus can be imaged as a smaller tunnel of a set width (16-bits), and able to transfer considerably less data based on a slower operating frequency and a smaller bandwidth in comparison to the now commonplace PCI bus.
If you remember back to the days of overclocking your favorite Socket-7 CPUs, one of the major concerns was that the higher FSB settings would result in a PCI bus frequency that would cause your other peripherals to malfunction. The reason for this was that essentially all of the peripherals that went in to your system and connected to your motherboard operated via the PCI bus, it would only be a short time before the PCI bus would become a bottleneck.
During 1997 Intel began planning ahead, by releasing their first Accelerated Graphics Port (AGP) enabled chipset, the i440LX, which gave graphics vendors the ability to communicate directly with the memory and then the CPU without being limited by the PCI bus. Since 3D games and applications increased in popularity, the need for a greater amount of available bandwidth became a more pressing issue, thus resulting in the, what some may argue, premature, release of the AGP enabled i440LX chipset. The AGP bus offered a few key improvements over the PCI bus when it came to graphics cards, the most obvious being that it was a bus independent of the PCI bus. The AGP bus can be presented in the same terms as the PCI bus, as an even wider tunnel of a set width (64-bits) able to transfer only a certain amount of traffic at a time (64-bits * 66.6MHz operating speed of the AGP 2X bus = 4266.6Mbits/s = 533MB/s). The basic nature of the AGP 2X bus, the standard AGP specification for most motherboards today, indicates that it is faster than the PCI bus, however you can only use the AGP bus for graphics cards what about the rest of your system?
Here's where Intel's new Accelerated Hub Architecture comes into play.