Hyper-Threading Technology

We all want our computers to be as speedy as they can be. There are many different ways to increase computer performance through different types of upgrades. Processors have become speedier because of demand and competition.

To make processors fast, chipmakers have been creating new CPU architectures to process information and milk every ounce of processing power available. Intel created Hyper-Threading technology as an upgrade in CPU architecture and quietly integrated it into some of their processors for development and testing purposes.

It is based on the idea of simultaneous multi-threading technology (SMT), where multiple physical CPUs are used to process multiple threads at once. As an alternative to using multiple physical processors, Intel created multiple logical processors inside a single physical CPU.

It allows multi-threaded software applications to execute threads in parallel. Consequently, resource utilization provides higher processing throughput.

It is basically a more superior form of Super-threading that was first introduced on the Intel Xeon processors and was later added to Pentium 4 processors. This type of threading technology was not present in general-purpose microprocessors.

To boost performance, threading was allowed in the software by splitting instructions into multiple streams so that multiple processors could act upon them. By using this technology, processor-level threading can be utilized which provides more efficient use of resources for greater parallelism and improved performance on today's multi-threaded software.

Hyper-Threading is a multi-threading technology in which SMT is achieved by duplicating the architectural state on each processor, while sharing one set of processor execution resources. It also produces faster response times for a multi-tasking workload environment.

By permitting the processor to use on-die resources that would otherwise have been idle, it offers a performance boost on multi-threading and multi-tasking operations for the microarchitecture.

In a CPU, every clock cycle has the ability to do one or more operations at a time. One processor can only handle so much during an individual clock cycle. Hyper-Threading permits a single physical CPU to fool an operating system, capable of SMT operations, into thinking there are two processors.

It produces logical processors to handle multiple threads in the same time slice, where a single physical processor would normally only be able to handle a single operation. There are some prerequisites that must be satisfied before taking advantage of this technology.

The first prerequisite is that you must have a Hyper-Threading enabled processor, HT enabled chipset, BIOS and operating system. Further, your operating system must support multiple threads. Finally, the number and types of applications being used make a difference on the increase in performance as well.

Hyper-Threading is a hardware upgrade that makes use of the wasted power of a CPU, but it also helps the operating system and applications to run more efficiently, to do more at once. There are millions of transistors inside a CPU that turn on and off to process commands.

By adding more transistors, chipmakers typically add more brute force computing power. More transistors equal a large CPU and more heat. The technology is aimed at increasing performance, without significantly increasing the number of transistors contained on the chip, making the CPU footprint smaller.

It offers two logical processors in one physical package. Each logical processor must share external resources like memory, hard disk, etc. and must also use the same physical processor for computations.

The performance boost will not scale the same way as a true multiprocessor architecture, because of the shared nature of Hyper-Threading processors. System performance will be somewhere between that of a single CPU without Hyper-Threading and a multi-processor system with two comparable CPUs.

This technology is independent of any platform. Some applications are already multi-threaded and will automatically benefit from this technology. Multi-threaded applications take full benefits of the increased performance that this technology has to offer, permitting users to see immediate performance gains when multitasking.

It also improves reaction and response time, and increased number of users a server can support. Today's multi-processing software programs are compatible with Hyper-Threading technology enabled platforms, but further performance gains can only be realized by specifically tuning the software to utilize it.

For future software optimization and business growth, this technology complements traditional multi-processing by providing additional headroom.