Internal bus, which willprocessorAll structural units of are internally connected.Its width can be 8, 16, 32, 64, or 128 bits.
If inside the CPU,registerThe bus used for data transmission between ALU and control unit is called chipInternal bus(i.e. the bus inside the chip)[1]。
Internal bus technology
Announce
edit
Several popular internal bus technologies:
1. I2C bus
I2C (Inter IC) bus was introduced by Philips Company in 1982, which is a new type widely used in the field of microelectronic communication control in recent yearsBus standard。It isSynchronous communicationIt is a special form of, which has the advantages of fewer interface lines, simplified control mode, small device packaging form, and higher communication rate.In master slave communication, multiple I2C bus devices can be connected to the I2C bus at the same time, and the communication object can be identified by address.
2. SPI bus
Serial peripheral interfaceSPI（serial peripheral interface）Bus technologyIt is a product of MotorolaSynchronous serial interface。Most MCU produced by Motorola（micro controller）All are equipped with SPI hardware interface, such as 68 series MCU.SPI bus is a three wire synchronous bus. Because of its strong hardware function, the software related to SPI is quite simple, allowing the CPU to have more time to deal with other transactions.
3. SCI bus
serial communicationThe serial communication interface (SCI) is also introduced by Motorola.It is a universal asynchronous communication interface UART, which has basically the same asynchronous communication function as MCS-51[2]。
Internal bus development
Announce
edit
External frequency and FSB bus
fsb andFront end busThey are two different concepts. The reason why people are confused is that in the old Pentium era, the frequency values of the two are often the same.
andFront end busSpeed refers to CPU andBeiqiao chipThe speed of inter bus indicates the speed of data transmission between CPU and the outside world.
reason whyFront end busIt is easy to confuse the two concepts of external frequency. The main reason is that for a long time (mainly before and just after the appearance of Pentium 4),Front end bus frequencyIt is the same as the external frequency, so the front-end bus is often called the external frequency directly, which eventually leads to such a misunderstanding.
With the development of computer technology, people found thatFront end bus frequencyNeed to be higher thanfsb Therefore, DDR (Double Date Rate) technology and QDR (Quad Date Rate) technology are generated, making the frequency of the front-end bus twice that of the external frequency（AMDK7 ofprocessor）, 4x（IntelPentium processing ofCore processor）Since then, the difference between the front-end bus and the external frequency has begun to be valued.
“Front end busThe name "FSB" is defined byAMDWhen the K7 CPU was introduced, the front end bus speed refers to the speed of data transmission, because the maximum bandwidth of data transmission depends on the width and transmission frequency of all data transmitted at the same time, that is, the data bandwidth=(bus frequency X dataBit width）÷8。
Front end bus frequencyThe larger the data transmission volume between the CPU and the North Bridge, the better the function of the CPU.Conversely, the lowerFront end busIt is impossible to supply enough data to the CPU, which limits the performance of the CPU and becomes the bottleneck of the system.
HT Bus (Hyper Transport)
fromAMDK8 processing starts,AMDandIntelThe development of the two internal buses began to diverge,IntelContinue to use the Core 2 CPU of FSB, whileAMDThen developed the HT bus (Hyper Transport) countermeasuresIntel。
HT bus isAMDSpecially designed for K8 platformHigh speed serial busIts development history can be traced back to 1999, formerly known as "Lightning Data Transport".In July 2001, this technology was officially launched,AMDAlso rename it Hyper Transport.Subsequently, Broadcom, Cisco, Sun, NVIDIA, ALi, ATI, Apple, Transmet and many other enterprises decided to adopt this new modelBus technology, andAMDIt also takes this opportunity to form a Hyper Transport open alliance, so as to promote Hyper Transport to the industry.
First generation: HT'sworking frequency In the range of 200MHz - 800MHz, two-way 16 bit mode, the maximumbandwidthIt can reach 6.4GB/s.
The second generation: In February 2004, the Hyper Transport Technology Alliance officially released the HT2.0 specification. Due to the adoption of Dual data technology, the frequency has been successfully raised to 1.0GHz, 1.2GHz and 1.4GHz, with two-way 16bit modeBus bandwidthIt has been upgraded to 8.0GB/s, 9.6GB/s and 11.2GB/s.
The third generation: November 19, 2007,AMDThe HT3.0 bus specification was officially released, providing 1.8GHz, 2.0GHz, 2.4GHz and 2.6GHz frequencies, which can support up to 32 channels.Under 32-bit channel, two-waybandwidthUp to 41.6GB/s.
QPI bus
becauseAMDThe maximum bandwidth provided by HT3.0 far exceedsIntel1600 FSB bandwidth. To combat HT 3.0,IntelIn another way, the QPI bus is proposed.
As we calculated earlier, the 1600FSB can provide 12.8G/s bandwidth, but such a high bandwidth can only meet the bandwidth requirements of DDR2 800 dual channel memory (800 X 64 X 2/8=12.8G/s). If it is combined with 1066 or even higher 1333 memory, the FSB needs to be increased to a higher frequency, not to mentionPCI busPCI-E bus, USB, SATA and other devices also need to occupy a certain bandwidth.Under the current production process and framework, it becomes even more difficult to increase the frequency. Even if some players increase the FSB to 2400, the heat generated is also terrible.
along withProcessor coreWith the improvement of performance and the rapid growth of the number of cores, FSB is becoming a bottleneck and must be solved.IntelIf you want toMulti coreThe era is in an invincible position. The first problem is to successfully solve the problem of system resource allocation and give full play to the advantages of multi-core. This is Intel's launch of QPIBus technologyThe ultimate purpose of.
The biggest improvement of QPI is that it provides amazing output transmission capability, ranging from 4.8 to 6.4 GT/s.A connectedBit widthIt can be 5, 10 or 20 bits.Therefore, the QPI full width link in each direction can provide a bandwidth of 12 to 16BG/s. Then the bandwidth of each QPI link is 24 to 32GB/s, equivalent to two to three times the 1600FSB bandwidth, which is basically the same as the HT 3.0 bandwidth.
In addition, another highlight of QPI is that it supports multiplesystem bus connect,IntelIt is called multi FSB.system bus It will be divided into multiple connections, and the frequency is no longer single and fixed, and there is no need to connect through FSB as before.
Compared with FSB, QPI bus is of great revolutionary significance, which has brought innovation in PC manufacturing structure and abandoned the previous concept of North Bridge and South Bridge[3]。
