Hyper-threading technology and dual-channel memory control technology can be said to be two different technologies. Of course, these two technologies can find their location and value from different application levels. In order to let everyone know the two technologies, the author believes that the corresponding profiling and vertical comparison test of these two technologies can find the answer we need. Of course, there is only this, in order to make us "taking the machine", do "some targeted" to avoid being wasted "silver two" left in your own pocket.
First, what is "super thread" processor technology
Simple definition "super thread" technology
The so-called ultra-thread technology uses special hardware instructions to simulate two logical kernels inside the multi-threaded processor into two physical chips, so that a single processor can "enjoy" the processor technology of the thread-level parallel computing. Multi-threading technology can enhance processors on multitasking, multi-threaded processing capabilities on multi-threaded operating systems and software.
Hyper-threading can make multiple threads of the operating system or application software, simultaneously run on a hyperthartic processor, shared a group of processor execution units, parallel completion, multiplier, etc. . This can make the processor's processing capability by 30%, because in the same time, the application can fully use the individual arithmeters of the chip.
For single-threaded chips, although thousands of instructions can be processed per second, at a certain moment, it can only process one instruction (single thread), and the result will inevitably make other processing units inside the processor. Idle. The "Ultra-Thread" technology can cause more instructions and data (multiple threads) to process the processor at a certain moment. It can be said that the ultra-thread is a technique that can temporarily temporarily process resources inside the CPU.
How is the ultra-thread work?
In the process of processing multiple threads, each logical processor inside the multi-thread processor can respond separately, and when the first logical processor traces a software thread, the second logical processor also begins. Track and process another software thread.
In addition, in order to avoid the CPU processing resource conflict, it is responsible for processing the logical processor of the second thread, which is used only the processing unit that is temporarily idle when the first thread is run. For example, when a logical processor performs a floating point operation (using the processor floating unit), another logical processor can perform an additional operation (using a processor's integer arithmetic unit). In this way, it is undoubtedly greatly improved the utilization rate of the processor internal processing unit and the corresponding data, and the instructions are throughput.
Five prerequisites for implementing hyperthreading
(1) Need CPU support
At present, the CPU that officially supports the ultra-thread technology has Pentium4 3.06GHz, 2.40C, 2.60C, 2.80C.
, 3.0GHz, 3.2GHz and Prescott processors, and some models of Xeon.
(2) Requires motherboard chipset support
The main model of the motherboard chipset that officially supports Hyper-Threading Technology includes Intel's 875P, E7205, 850E, 865PE / G / P, 845PE / GE / GV, 845G (B-Stepping), 845E. 875P, E7205, 865PE / G / P, 845PE / GE / GV chipset can normally support the use of hyper-thread technology, while the early 845E and 850E chipset can solve the support problem as long as the BIOS is upgraded. SIS645DX (B), SIS648 (B), SIS655, SIS658, SIS648FX. VIA has P4X400A, P4X600, P4X800.
(3) Requires motherboard BIOS support
Motherboard vendors must support ultra-threaded talents in the BIOS.
(4) Require operating system support
Currently, Microsoft's operating system is only Windows
XP Professional and subsequent versions support this feature, and the plan to implement the HMT support on Windows 2000 has been canceled.
(5) Require application software support
In general, as long as the software that can support multi-processors can support ultra-thread technology, it is actually not much software, and it is highly supported by professional software such as graphics, video processing, and the game software is very supportive. Application software has Office2000, Office XP, etc. The version after Linux Kernel 2.4.x is also supported by Hyperformance Technology.
Second, what is "dual channel" memory technology?
Dual-channel memory technology is to make two memory controllers in the North Bridge (also known as GMH) chipset, which can work independently of each other. On these two memory channels, the CPU can address, read data, so that the bandwidth of the memory can be doubled, and the data access speed is doubled accordingly (theoretically this).
At present, the popular dual-channel DDR memory architecture is built in two 64BitDDR memory controllers, and its bandwidth can reach 128bit, but the working mode is different from memory control technology of single-channel 128bit. Because two memory controllers of the dual-channel system are independent, two memory controllers can operate simultaneously in the case of zero waiting time between each other. For example, when the controller B is ready to perform the next access memory, the controller
A is reading / writing the main memory, and vice versa. This complementary "nature" of the two memory controllers allows effective wait time to reduce 50%, so that the bandwidth of the memory is turned over.
Two memory controllers in dual-channel DDR are functionally exactly the same, and the timing parameters of the two controllers can be set separately. Such flexibility allows the user to use two different constructs, capacity, and speed DIMM memory strips. At this time, the dual channel DDR simply adjusts to the lowest density to achieve a 128bit bandwidth, allowing DIMM memory strips for different density / waiting time characteristics. Reliable work together.
Briefly, dual-channel technology is a technology that is related to the motherboard chipset. It has nothing to do with memory itself. As long as the manufacturer integrates two memory controllers inside the chip, it can constitute a dual-channel DDR system. The motherboard manufacturers need to divide DIMM into Channel according to the memory channel.
1 with CHANEL
2. Users also need to be dual into the plug-in, just like RDRAM. If you only use a single memory, you will work only in the two memory controllers, and there is no double channel effect.
Dual-channel memory control technology can improve the memory bandwidth, especially those that require frequent exchange of data with memory and integrated chipsets with graphical cores (integrated graphics cards). At 865g, the high bandwidth brought by the dual-channel memory control technology, can help the integration of the graphics card to get a higher data bandwidth when dividing the main memory is displayed, and the data bandwidth of the memory It is the bottleneck that restricts a graphics card performance.
For the main board of the integrated graphic core, its memory is not only frequently converted with the CPU, but also the graphical core sharing of the integrated graphics on the motherboard. At this time, the memory will also change frequently, and this is undoubtedly a severe test for limited memory bandwidth.
Dual Channel Memory Control Technology is a motherboard chipset technology that only supports dual-channel memory control technology to architecture to architectural memory platform, and the Intel camp has I850.
I875P, I7205, I865PE, I865G, SIS655, SIS655FX, VIA PT600 (P4X600), VIA
PT800 (P4X800), VIA PT880 and other chipset, it is really good to be talented, and AMD camp has only nForce2
The NFORCE3 chipset is standing on the situation.
