| | | | Technology (HTT) used in Intel's Pentium 4 processors and all of their Xeon processors that makes a single CPU appear as two logical processors to the Operating System. A twin-thread version of Simultaneous Multithreading (SMT), that can trace it's earliest routes back to research carried out in the 1950s. On hyperthreading processors, some parts of the chip architecture, namely the architectural state (contains the control registers and general purpose registers) are duplicated. This has the advantage of allowing two sets of instructions or threads to be performed by the processor in parallel for every clock cycle, rather than just one. There is a significant gain in speed over machines with single processors without hyperthreading of up to 30% when applications that can take advantage of the technology are running, however the increase is not as great as when two physically distinct processors are present and only applies when certain operations are being performed by certain applications (namely those written specifically to take advantage of the technology). Paradoxically, because of certain features of the P4 architecture, some applications can actually see a small decrease in performance when the processing operations of one thread tie up the resources required by the other, preventing it from performing as fast as it could otherwise do (later versions of the P4 chip address this issue partially). In addition, some hyperthreading processors have been labelled energy-inefficient when compared to running a dual core design. Although theoretically multiple threads (i.e. more than 2) are possible, to all intents and purposes, from a practical point of view, the number of threads per processor/core is limited to two because of the constraints of physical size on design. | | | |
|
