CPU graphic production process - to intel CORE i7, for example, look to CPU architecture

Science Sciences REVIEW: graphic CPU whole production process - to intel CORE i7, for example, look to CPU architecture. This article Intel Intel x86 architecture CPU manufacturing raw materials and preparation, CPU manufacturing process, and the prospect of CPU x86 architecture and RISC architecture. Wherein Refresh "intel Core i7 CPU illustrates the whole production process," increased production step described nano-level data. Keywords: CPU, x86, RISC-V, the Turing Award, Patterson, Tsinghua - Berkeley International Laboratory. No public input field development "CPU architecture production" Get this PDF document; welcome to appreciate the support of science, learning download scientific and technological knowledge.

table of Contents

CPU graphic production process - to intel CORE i7, for example, look to CPU architecture (11k words)

First, Intel architecture Intelx86

Second, the basic raw material for producing the CPU

Three, CPU manufacturing preparation stage

Four, intelCore i7 production process diagram

Fifth, the Intel X86 outlook

 

Graphic CPU production process - to intel CORE i7 , for example, look to CPU architecture

Wen | Qin Long Ji et al., Science Sciences © 20190617Mon

 

The central processing unit is the core component --CPU modern computers, also known as "microprocessors." A measure of computer performance, depending on CPU manufacturing specifications and operational control frequency. And manufacturing computer, computer system architecture design first, reproduced it. After the completion of the hardware manufacturing, but also the operating system and applications, human beings can use the computer - the most powerful in the history of information processing tools. Outside the territory, to strictly determine the protection of intellectual property, otherwise unsustainable. This article CPU architecture and the production of knowledge.

 

First, Intel Intel x86 architecture

From the Intel 8086, the 286,386,486,586, P1, P2, P3, P4, i3, i5, i7, i9 are using the same CPU architecture, collectively X86. (INTEL and AMD's current processors are basically the X86 architecture .IBM POWER series and the ARM architecture X86 is not the company's products.) From June 8, 1978, the US Intel (Intel) has released new 16-bit microprocessor " 8086 ", counting the times to create the x86 architecture, Intel x86 architecture has more than forty years, is the world's design, production, manufacture one of the most CPU system.

FIG 2: Intel 8086 processor

X86 architecture (The X86 architecture) is a computer language instruction set executed by the microprocessor, it means a standard code abbreviation intel series general purpose computer, a general-purpose computer also identifies a set of instructions. x86 refers to some specific microprocessor instruction set computer language execution, defines the basic rules of the use of the chip, as today's x64, IA64 and so on.

Although 8086 was released did not get much attention, has not been adopted widely, but it is in the PC industry to bring the x86 architecture, Intel heyday of the achievements of the position, has become an industry standard, even in today's powerful on multi-core x86 processors can see the shadow. In the history of 40 years of development, x86 family has grown, he moved notebooks, servers, supercomputers from the desktop, but also a lot of frustration or restrict the development of competitors' writing device, during which so many processor architectures technology vendors and their names into history , even though some are also closed unsustainable development, such as Apple has abandoned the PowerPC.

Of course, we can not forget the struggle of x86-64 and EM64T. In 2003, AMD introduced the industry's first 64-bit processor Athlon 64, also brought x86-64, that is, 64-bit x86 instruction set extensions superset, with a downward-compatible features. Intel was also in the implementation of 64-bit technology, but it is not compatible with IA64 architecture x86, only used on the server processor Itanium. In order to compete with AMD, Intel also in 2004 launched its own 64-bit version of x86, which is EM64T.

In this regard, AMD and Intel accuse each other, but in any case at least 64 to promote the development and popularization of technology, but also to x86 technology to continue to flourish. UC Berkeley professor of computer science, one of RISC inventor David · Patterson (David Patterson) , said: "This proves resilient x86 instruction set can be used to fully deal with Intel, even though Intel dominated the market, other the company can still change the direction of x86. "

intel x86 is a general-purpose computer series of standard number abbreviation, also identifies a common set of computer instructions, X has nothing to do with the processor, it is a simple wildcard definitions for all * 86 system, for example: i386,586, Pentium (pentium). Since the early intel's CPU is numbered as 8086,80286, the entire series of CPU instructions are compatible, so are used to identify the X86 instruction set used today's Pentium, P2, P4, Celeron series are supported by X86 instruction set, so the X86 family belong.

US Intel X86 instruction set for its first piece of 16-bit CPU (i8086) specially developed for the United States, IBM in 1981 launched the world's first PC in the CPU - is the X86 i8088 (i8086 Starter Edition) to use commands, while the computer in order to improve the floating point data processing capacity increased X87 math coprocessor chip series are additionally used X87 instructions, X86 instruction set and will later X87 instruction set referred to as X86 instruction set. Although with the development of CPU technology, Intel have developed a newer i80386, i80486 until today's Pentium 4 (hereinafter as P4) series, but in order to ensure that the computer can continue to run all kinds of applications developed in the past to protect and inherit a wealth of software resources, the company produced all Intel CPU still continue to use the X86 instruction set, so it's still a CPU X86 series.

Also in addition to Intel Corp., AMD and Cyrix and other manufacturers have also produced can use the X86 instruction set of CPU, because the CPU can run all these various software developed by Intel CPU, so the industry will these computer CPU column Intel's CPU is compatible products. Since the Intel X86 series and compatible CPU use the X86 instruction set, so today the formation of a large series and compatible with X86 CPU lineup. Of course, not all use the X86 series CPU in a desktop (portable) computers, some servers and Apple (Macintosh) machine also uses American DIGITAL (digital) the company's Alpha 61164 and PowerPC 604e series CPU.

x86 architecture CPU for most of us work and live quite far-reaching impact. Most know little about computer knowledge for many friends inside the CPU will know the most important thing is the transistor, and increase CPU speed, the most important thing is how to put it plainly to more and more transistors on the same CPU inside the area, due to the CPU is too small, too precise, which make up a considerable number of transistors, so the staff is absolutely impossible, can only be processed by a photolithography process.

This is why a number of CPU inside Why can so many transistors. In fact, the transistor switch is a two-position: i.e., on and off. If you recall the basic computing era, that is a computer needs to perform all the work. Two options, on and off, and for the machine is i.e. 0 1. So how will you make a CPU? Here, step by step about the central processor from a pile of sand into a powerful whole process of integrated circuit chips.

 

Second, manufacturing CPU basestocks

If asked what the CPU is the raw material, it will easily give the answer - is silicon. This is not false, but the silicon where they come from? In fact, those sand is the most obscure. It is hard to imagine, expensive, complex, powerful, full of mystery from the CPU actually it is simply worthless sand. Of course, this intermediate must go through a complex manufacturing process for the job. But not just grab a handful of sand can do the raw materials must be carefully selected to extract the most pure silicon raw material for the job. Imagine if with that the most inexpensive and abundant reserves of raw materials made CPU, so the quality of the finished product will be, you can spend as they are now high-performance processor it?

Than silicon is removed, an important need for producing CPU material is a metal. So far, aluminum has become a major processor of metal material interior fittings, while copper is being phased out, this is for several reasons, in the current CPU voltage, electro-migration properties of aluminum significantly better than copper. The so-called electro-migration problem, refers to when a large number of electrons to flow through the conductor section of the conductor material atom by electron impact and away from its position, leaving a vacancy, the vacancy will lead to excessive conductive lead is disconnected, and the dislodged atoms stay elsewhere, it will cause a short circuit affecting other parts of the logic function of the chip, leading to the chip can not be used.

In addition to these two primary material of the chip design process also requires some kind of chemical raw materials, their different roles, is not repeated here.

 

Three, the CPU manufacturing preparation stage

After the necessary work is completed the acquisition of raw materials, these raw materials in some part of the required preprocessing. As the main raw material, silicon processing is essential. First, the raw material silicon to chemical purification, this step to reach the level of the raw material used for the semiconductor industry. In order to meet these raw silicon integrated circuit fabrication processing needs, it must also be shaped, This is done by melting the silicon raw material, liquid silicone is then injected into a large high quartz container accomplished.

Then, the raw materials for high-temperature melt. High school chemistry class we learned too, are many of atoms within the crystal structure of solid silicon is true. In order to meet the requirements of high-performance processors, the entire material must be highly pure silicon, and single crystal silicon. And then stretched using a rotary manner from the high temperature vessel will be removed silicon feedstock, at this time a cylindrical silicon ingot is produced. From the point of view of the process currently used, the diameter of the circular cross-section ingot of 200 mm.

But now intel and other companies have started a 300 mm diameter silicon ingot of. Increased cross-section in the case of silicon ingots of various properties to remain unchanged area is quite difficult, but as long as companies are willing to invest large sums of money to study, or can be achieved. intel for the development and production of 300 mm silicon ingot factory established cost about $ 3.5 billion, the success of new technology makes possible manufacturing intel higher complexity, more powerful integrated circuit chips. And 200 mm silicon ingot factories spent $ 1.5 billion.

制成硅锭并确保其是一个绝对的圆柱体之后,下一个步骤就是将这个圆柱体硅锭切片,切片越薄,用料越省,自然可以生产的处理器芯片就更多。切片还要镜面精加工的处理来确保表面绝对光滑,之后检查是否有扭曲或其它问题。这一步质量检验尤为重要,它直接决定了成品CPU的质量。

新的切片中要掺入一些物质而使之成为真正的半导体材料,而后在其上刻划代表着各种逻辑功能的晶体管电路。掺入的物质原子进入硅原子之间的空隙,彼此之间发生原子力的作用,从而使得硅原料具有半导体的特性。今天的半导体制造多选择CMOS工艺(互补型金属氧化物半导体)。

其中互补一词表示半导体中N型MOS管和P型MOS管之间的交互作用。而N和P在电子工艺中分别代表负极和正极。多数情况下,切片被掺入化学物质而形成P型衬底,在其上刻划的逻辑电路要遵循nMOS电路的特性来设计,这种类型的晶体管空间利用率更高也更加节能。同时在多数情况下,必须尽量限制pMOS型晶体管的出现,因为在制造过程的后期,需要将N型材料植入P型衬底当中,而这一过程会导致pMOS管的形成。

在掺入化学物质的工作完成之后,标准的切片就完成了。然后将每一个切片放入高温炉中加热,通过控制加温时间而使得切片表面生成一层二氧化硅膜。通过密切监测温度,空气成分和加温时间,该二氧化硅层的厚度是可以控制的。

在intel的90纳米制造工艺中,门氧化物的宽度小到了惊人的5个原子厚度。这一层门电路也是晶体管门电路的一部分,晶体管门电路的作用是控制其间电子的流动,通过对门电压的控制,电子的流动被严格控制,而不论输入输出端口电压的大小。

准备工作的最后一道工序是在二氧化硅层上覆盖一个感光层。这一层物质用于同一层中的其它控制应用。这层物质在干燥时具有很好的感光效果,而且在光刻蚀过程结束之后,能够通过化学方法将其溶解并除去。

3.1光刻蚀

这是目前的CPU制造过程当中工艺非常复杂的一个步骤,为什么这么说呢?光刻蚀过程就是使用一定波长的光在感光层中刻出相应的刻痕,由此改变该处材料的化学特性。这项技术对于所用光的波长要求极为严格,需要使用短波长的紫外线和大曲率的透镜。刻蚀过程还会受到晶圆上的污点的影响。每一步刻蚀都是一个复杂而精细的过程。

设计每一步过程的所需要的数据量都可以用10GB的单位来计量,而且制造每块处理器所需要的刻蚀步骤都超过20步(每一步进行一层刻蚀)。而且每一层刻蚀的图纸如果放大许多倍的话,可以和整个纽约市外加郊区范围的地图相比,甚至还要复杂,试想一下,把整个纽约地图缩小到实际面积大小只有100个平方毫米的芯片上,那么这个芯片的结构有多么复杂,可想而知了吧。

当这些刻蚀工作全部完成之后,晶圆被翻转过来。短波长光线透过石英模板上镂空的刻痕照射到晶圆的感光层上,然后撤掉光线和模板。通过化学方法除去暴露在外边的感光层物质,而二氧化硅马上在陋空位置的下方生成。

 

3.2掺杂

在残留的感光层物质被去除之后,剩下的就是充满的沟壑的二氧化硅层以及暴露出来的在该层下方的硅层。这一步之后,另一个二氧化硅层制作完成。然后,加入另一个带有感光层的多晶硅层。多晶硅是门电路的另一种类型。由于此处使用到了金属原料(因此称作金属氧化物半导体),多晶硅允许在晶体管队列端口电压起作用之前建立门电路。感光层同时还要被短波长光线透过掩模刻蚀。再经过一部刻蚀,所需的全部门电路就已经基本成型了。然后,要对暴露在外的硅层通过化学方式进行离子轰击,此处的目的是生成N沟道或P沟道。这个掺杂过程创建了全部的晶体管及彼此间的电路连接,没个晶体管都有输入端和输出端,两端之间被称作端口。

3.3重复这一过程

从这一步起,你将持续添加层级,加入一个二氧化硅层,然后光刻一次。重复这些步骤,然后就出现了一个多层立体架构,这就是你目前使用的处理器的萌芽状态了。在每层之间采用金属涂膜的技术进行层间的导电连接。今天的P4处理器采用了7层金属连接,而Athlon64使用了9层,所使用的层数取决于最初的版图设计,并不直接代表着最终产品的性能差异。

3.4测试封装测试过程

接下来的几个星期就需要对晶圆进行一关接一关的测试,包括检测晶圆的电学特性,看是否有逻辑错误,如果有,是在哪一层出现的等等。而后,晶圆上每一个出现问题的芯片单元将被单独测试来确定该芯片有否特殊加工需要。

而后,整片的晶圆被切割成一个个独立的处理器芯片单元。在最初测试中,那些检测不合格的单元将被遗弃。这些被切割下来的芯片单元将被采用某种方式进行封装,这样它就可以顺利的插入某种接口规格的主板了。大多数intel和AMD的处理器都会被覆盖一个散热层。

在处理器成品完成之后,还要进行全方位的芯片功能检测。这一部会产生不同等级的产品,一些芯片的运行频率相对较高,于是打上高频率产品的名称和编号,而那些运行频率相对较低的芯片则加以改造,打上其它的低频率型号。这就是不同市场定位的处理器。而还有一些处理器可能在芯片功能上有一些不足之处。比如它在缓存功能上有缺陷(这种缺陷足以导致绝大多数的CPU瘫痪),那么它们就会被屏蔽掉一些缓存容量,降低了性能,当然也就降低了产品的售价,这就是Celeron和Sempron的由来。

当CPU被放进包装盒之前,一般还要进行最后一次测试,以确保之前的工作准确无误。根据前面确定的最高运行频率不同,它们被放进不同的包装,销往世界各地。

读到这,相信你已经对CPU制造流程有了一些比较深入的认识。CPU的制造,可以说是集多方面尖端科学技术之大成,CPU本身也就那么点大,如果把里面的材料分开拿出来卖,恐怕卖不了几个钱。然而CPU的制造成本是非常惊人的,从这里或许我们可以理解,为什么这东西卖这么贵了。

在测试这个环节很重要,比如你的处理器是6300还是6400就会在这个环节被划分,而6300天生并不是6300,而是在测试之后,发现处理器不能稳定的在6400标准下工作,只能在6300标准下稳定工作,于是对处理器定义,锁频,定义 ID,封装,印上6300。

我们用AMD的来举例:同样核心的处理器都是一个生产线下来的,如果稳定工作在2.8GHz,1M*2的缓存下,就被定义为5600+,如果缓存有瑕疵,切割有问题的那一半,成为5400+,如果缓存没问题而频率只能在2.6G通过测试,那么就是5200+,如果缓存有瑕疵,就切割成为5000+…………一直把它测到3800+,如果还不稳定,要么想办法变成速龙64单核或者单核闪龙,或者就是出现过的ES版的双核闪龙,如果出现批量不能工作在3800+条件下,而工作在3600+条件下,那么3600+就上市了,如果出现批量能工作在3G,1M*2条件下,那么6000+就上市了,这就是为什么处理器总是中等型号的先上市,高端和底端的后上市,当然后期工厂可能会节约成本专门开出底端的流水线,专门生产底端处理器,赛扬,闪龙的各种型号就相继上市,而高端的流水线因为个别处理器不稳定转变为底端处理器,例如将速龙64缓存切割就变为闪龙64。

 

四、intel Core i7生产全过程图解

沙子:硅是地壳内第二丰富的元素,而脱氧后的沙子(尤其是石英)最多包含25%的硅元素,以二氧化硅(SiO2)的形式存在,这也是半导体制造产业的基础。

4.1硅熔炼:

12英寸/300毫米晶圆级,下同。通过多步净化得到可用于半导体制造质量的硅,学名电子级硅(EGS),平均每一百万个硅原子中最多只有一个杂质原子。此图展示了是如何通过硅净化熔炼得到大晶体的,最后得到的就是硅锭(Ingot)。

 

4.2单晶硅锭:

整体基本呈圆柱形,重约100千克,硅纯度 99.9999%。

4.3硅锭切割:

横向切割成圆形的单个硅片,也就是我们常说的晶圆 (Wafer)。顺便说,这下知道为什么晶圆都是圆形的了吧?

4.4晶圆:

切割出的晶圆经过抛光后变得几乎完美无瑕,表面甚至可以当镜子。事实上,Intel自己并不生产这种晶圆,而是从第三方半导体企业那里直接购买成品,然后利用自己的生产线进一步加工,比如现在主流的45nm HKMG(高K金属栅极)。值得一提的是,Intel公司创立之初使用的晶圆尺寸只有2英寸/50毫米

4.5光刻胶(PhotoResist)

图中蓝色部分就是在晶圆旋转过程中浇上去的光刻胶液体,类似制作传统胶片的那种。晶圆旋转可以让光刻胶铺的非常薄、非常平。

4.6光刻:

光刻胶层随后透过掩模(Mask)被曝光在紫外线(UV)之下,变得可溶,期间发生的化学反应类似按下机械相机快门那一刻胶片的变化。掩模上印着预先设计好的电路图案,紫外线透过它照在光刻胶层上,就会形成微处理器的每一层电路图案。一般来说,在晶圆上得到的电路图案是掩模上图案的四分之一。

 

4.7光刻:

由此进入50-200纳米尺寸的晶体管级别。一块晶圆上可以切割出数百个处理器,不过从这里开始把视野缩小到其中一个上,展示如何制作晶体管等部件。晶体管相当于开关,控制着电流的方向。现在的晶体管已经如此之小,一个针头上就能放下大约3000万个。

4.8溶解光刻胶:

光刻过程中曝光在紫外线下的光刻胶被溶解掉,清除后留下的图案和掩模上的一致

4.9蚀刻:

使用化学物质溶解掉暴露出来的晶圆部分,而剩下的光刻胶保护着不应该蚀刻的部分。

4.10清除光刻胶:

蚀刻完成后,光刻胶的使命宣告完成,全部清除后就可以看到设计好的电路图案。

 

4.11光刻胶:

再次浇上光刻胶(蓝色部分),然后光刻,并洗掉曝光的部分,剩下的光刻胶还是用来保护不会离子注入的那部分材料。

 

离子注入(Ion Implantation):在真空系统中,用经过加速的、要掺杂的原子的离子照射(注入)固体材料,从而在被注入的区域形成特殊的注入层,并改变这些区域的硅的导电性。经过电场加速后,注入的离子流的速度可以超过30万千米每小时。

 

4.12清除光刻胶:

离子注入完成后,光刻胶也被清除,而注入区域(绿色部分)也已掺杂,注入了不同的原子。注意这时候的绿色和之前已经有所不同。

 

4.13晶体管就绪:

至此,晶体管已经基本完成。在绝缘材(品红色)上蚀刻出三个孔洞,并填充铜,以便和其它晶体管互连。

 

4.14电镀:

在晶圆上电镀一层硫酸铜,将铜离子沉淀到晶体管上。铜离子会从正极(阳极)走向负极(阴极)。

 

4.15铜层:

电镀完成后,铜离子沉积在晶圆表面,形成一个薄薄的铜层。

 

4.16抛光:

将多余的铜抛光掉,也就是磨光晶圆表面。

 

晶体管级别,六个晶体管的组合,大约500纳米。在不同晶体管之间形成复合互连金属层,具体布局取决于相应处理器所需要的不同功能性。芯片表面看起来异常平滑,但事实上可能包含20多层复杂的电路,放大之后可以看到极其复杂的电路网络,形如未来派的多层高速公路系统。

 

4.17晶圆测试:

内核级别,大约10毫米/0.5英寸。图中是晶圆的局部,正在接受第一次功能性测试,使用参考电路图案和每一块芯片进行对比。

 

4.18晶圆切片(Slicing)

晶圆级别,300毫米/12英寸。将晶圆切割成块,每一块就是一个处理器的内核(Die)。

 

晶圆级别。测试过程中发现的有瑕疵的内核被抛弃,留下完好的准备进入下一步。

 

4.19单个内核:

内核级别。从晶圆上切割下来的单个内核,这里展示的是Core i7的核心。

封装:

封装级别,20毫米/1英寸。衬底(基片)、内核、散热片堆叠在一起,就形成了我们看到的处理器的样子。衬底(绿色)相当于一个底座,并为处理器内核提供电气与机械界面,便于与PC系统的其它部分交互。散热片(银色)就是负责内核散热的了。

 

4.20等级测试:

最后一次测试,可以鉴别出每一颗处理器的关键特性,比如最高频率、功耗、发热量等,并决定处理器的等级,比如适合做成最高端的Core i7-975 Extreme,还是低端型号Core i7-920。

 

4.21装箱:

 

制造、测试完毕的处理器要么批量交付给OEM厂商,要么放在包装盒里进入零售市场。

 

五、英特尔X86展望

英特尔推出X86架构已满40年了,同486相比,Pentium向前迈进了一大步,而PⅡ的前进步伐则没有这么大了,X86 CPU的发展似乎已到了尽头。英特尔非常清楚,是X86指令集限制了CPU性能的进一步提高,因此,他们正同惠普共同努力开发下一代指令集架构(Instruction Set Architecture,ISA): EPIC(Explicitly ParallelInstruction Computing,显性并行指令计算)。对英特尔而言,IA-64(英特尔的64位架构)是下一个10到15年的架构。新的ISA将使英特尔摆脱X86架构的限制,从而设计出超越所有现有RISC CPU和X86 CPU的新型处理器。

 

那么EPIC的先进之处在什么地方呢?为什么英特尔会放弃使它成为芯片巨人的X86架构呢?

IA32的问题:工程师可以通过提高每个时钟的指令执行数来提高性能,英特尔新的指令集的首要目的在于,让指令更容易解码,更容易并行执行。这样就可以不受限制地开发新型处理器。但是,对工程师而言,兼容8086的X86指令集一直是必须完成的任务。毕竟,兼容前代产品是使英特尔成长壮大起来的关键因素,而且还可以保护用户原先的投资和使用数以百万计应用软件。

既然如此,为什么又要放弃整个X86指令集重新开始呢?X86的不足在什么地方?

(1)可变的指令长度 X86指令的长度是不定的,而且有几种不同的格式,结果造成X86 CPU的解码工作非常复杂,为了提高CPU的工作频率,不得不延长CPU中的流水线,而过长的流水线在分支预测出错的情况下,又会带来CPU工作停滞时间较长的弊端。

(2)寄存器的贫乏 X86指令集架构只有8个通用寄存器,而且实际只能使用6个。这种情况同现代的超标量CPU极不适应,虽然工程师们采用寄存器重命名的技术来弥补这个缺陷,但造成了CPU过于复杂,流水线过长的局面。

(3)内存访问 X86指令可访问内存地址,而现代RISC CPU则使用LOAD/STORE模式,只有LOAD和STORE指令才能从内存中读取数据到寄存器,所有其他指令只对寄存器中的操作数计算。在CPU的速度是内存速度的5倍或5倍以上的情况下,后一种工作模式才是正途。

(4)浮点堆栈 X87 FPU是目前最慢的FPU,主要的原因之一就在于X87指令使用一个操作数堆栈。如果没有足够多的寄存器进行计算,你就不得不使用堆栈来存放数据,这会浪费大量的时间来使用FXCH指令(即把正确的数据放到堆栈的顶部)。

(5)4GB限制这似乎不是问题,但是,在2000年前,主流PC只有4MB内存,现在绝大部分PC装备了2G以上的内存,是以前的512倍,所以,PC内存突破16GB绝对不会令人惊讶,大型服务器已经使用了32GB以上的内存,突破64GB内存的情况已经出现。

(6)芯片变大所有用于提高X86 CPU性能的方法,如寄存器重命名、巨大的缓冲器、乱序执行、分支预测、X86指令转化等等,都使CPU的芯片面积变得更大,也限制了工作频率的进一步提高,而额外集成的这些晶体管都只是为了解决X86指令的问题。

 

1,3. 最近更新:xs0kiss(2019-06-12). X86架构. [EB/OL], baike, https://baike.baidu.com/item/X86%E6%9E%B6%E6%9E%84,2019-06-12, visit date: 2019-06-17

2. 搜狐>汽车>正文. 图解 intel Core i7 CPU生产全过程. [EB/OL], sohu, http://www.sohu.com/a/320900924_465219, 2019-06-1610:00:00, visit date: 2019-06-17

4. 清华大学. 图灵奖得主依托清华-伯克利深圳学院建设RISC-V国际开源实验室. [EB/OL], sina, http://edu.sina.com.cn/l/2019-06-14/doc-ihvhiews8841336.shtml,2019-06-14 14:36, visit date: 2019-06-17

5. 秦陇纪. 人工智能起源与发展正史. [EB/OL], 科学Sciences. http://weixin.qq.com/,2019-06-06, visit date: 2019-06-07

x. 秦陇纪. 西方哲学与人工智能、计算机; 数据科学与大数据技术专业概论; 人工智能研究现状及教育应用; 数据资源概论; 文本数据溯源与简化; 大数据简化技术体系; 数据简化社区概述. [EB/OL], 数据简化DataSimp(微信公众号),https://dsc.datasimp.org/, http://www.datasimp.org, 2017-06-06

Guess you like

Origin blog.csdn.net/weberhuangxingbo/article/details/92709285