What is a Hypertransport?

HyperTransport is a technology designed to replace traditional input/output bus technology in computers. This technology is much faster than previous technologies. It is compatible with most microprocessors, regardless of which company made them.

An input/output bus or I/O bus is a data pathway that links computer processors to each other or to coprocessors or other components. Until HyperTransport was invented, the industry standard for connecting different hardware devices was the Peripheral Component Interconnect (PCI) bus. This technology is still used in many computers and may connect network cards or modems to the computer.

As computer technology evolved, traditional I/O buses could not keep up with evolutions in processor speed. To solve this problem, Advanced Micro Devices (AMD) developed HyperTransport technology in 1997. In 2001, seven other technology companies joined with AMD to create the HyperTransport Consortium. The Consortium is a non-profit group that manages and licenses transport technology within the computing industry. Its membership consists of a variety of companies and academic organizations.

HyperTransport has many advantages over the older bus technologies. It requires less electrical power, so users do not have to worry as much about cooling the computer. It is also narrow, allowing for more flexibility. In the past bus makers would just widen bus size in order to increase speed. At a certain point, this became impractical. The newer technology is able to move more data in less space so the bus can stay narrow.

A PCI bus contains dedicated pins that only send specific kinds of information. This means that pins go unused if the type of data being sent is not the type they generally handle. With HyperTransport technology all pins are able to transport all types of data, so all pins can be used every time data is sent. It moves data in 32-bit packets, which make the most of the width of the bus to send the largest amount of information physically possible.

Each HyperTransport has two data lines. Each line can move 1.6 billion data transfers per second for each pin pair. Since it can send more data on fewer pins, production costs are reduced.

The technology also creates dedicated lines for each device. With traditional bus technology all devices share one bus. This can slow down processing if more than one device is accessing the bus at a time. HyperTransport creates a dedicated line for each device, solving this problem.