cpu与gpu的作用和位置分析Role and position analysis of CPUand GPU
The effect of CPU: CPU as the core of a computer, it proved tobe irreplaceable, this is the past, today is still such, thefuture should be like this, but may be increased and endowedwith more complex function. Why is CPU able to control the coreof a computer and manage a complex and sophisticated computersystem with ease? Why can CPU be properly termed "brain" ofcomputers rather than other components?This is because CPU isdesigned primarily for the execution of operating systems,system software, scheduling and running a wide variety ofapplications, and for coordinating and controlling the entirecomputer system. CPU has a universal characteristic, which is"man" or "generalist", what should be, but this doesn't meanthat CPU each task has a top level.
The integration of millions, tens of millions or even hundredsof millions of transistors, CPU chip, in addition to computingpower circuit and structure, also has a central controller andthe other hardware circuit with the circuit structure andmodern CPU also has more thinking ability, such as logic,speculative execution, prediction of execution and so on. Onlywith these qualities will CPU be able to work as a "brain" forc omputers.
So what does CPU rely on for thought, command, and control?Theanswer is the instruction set. Instruction set is the set ofall CPU instruction processing, no instruction set chip couldnot be called CPU, but that is the CPU instruction set thinkinglanguage, CPU is the "smart", it is also a basic attribute
different from other chips. Similar to the human brain,involved in the thinking process of any man' s language, Chinesewith Chinese thinking, Americans English thinking, if used tospeak dialect, dialect people even think, people in thesubconscious instinct or conditions are most familiar withtheir own language thinking. The instruction set is the brainCPU language of the computer, and the CPU uses the instructionset to "think"".
The familiar x86 instruction set is the language of CPU thatmost of us use today, and the x86 instruction set was invented,developed, and continuously improved by Intel. All Intelarchitectures of CPU and CPU are compatible with the x86instruction set. Any program, written in any high-levelprogramming language, needs to be translated into x86instructions by a high-level language compiler or interpreterbefore it can be executed by CPU.
Such as C language, C++ language, Pascal language, advancedprogramming language is for programmers to use, people canexpress their thoughts and instruction through the high-levelprogramming language, can understand the instructions by thecompiler or interpreter into CPU, CPU can follow people"thinking and instruction" implementation of hundred-percent,strict in demands. In fact, compilers and interpreters areexecuted by CPU.
With the instruction system, CPU can control, command,coordinate and schedule the subsystems of the whole computersystem, so that they can cooperate with each other andaccomplish all kinds of tasks in an orderly way.
The role of GPU: in addition to CPU (central processing unit,also known as the central processing unit) , the computer systemalso has many PU - processing units, collectively referred toas xPU. Since they do not have the versatility of CPU, they areproprietary, and are customarily called controllers or chips.Such as memory controller, interrupt controller, Ethernet chip,USB controller and so on, although the term is not common, butwe can still put them into Memory PU, also called: InterruptPU memory processing unit -interrupt processing unit EthernetPU - USB-PU - USB Ethernet processing unit processing unitSo it' s no wonder that graphics chips, which are more graphiccomputing power, are called GPU, or graphics processing units.GPU has the characteristics of specificity and is good atgraphic calculation and processing.
The predecessor of GPU is the main chip of display card. Displaycards and displays, etc. , form a graphics subsystem in acomputer system.
The earliest display card function is relatively simple, soalso called display adapter card (referred to as video card) ,it is connected to the host and display interface card. Now thegraphics cards are 3D graphics acceleration card, the main chiphas also been dubbed GPU' s new name.
Today, there is no fundamental change in the main function ofthe video card, and its role is to transfer the outputinformation and instructions of CPU to characters, graphics andcolors, which are transmitted to the display. However, today' s
graphics cards are more capable of executing and computinggraphics when performing graphical instructions from CPU. Nowlet me explain to you.
Early graphics graphics processing ability is very weak,basically just to play the role of transfer CPU, if you wantto draw a simple graphics in graphic mode such as square, parkline graph, both need to position, each consisting of graphicsto display the size, color are all told graphics, graphics andfollow the prescribed order drawn on the display.
With the operating system and the application of complex andhigh quality graphics have become increasingly demanding,full-time CPU to do the graphics processing work hard, also lowefficiency but also The loss outweighs the gain. , cause CPU.Because CPU is designed to handle system tasks and programscheduling, not for graphics processing optimization.As a result, graphics acceleration has been given to newgraphics cards (now the main chip can be called GPU) , andgraphics cards that support 2D graphics acceleration haveemerged, greatly reducing the graphics processing pressure ofCPU. A 2D accelerated graphics card, CPU if you want to drawtwo-dimensional graphics, now only need to send a command tothe graphics card, such as "in the coordinates (x, y) to drawa length and width is a*B size of the rectangle, the graphicsgraphics accelerator (GPU) can draw with the size of theposition the graphics quickly on the display, after drawing GPUCPU notice, " I have finished, and then wait for the nextinstruction from CPU graphics.
In addition to the 2D graphics acceleration feature, the GPUnow has the capability to continuously accelerate 3D graphicsacceleration. Similarly, GPU is also the complicated 3Dgraphics processing work share from CPU over the CPU now as acommand, such as "draw a ball", to GPU, GPU to completethree-dimensional graphics rendering, and then notify the CPUfinish, waiting for the next instruction.
With a graphics accelerator, CPU is freed from the task ofgraphical processing, and can perform other more system tasks,thus improving the overall performance of the computer.However, not all 3D and graphics related operations at GPU isover, "task" transition is also gradually, for the calculationof GPU can not be achieved or not yet implemented or "work" tocomplete CPU. For example, the geometric coordinates of thegraph changes and the computation of the illumination model(T&L) are implemented in the mainstream GPU, which waspreviously done by CPU. For some of the current GPU integratedin the chipset, if there is no T&L acceleration, the T&Lcalculation is still implemented in software -- CPU. Inaddition, the coloring calculation of 3D graphics is graduallyshifted from CPU to GPU, such as vertex coloring and pixelcoloring (Vertex Shader & Pixel Shader) .
From the brief introduction of the above we can see that thefocus of the work of CPU and GPU are not the same, CPUresponsibility is much larger, with the whole computer system,to take care of all aspects, in addition to guarantee thehigh-speed operation of the whole system, but also to ensurethe stable operation of the system. Any mistake can be fatal,
so it' s hard for CPU to focus". It will often be interrupted,stop the work at hand, deal with normal or abnormal emergencytasks, otherwise the system will crash. Compared to CPU, GPU' sresponsibility to be lighter and more, if the graphicalcalculation is wrong, and run the program itself does not affectthe most, is displayed on the screen or the color graphicsdislocation chaos and so on, and is not responsible for GPUgraphics program results.
There is no GPU acceleration 2D and 3D' s, CPU swept and computergraphics all living and accelerated the "live" is not easy,today many of these jobs are GPU to do, CPU was liberated theCPU operation and control of valuable resources for more up thecore task execution system and the application of other non GPUclass. GPU is equivalent to a senior assistant at CPU withgraphics computing and processing expertise.
For non graphical tasks, CPU requires a high-level assistantfor other specialties. GPU' s contribution to the computersystem today is mainly the three-dimensional graphics ofhigh-end 3D games, and it is powerless for other broaderapplications. Such as system security encryption anddecryption, processing and processing of multimedia digitalcontent, virtual system, artificial intelligence and so on toonumerous to mention in the game.
In the future, I' ll detail some aspects of "CPU" and "GPU" aregood and not good at it".
This article is from BLOG @ Intel China. For more information,please click here
Brief introduction of the author
Zhao Jun graduated from the Department of computer science ofPeking University in June 1995 with a master' s degree incomputer science. He later taught at the Peking University' sDepartment of information management until June 1997.Zhao Jun joined Intel in July 1997 as a computer platformtechnical engineer. From 2000 to 2003, he served as a seniortechnical engineer in Intel China' s OEM division, and won theIntel achievement award in 2001".
Since October 2003 October ~2006, Zhao Jun has been in theleading technology sales division of the Intel marketing group,serving as engineering and technical manager and technicalmarketing manager. Since November 2006, he has been a seniorArchitecture Manager at the advanced technology support andservices division.
Company informat ion:
Intel Intel
已经有一段时间没有听到Gigsgigscloud服务商的信息,这不今天看到商家有新增一款国际版线路的美国VPS主机,年付也是比较便宜的只需要26美元。线路上是接入Cogentco、NTT、AN2YIX以及其他亚洲Peering。这款方案的VPS主机默认的配置是1Gbps带宽,比较神奇的需要等待手工人工开通激活,不是立即开通的。我们看看这款服务器在哪里选择看到套餐。内存CPUSSD流量价格购买地址1...
CloudCone商家我们很多喜欢低价便宜VPS主机的肯定是熟悉的,个人不是特别喜欢他。因为我之前测试过几次,开通的机器IP都是不通的,需要删除且开通好几次才能得到一个可用的IP地址。当然他们家的优势也是有的,就是价格确实便宜,而且还支持删除重新开通,而且机房只有一个洛杉矶MC。实话,如果他们家能多几个机房,保持现在的特点,还是有很多市场的。CloudCone是来自美国的主机销售商,成立于2017...
HostKvm是一家成立于2013年的国外VPS服务商,产品基于KVM架构,数据中心包括日本、新加坡、韩国、美国、俄罗斯、中国香港等多个地区机房,均为国内直连或优化线路,延迟较低,适合建站或者远程办公等。本月,商家旗下俄罗斯、新加坡、美国、香港等节点带宽进行了大幅度升级,俄罗斯机房国内电信/联通直连,CN2线路,150Mbps(原来30Mbps)带宽起,目前俄罗斯和香港高防节点5折骨折码继续优惠中...