Main Page | See live article | Alphabetical index

HyperTransport

HyperTransport is a bidirectional serial/parallel high-bandwidth computer bus. The technology is used (or is planned to be used) by a number of companies that notably include NVIDIA, AMD and Apple Computer, to replace existing local bus systems. In August 2003, IBM joined the HT group. It is believed this move was made to help in the future development of IBM's POWER processors, PPC chips (Used by Apple in their iBook and G5 lines), and compete better in the marketplace.

HyperTransport runs at 200-800 MHz (compared to most systems at 133 MHz or so) and can be "double-pumped", meaning it sends data on both the up and down "tick" of the 800 MHz clock. Even in a simple serial configuration this allows for transfer rates of 1600 Mbit/s, or 200 MByte/s.

HyperTransport allows for various bus widths depending on needs, from 2 (bidirectional serial, 1 bit each way) to 32-bit (16 each way) busses are allowed. The full-sized 32-bit bus has a transfer rate of 6400 MByte/s, making it much faster than existing standards. Busses of various widths can be mixed together in a single application, which allows for high speed busses between main memory and the CPU, and lower speed busses to peripherals, as appropriate.

HyperTransport is packet-based, with each packet always consisting of a set of 32-bit words, regardless of the physical width of the bus interconnect. The first word in a packet is always a command word. If a packet contains an address, the last 8 bits of the command word are chained with the next 32-bit word to make a 40-bit address. The remaining 32-bit words in a packet are the data payload. Transfers are always padded to a multiple of 32 bits, regardless of their actual length.

Its electrical interface uses 1.2 volt Low Voltage Differential Signaling (LVDS).

Applications for HyperTransport

Backplane replacement

The primary uses for HyperTransport will be to replace the backplane bus, which is currently different for every machine (or some set of them). For instance, a Pentium cannot be plugged into a Pentium 4 backplane. In order to expand the system the backplane must include adaports for the various standard busses, like AGP or PCI. These are typically included in a controller called the northbridge.

A similar computer implemented with HyperTransport is more flexible, as well as being faster. A single PCI<->HyperTransport adaptor chip will work in any machine, and allow the PCI cards to talk to any CPU. nVIDIA already implements HyperTransport in their GEFORCE chip sets, and AMD is using it as their backplane in their future chipsets.

Multiprocessor interconnects

Another use for HyperTransport is as an interconnect for NUMA multiprocessor computers.

Router or switch backplane replacements

HyperTransport could also replace the backplanes in routers and switches. Routers and switches have multiple connections ports and data has to be forwarded between these ports as fast as can. E.g. a four port 100 MBit/s Ethernet switch needs a backplane that is 800 MBit/s fast (100 MBit/s * 4 ports * 2). With HyperTransport, vendors could use four HyperTransport-to-Ethernet bridges and use HyperTransport as backplane instead.

See also:

Rapid I/O

External links:
http://www.hypertransport.org/technology.html
http://www.hypertransport.org/faqs.html