A few years ago, Intel introduced a step-by-step microprocessor manufacturing process, from sand to final product. In fact, the process of manufacturing semiconductor elements looks truly amazing.
Step 1. Sand
Silicon, which makes up about 25 percent of all chemical elements in the earth's crust by total mass, is the second most abundant after oxygen. The sand has a high percentage of silicon dioxide (SiO 2 ), which is the main ingredient not only for the production of Intel processors, but also for semiconductor production in general.
molten silicon
The substance is purified over several steps until a semiconductor-grade silicon is obtained, used in semiconductors. Ultimately, it comes in the form of single-crystal ingots about 300 millimeters (12 inches) in diameter. Previously, ingots had a diameter of 200 millimeters (8 inches), and in the distant 1970 - even less - 50 millimeters (2 inches).
At this level of processor production, after cleaning, the purity of the crystal is one impurity atom per billion silicon atoms. The weight of the ingot is 100 kilograms.
Step 3. Cutting an ingot
The ingot is cut with a very thin saw into individual slices called substrates. Each of them is subsequently polished to obtain a flawless mirror-smooth surface. It is on this smooth surface that tiny copper wires will subsequently be applied.
Exposure of the photoresist layer
A photoresistive liquid (the same materials used in traditional photography) is poured onto a substrate rotating at high speed. During rotation, a thin and uniform resistive layer is formed on the entire surface of the substrate.
An ultraviolet laser through masks and a lens acts on the surface of the substrate, forming small illuminated ultraviolet lines on it. The lens makes the focused image 4 times smaller than the mask. Wherever ultraviolet lines hit the resistive layer, a chemical reaction occurs, as a result of which these areas become soluble.
Step 5 Etching
The soluble photoresist material is then completely dissolved with a chemical solvent. Thus, a chemical etchant is used to partially dissolve or etch a small amount of polished semiconductor material (substrate). The rest of the photoresist material is removed by a similar washing process, exposing (exposing) the etched surface of the substrate.
Layer formation
To create the tiny copper wires that will eventually carry electricity to/from the various connectors, additional photoresists (light-sensitive materials) are added, which are also washed and exposed. Subsequently, an ion doping process is performed to add impurities and protect the copper ion deposits from copper sulphate during the electroplating process.
At various stages of these processor manufacturing processes, additional materials are added that are etched and polished. This process is repeated 6 times to form 6 layers.
The final product looks like a grid of many microscopic copper strips that conduct electricity. Some of them are connected to others, and some are located at a certain distance from others. But they are all used for the same purpose - to transfer electrons. In other words, they are designed to provide so-called "useful work" (for example, adding two numbers as fast as possible, which is the essence of the computing model these days).
The multilevel processing is repeated on each individual small area of the substrate surface on which the chips will be fabricated. Including such areas include those that are partially located outside the substrate.
Step 7. Testing
As soon as all metal layers are applied and all transistors are created, it's time for the next stage in the production of Intel processors - testing. A multi-pin device is placed at the top of the chip. A lot of microscopic wires are attached to it. Each such wiring has an electrical connection to the chip.
To reproduce the operation of the chip, a sequence of test signals is transmitted to it. The test not only tests traditional computing capabilities, but also performs internal diagnostics for voltage values, cascade sequences, and other functions. The response of the chip in the form of a test result is stored in a database specially allocated for a given section of the substrate. This process is repeated for each section of the substrate.
Insert cutting
A very small diamond-tipped saw is used to cut the plates. The database populated in the previous step is used to determine which chips cut from the substrate are retained and which are discarded.
Step 9 Enclosure
All workplates are placed in physical cases. Although the platters have been pre-tested and determined to work correctly, this does not mean that they are good processors.
The encapsulation process means placing a silicon chip in a substrate material with miniature gold wires attached to its contacts or ball array. An array of ball leads can be found on the back of the case. A heat sink is installed in the upper part of the case. It is a metal case. At the end of this process, the central processing unit looks like a finished product intended for consumption.
Note: The metal heat sink is a key component of today's high speed semiconductor devices. Previously, heatsinks were ceramic and did not use forced cooling. It was required for some 8086 and 80286 models and for models starting with 80386. Previous generations of processors had much fewer transistors.
For example, the 8086 processor had 29,000 transistors, while modern CPUs have hundreds of millions of transistors. Such a small number of transistors by today's standards did not generate enough heat to require active cooling. To separate these processors from those in need of this type of cooling, ceramic chips were subsequently labeled "Heat Sink Required".
Modern processors generate enough heat to melt in seconds. Only the presence of a heat sink connected to a large heatsink and fan allows them to function for a long time.
Sort processors by characteristics
By this stage of production, the processor looks like it is bought in a store. However, one more step is required to complete its production process. It's called sorting.
In this step, the actual characteristics of the individual CPU are measured. Parameters such as voltage, frequency, performance, heat dissipation and other characteristics are measured.
The best chips are shelved as higher-end items. They are sold not only as the fastest components, but also as low and extra low voltage models.
Chips that don't make it into the top processor group are often sold as processors with lower clock speeds. In addition, lower-end quad-core processors may be sold as dual- or triple-core.
Processor performance
The sorting process determines the final values of speed, voltage and thermal characteristics. For example, on a standard substrate, only 5% of manufactured chips can operate at frequencies above 3.2 GHz. At the same time, 50% of the chips can operate at a frequency of 2.8 GHz.
Processor manufacturers are constantly investigating the reasons why most of their processors are running at 2.8 GHz instead of the required 3.2 GHz. Occasionally, changes may be made to the processor design to improve performance.
Profitability of production
The profitability of the business for the production of processors and most semiconductor elements lies in the range of 33-50%. This means that at least 1/3 to 1/2 of the plates on each substrate are working, and the company is profitable in this case.
Intel has an operating profit of 95% using 45nm technology on a 300mm wafer. This means that if 500 silicon wafers can be made from a single substrate, 475 of them will be operational and only 25 will be discarded. The more plates can be obtained from one substrate, the more profit the company will have.
Intel technologies used today
The history of the application of new Intel technologies for mass production of processors:
- 1999 - 180 nm;
- 2001 - 130 nm;
- 2003 - 90 nm;
- 2005 - 65 nm;
- 2007 - 45 nm;
- 2009 - 32 nm;
- 2011 - 22 nm;
- 2014 - 14 nm;
- 2019 - 10 nm (planned).
At the beginning of 2018, Intel announced the postponement of mass production of 10nm processors to 2019. The reason for this is the high cost of production. At the moment, the company continues to ship 10nm processors in small volumes.
Let's characterize technologies of production of processors of Intel from the point of view of cost. The high cost of the company's management explains the long production cycle and the use of a large number of masks. The 10nm technology is based on deep ultraviolet lithography (DUV) using lasers operating at a wavelength of 193nm.
For the 7nm process, extreme ultraviolet lithography (EUV) will be used using lasers operating at a wavelength of 13.5nm. Thanks to this wavelength, it will be possible to avoid the use of multipatterns, which are widely used for the 10-nm process.
The company's engineers believe that for now, the DUV technology needs to be polished rather than jumping directly to the 7nm process. Thus, processors using 10nm technology will be discontinued for now.
Prospects for AMD microprocessor manufacturing
Intel's only real competitor in the processor manufacturing market today is AMD. Due to Intel's bugs with 10nm technology, AMD has slightly improved its position in the market. At Intel, mass production using the 10 nm process technology is very late. AMD is known to use a third party to manufacture its chips. And now there is a situation when AMD uses 7-nm processor manufacturing technologies to the fullest for production, which are not inferior to the main competitor.
The main third-party manufacturers of semiconductor devices using new technologies for complex logic are Taiwan Semiconductor Manufacturing Company (TSMC), US GlobalFoundaries, and Korea's Samsung Foundry.
AMD plans to use TSMC solely for the production of next-generation microprocessors. At the same time, new technologies for the production of processors will be applied. The company has already released a number of products using the 7nm process, including a 7nm GPU. The first one is planned to be released in 2019. Already in 2 years, it is planned to start mass production of 5nm chips.
GlobalFoundaries moved away from developing the 7nm process to focus on developing its 14/12nm processes for customers targeting high-growth markets. AMD is investing more in GlobalFoundaries to produce AMD's current-generation Ryzen, EPYC, and Radeon processors.
Production of microprocessors in Russia
The main microelectronic production facilities are located in Zelenograd (Mikron, Angstrem) and Moscow (Krokus). Belarus also has its own microelectronic production - the Integral company, which uses a 0.35 micron technological process.
Processors are produced in Russia by MCST and Baikal Electronics. The latest development of MCST is the Elbrus-8C processor. This is an 8-core microprocessor with a clock speed of 1.1-1.3 GHz. The performance of the Russian processor is 250 gigaflops (floating point operations per second). Representatives of the company declare that in a number of indicators the processor can compete even with the industry leader - Intel.
Production will continue with the Elbrus-16 model with a frequency of 1.5 GHz (the numeric index in the name indicates the number of cores). Mass production of these microprocessors will be carried out in Taiwan. This should help keep the price down. As you know, the price of the company's products is exorbitant. At the same time, according to the characteristics, the components are significantly inferior to the leading companies in this sector of the economy. So far, such processors will be used only in government organizations and for defense purposes. The 28-nm process technology will be used as a production technology for the processors of this line.
Baikal Electronics manufactures processors intended for industrial use. In particular, this applies to the Baikal T1 model. Its scope is routers, CNC systems and office equipment. The company does not stop there and is already developing a processor for personal computers - "Baikal M". Little is known about its characteristics. It is known that it will have an 8-core processor with support for up to 8 graphics cores. The advantage of this microprocessor will be its power efficiency.
Modern microprocessors are one of the most complex devices manufactured by man. The production of a semiconductor chip is much more resource-intensive than, say, the construction of a multi-storey building or the organization of the largest exhibition event. However, thanks to the mass production of the CPU in monetary terms, we do not notice this, and rarely does anyone think about the grandeur of the elements that occupy such a prominent place inside the system unit. We decided to study the details of the production of processors and tell about them in this material. Fortunately, there is enough information on the Web today on this topic, and a specialized selection of presentations and slides from Intel Corporation allows you to complete the task as clearly as possible. The enterprises of other giants of the semiconductor industry work on the same principle, so it can be said with confidence that all modern microcircuits follow an identical path of creation.
 |
 |
 |
 |
 |
 |
 |
 |
The first thing worth mentioning is the building material for processors. Silicon is the second most common element on the planet after oxygen. It is a natural semiconductor and is used as the main material for the production of chips of various microcircuits. Most silicon is found in ordinary sand (especially quartz) in the form of silicon dioxide (SiO2).
However, silicon is not the only material. Its closest relative and substitute is germanium, however, in the process of improving production, scientists identify good semiconductor properties in compounds of other elements and are preparing to test them in practice or are already doing so.
1 Silicon undergoes a multi-stage purification process: raw materials for microcircuits cannot contain more impurities than one foreign atom in a billion.
2 Silicon is melted in a special container and, having lowered a constantly cooled rotating rod inside, the substance is “wound” around it due to the forces of surface tension.
3 As a result, longitudinal blanks (single crystals) of circular cross section are obtained, each weighing about 100 kg.
4 The workpiece is cut into separate silicon disks - plates on which hundreds of microprocessors will be located. For these purposes, machines with diamond cutting discs or wire-abrasive installations are used.
5 The substrates are polished to a mirror finish in order to eliminate all defects on the surface. The next step is to apply the thinnest photopolymer layer.
6 The treated substrate is exposed to harsh ultraviolet radiation. A chemical reaction takes place in the photopolymer layer: light, passing through numerous stencils, repeats the patterns of the CPU layers.
7 The actual size of the applied image is several times smaller than the actual stencil.
8 Areas "etched" by radiation are washed out. On a silicon substrate, a pattern is obtained, which is then subjected to fixing.
9 The next stage in the manufacture of one layer is ionization, during which polymer-free silicon areas are bombarded with ions.
10 In places where they hit, the properties of electrical conductivity change.
11 The remaining polymer is removed, and the transistor is almost ready. Holes are made in the insulating layers, which, through a chemical reaction, are filled with copper atoms used as contacts.
12 The connection of transistors is a multi-level wiring. If you look through a microscope, you can see a lot of metal conductors on a crystal and silicon atoms or its modern substitutes placed between them.
13 Part of the finished substrate passes the first test for functionality. At this stage, current is applied to each of the selected transistors, and the automated system checks the operating parameters of the semiconductor.
14 The substrate is cut into separate parts with the help of the thinnest cutting wheels.
15 Good chips obtained as a result of this operation are used in the production of processors, and defective ones are sent to waste.
16 A separate chip, from which the processor will be made, is placed between the base (substrate) of the CPU and the heat-distributing cover and “packed”.
17 During the final testing, finished processors are checked for compliance with the required parameters and only then sorted. Based on the received data, microcode is flashed into them, allowing the system to properly determine the CPU.
18 Finished devices are packaged and sent to the market.
 |
 |
 |
 |
Interesting facts about processors and their production
"Silicon Valley" (Silicon Valley, USA, California)
It got its name from the main building element used in the production of microchips.
"Why are processor wafers round?"- you will surely ask.
For the production of silicon crystals, a technology is used that allows only cylindrical billets to be obtained, which are then cut into pieces. Until now, no one has been able to produce a square plate free of defects.
Why are microchips square?
It is this kind of lithography that makes it possible to use the area of the plate with maximum efficiency.
Why do processors need so many pins/pins?
In addition to the signal lines, each processor needs a stable power supply to work. With a power consumption of about 100-120 W and low voltage, a current of up to 100 A can flow through the contacts. A significant part of the CPU contacts is allocated specifically for the power supply system and is duplicated.
Production waste disposal
Previously, defective plates, their remains and defective microchips went to waste. To date, developments are underway to use them as the basis for the production of solar cells.
"Rabbit Suit"
This name was given to the white overalls, which are required to be worn by all workers in production facilities. This is done to maintain maximum cleanliness and to protect against accidental entry of dust particles into production facilities. The "bunny suit" was first used in processor factories in 1973 and has since become the accepted standard.
99,9999%
Only silicon of the highest purity is suitable for the production of processors. The blanks are cleaned with special chemicals.
300 mm
This is the diameter of modern silicon wafers for the production of processors.
1000 times
That's how much cleaner the air is in the chip factories than in the operating room.
20 layers
The processor chip is very thin (less than a millimeter), but more than 20 layers of the most complex structural combinations of transistors fit in it, which look like multi-level highways.
2500
That is how many Intel Atom processor chips (they have the smallest area among modern CPUs) are placed on one 300 mm wafer.
10 000 000 000 000 000 000
One hundred quintillion transistors in the form of microchip building blocks are shipped from factories every year. This is about 100 times more than the estimated number of ants on the planet.
A
The cost of producing one transistor in a processor today is equal to the price of printing one letter in a newspaper.
In the process of preparing the article, materials were used from the official website of Intel Corporation, www.intel.ua
As promised - a detailed story about how processors are made ... starting with sand. Everything you wanted to know but were afraid to ask
I've already talked about " Where are processors made?"and about what" Production difficulties are on this path. Today we will talk directly about the production itself - "from and to".
Processor production
When a factory for the production of processors using a new technology is built, it has 4 years to recoup the investment (over $5 billion) and make a profit. From simple secret calculations, it turns out that the factory should produce at least 100 working plates per hour.Briefly, the process of manufacturing a processor looks like this: a single crystal of a cylindrical shape is grown from molten silicon using special equipment. The resulting ingot is cooled and cut into "pancakes", the surface of which is carefully leveled and polished to a mirror finish. Then, integrated circuits are created on silicon wafers in the "clean rooms" of semiconductor factories by photolithography and etching. After re-cleaning the wafers, laboratory specialists perform selective testing of processors under a microscope - if everything is OK, then the finished wafers are cut into separate processors, which are later enclosed in cases.
Chemistry lessons
Let's look at the whole process in more detail. The content of silicon in the earth's crust is about 25-30% by weight, due to which this element ranks second in abundance after oxygen. Sand, especially quartz, has a high percentage of silicon in the form of silicon dioxide (SiO 2 ) and, at the beginning of the production process, is the basic component for creating semiconductors.Initially, SiO 2 is taken in the form of sand, which is reduced with coke in arc furnaces (at a temperature of about 1800 ° C):
These reactions using the recycle of the formed by-product silicon-containing substances reduce the cost and eliminate environmental problems:
2SiHCl 3 SiH 2 Cl 2 + SiCl 4The resulting hydrogen can be used in many places, but the most important thing is that "electronic" silicon was obtained, pure-pure (99.9999999%). A little later, a seed (“growth point”) is lowered into the melt of such silicon, which is gradually drawn out of the crucible. As a result, the so-called "boule" is formed - a single crystal as high as an adult. The weight is appropriate - in production, such a muzzle weighs about 100 kg.
2SiH 2 Cl 2 SiH 3 Cl + SiHCl 3
2SiH 3 Cl SiH 4 + SiH 2 Cl 2
SiH 4 Si + 2H 2
The ingot is skinned with a "zero" :) and cut with a diamond saw. The output is wafers (codenamed "wafer") with a thickness of about 1 mm and a diameter of 300 mm (~12 inches; these are the ones used for the 32nm process technology with HKMG, High-K / Metal Gate technology). Once upon a time, Intel used disks with a diameter of 50mm (2"), and in the near future it is already planned to switch to wafers with a diameter of 450mm - this is justified at least in terms of reducing the cost of manufacturing chips. Speaking of savings - all these crystals are grown outside of Intel; for processor manufacturing, they are purchased elsewhere.
Each plate is polished, made perfectly smooth, bringing its surface to a mirror finish.
The production of chips consists of more than three hundred operations, as a result of which more than 20 layers form a complex three-dimensional structure - the volume of the article available on Habré will not allow us to briefly talk about even half of this list :) Therefore, very briefly and only about the most important stages.
So. It is necessary to transfer the structure of the future processor into polished silicon wafers, that is, to introduce impurities into certain parts of the silicon wafer, which eventually form transistors. How to do it? In general, applying various layers to a processor substrate is a whole science, because even in theory such a process is not easy (not to mention practice, taking into account the scale) ... but it's so nice to understand the complex;) Well, or at least try to figure it out.
Photolithography
The problem is solved with the help of photolithography technology - the process of selective etching of the surface layer using a protective photomask. The technology is built on the principle of "light-pattern-photoresist" and proceeds as follows:- A layer of material is applied to the silicon substrate, from which a pattern is to be formed. It is applied to it photoresist- a layer of polymeric light-sensitive material that changes its physical and chemical properties when irradiated with light.The desired structure is drawn on a photomask - as a rule, this is a plate of optical glass, on which opaque areas are photographically applied. Each such template contains one of the layers of the future processor, so it must be very accurate and practical.
- Produced exposure(illumination of the photolayer for a precisely set period of time) through a photomask
- Removal of spent photoresist.
Sometimes it is simply impossible to deposit certain materials in the right places of the plate, so it is much easier to apply the material to the entire surface at once, removing excess from those places where it is not needed - the image above shows the application of photoresist in blue.
The wafer is irradiated with a stream of ions (positively or negatively charged atoms), which penetrate under the wafer surface at specified places and change the conductive properties of silicon (green areas are embedded foreign atoms).
How to isolate areas that do not require post-processing? Before lithography, a protective dielectric film is applied to the surface of the silicon wafer (at a high temperature in a special chamber) - as I already said, instead of traditional silicon dioxide, Intel began to use High-K dielectric. It is thicker than silicon dioxide, but at the same time it has the same capacitive properties. Moreover, due to the increase in thickness, the leakage current through the dielectric is reduced, and as a result, it has become possible to obtain more energy-efficient processors. In general, it is much more difficult to ensure the uniformity of this film over the entire surface of the plate - in this regard, high-precision temperature control is used in production.
So. In those places that will be treated with impurities, the protective film is not needed - it is carefully removed by etching (removal of layer areas to form a multilayer structure with certain properties). And how to remove it not everywhere, but only in the right areas? To do this, another layer of photoresist must be applied over the film - due to the centrifugal force of the rotating plate, it is applied in a very thin layer.
In photography, light passed through the negative film, fell on the surface of the photographic paper, and changed its chemical properties. In photolithography, the principle is similar: light is passed through a photomask onto a photoresist, and in those places where it passed through the mask, individual sections of the photoresist change properties. Light radiation is passed through the masks, which is focused on the substrate. Accurate focusing requires a special system of lenses or mirrors that can not only reduce the image cut on the mask to the size of a chip, but also accurately project it on the workpiece. The printed plates are typically four times smaller than the masks themselves.
The entire spent photoresist (which has changed its solubility under the action of irradiation) is removed with a special chemical solution - together with it, a part of the substrate under the illuminated photoresist is dissolved. The part of the substrate that was covered from light by the mask will not dissolve. It forms a conductor or a future active element - the result of this approach is different patterns of short circuits on each layer of the microprocessor.
Strictly speaking, all the previous steps were necessary in order to create semiconductor structures in the necessary places by introducing a donor (n-type) or acceptor (p-type) impurity. Suppose we need to make a p-type carrier concentration region in silicon, that is, a hole conduction band. To do this, the plate is processed using a device called implanter- boron ions are fired from a high-voltage accelerator with great energy and are evenly distributed in unprotected zones formed during photolithography.
Where the dielectric has been removed, the ions penetrate the layer of unprotected silicon - otherwise they get "stuck" in the dielectric. After the next etching process, the remnants of the dielectric are removed, and zones remain on the plate in which there is local boron. It is clear that modern processors can have several such layers - in this case, a dielectric layer is grown again in the resulting figure, and then everything goes along the trodden path - one more photoresist layer, the photolithography process (already using a new mask), etching, implantation ... well, you understood.
The characteristic size of the transistor is now 32 nm, and the wavelength that silicon is processed is not even ordinary light, but a special ultraviolet excimer laser - 193 nm. However, the laws of optics do not allow resolution of two objects that are less than half a wavelength apart. This is due to the diffraction of light. How to be? Apply various tricks - for example, in addition to the mentioned excimer lasers that shine far in the ultraviolet spectrum, modern photolithography uses multilayer reflective optics using special masks and a special process of immersion (immersion) photolithography.
The logical elements that are formed during the photolithography process must be connected to each other. To do this, the plates are placed in a solution of copper sulfate, in which, under the influence of an electric current, the metal atoms "settle" in the remaining "passages" - as a result of this galvanic process, conductive regions are formed that create connections between the individual parts of the processor "logic". Excess conductive coating is removed by polishing.
finish line
Hooray - the hardest part is behind us. It remains a tricky way to connect the "remnants" of transistors - the principle and sequence of all these connections (buses) is called the processor architecture. For each processor, these connections are different - although the circuits seem completely flat, in some cases up to 30 levels of such "wires" can be used. Remotely (at a very high magnification), all this looks like a futuristic road junction - and after all, someone is designing these balls!
When the processing of the plates is completed, the plates are transferred from production to the assembly and testing shop. There, the crystals are first tested, and those that pass the test (and this is the vast majority) are cut out of the substrate with a special device.
At the next stage, the processor is packaged in a substrate (in the figure - an Intel Core i5 processor, consisting of a CPU and an HD graphics chip).
Hello socket!
The substrate, die, and heat-distributing cover are connected together - this is the product we will mean when we say the word "processor". The green substrate creates an electrical and mechanical interface (gold is used to electrically connect the silicon chip to the case), thanks to which it will be possible to install the processor in the motherboard socket - in fact, this is just a platform on which the contacts from the small chip are separated. The heat-distributing cover is a thermal interface that cools the processor during operation - it is to this cover that the cooling system will adjoin, whether it be a cooler radiator or a healthy water block.
socket(central processor socket) - a socket or slot-type connector designed to install a central processor. Using a socket instead of directly soldering the processor on the motherboard makes it easier to replace the processor for computer upgrades or repairs. The connector can be designed to install the actual processor or CPU card (for example, in Pegasos). Each slot allows only a certain type of processor or CPU card to be installed.
At the final stage of production, finished processors undergo final tests for compliance with the main characteristics - if everything is in order, then the processors are sorted in the right order into special trays - in this form, the processors will go to manufacturers or go to OEM sales. Another batch will go on sale in the form of BOX versions - in a beautiful box along with a stock cooling system.
The end
Now imagine that a company announces, for example, 20 new processors. All of them are different from each other - the number of cores, cache volumes, supported technologies ... Each processor model uses a certain number of transistors (calculated in millions and even billions), its own principle of connecting elements ... And all this needs to be designed and created / automated - templates, lenses, lithographs, hundreds of parameters for each process, testing… And all this should work around the clock, at several factories at once… As a result, devices should appear that have no right to make mistakes in operation… And the cost of these technological masterpieces should be within the limits of decency… I’m almost sure is that you, like me, also cannot imagine the whole amount of work being done, which I tried to talk about today.Well, and something even more amazing. Imagine that you are a great scientist without five minutes - you carefully removed the heat-distributing cover of the processor and could see the structure of the processor through a huge microscope - all these connections, transistors ... even sketched something on a piece of paper so as not to forget. Do you think it is easy to learn the principles of the processor, having only this data and data on what tasks can be solved with the help of this processor? It seems to me that such a picture is now visible to scientists who are trying to study the work of the human brain at a similar level. Only if the Stanford microbiologists are to be believed, in one human brain
WHERE Intel processors are manufactured
As I wrote in a previous post, at the moment Intel has 4 factories capable of mass-producing 32nm processors: D1D and D1C in Oregon, Fab 32 in Arizona and Fab 11X in New Mexico.
Let's see how they are set up.
The height of each Intel processor factory Ventilation system level A microprocessor is made up of millions of transistors. Level of "clean rooms" The floor covers the area of several football fields In addition to the increased requirements for sterility, room Lower level Designed for systems supporting the operation of the fa- Engineering level |
It takes about 3 years and about 5 billion to build a factory of this level - it is this amount that the plant will have to “recapture” in the next 4 years (by the time a new technological process and architecture appear, the productivity required for this is about 100 working silicon wafers per hour). To build a factory you will need:
– more than 19,000 tons of steel
– more than 112,000 cubic meters of concrete
– more than 900 kilometers of cable
HOW microprocessors are made
A technically modern microprocessor is made in the form of a single ultra-large integrated circuit, consisting of several billion elements - this is one of the most complex structures created by man. The key elements of any microprocessor are discrete switches - transistors. By blocking and passing electric current (on-off), they enable the computer's logic circuits to work in two states, that is, in a binary system. Transistors are measured in nanometers. One nanometer (nm) is one billionth of a meter.
Briefly, the process of manufacturing a processor looks like this: a single crystal of a cylindrical shape is grown from molten silicon using special equipment. The resulting ingot is cooled and cut into "pancakes", the surface of which is carefully leveled and polished to a mirror finish. Then, integrated circuits are created on silicon wafers in the "clean rooms" of semiconductor factories by photolithography and etching. After re-cleaning the wafers, laboratory specialists perform selective testing of processors under a microscope - if everything is OK, then the finished wafers are cut into separate processors, which are later enclosed in cases.
Let's look at the whole process in more detail.
Initially, SiO2 is taken in the form of sand, which is reduced with coke in arc furnaces (at a temperature of about 1800°C):
SiO2 + 2C = Si + 2CO
Such silicon is called "technical" and has a purity of 98-99.9%. The production of processors requires a much cleaner raw material called "electronic silicon" - this should contain no more than one foreign atom per billion silicon atoms. To be refined to this level, silicon is literally "born again". Silicon tetrachloride (SiCl4) is obtained by chlorination of technical silicon, which is further converted into trichlorosilane (SiHCl3):
3SiCl4 + 2H2 + Si ↔ 4SiHCl3
These reactions using the recycle of the formed by-product silicon-containing substances reduce the cost and eliminate environmental problems:
2SiHCl3 ↔ SiH2Cl2 + SiCl4
2SiH2Cl2 ↔ SiH3Cl + SiHCl3
2SiH3Cl ↔ SiH4 + SiH2Cl2
SiH4 ↔ Si + 2H2
The resulting hydrogen can be used in many places, but the most important thing is that "electronic" silicon was obtained, pure-pure (99.9999999%). A little later, a seed (“growth point”) is lowered into the melt of such silicon, which is gradually drawn out of the crucible. As a result, the so-called "boule" is formed - a single crystal as high as an adult. The weight is appropriate - in production, such a boule weighs about 100 kg.
The ingot is skinned with a "zero" :) and cut with a diamond saw. At the output - plates (codenamed "wafer") with a thickness of about 1 mm and a diameter of 300 mm (~12 inches; these are the ones used for the 32nm process technology with HKMG, High-K / Metal Gate technology).
Now the most interesting thing is that it is necessary to transfer the structure of the future processor into polished silicon wafers, that is, to introduce impurities into certain parts of the silicon wafer, which eventually form transistors. How to do it?
The problem is solved using photolithography technology - the process of selective etching of the surface layer using a protective photomask. The technology is built on the principle of "light-pattern-photoresist" and proceeds as follows:
- A layer of material is applied to the silicon substrate, from which a pattern is to be formed. A photoresist is applied to it - a layer of photosensitive polymer material that changes its physical and chemical properties when irradiated with light.
- Exposure is made (photo layer is illuminated for a precisely set period of time) through a photo mask
— Removal of spent photoresist.
The desired structure is drawn on a photomask - as a rule, this is a plate of optical glass, on which opaque areas are photographically applied. Each such template contains one of the layers of the future processor, so it must be very accurate and practical.
The wafer is irradiated with a stream of ions (positively or negatively charged atoms), which penetrate under the wafer surface at specified places and change the conductive properties of silicon (green areas are embedded foreign atoms).
In photography, light passed through the negative film, fell on the surface of the photographic paper, and changed its chemical properties. In photolithography, the principle is similar: light is passed through a photomask onto a photoresist, and in those places where it passed through the mask, individual sections of the photoresist change properties. Light radiation is passed through the masks, which is focused on the substrate. Accurate focusing requires a special system of lenses or mirrors that can not only reduce the image cut on the mask to the size of a chip, but also accurately project it on the workpiece. The printed plates are typically four times smaller than the masks themselves. 
The entire spent photoresist (which has changed its solubility under the action of irradiation) is removed with a special chemical solution - along with it, a part of the substrate under the illuminated photoresist is dissolved. The part of the substrate that was covered from light by the mask will not dissolve. It forms a conductor or a future active element - the result of this approach is different patterns of closures on each layer of the microprocessor.
Strictly speaking, all the previous steps were necessary in order to create semiconductor structures in the necessary places by introducing a donor (n-type) or acceptor (p-type) impurity. Suppose we need to make a p-type carrier concentration region in silicon, that is, a hole conduction band. To do this, the plate is processed using a device called an implanter - boron ions are fired with great energy from a high-voltage accelerator and are evenly distributed in unprotected zones formed during photolithography.
Where the dielectric has been removed, the ions penetrate the layer of unprotected silicon - otherwise they "get stuck" in the dielectric. After the next etching process, the remnants of the dielectric are removed, and zones remain on the plate in which there is local boron. It is clear that modern processors can have several such layers - in this case, a dielectric layer is again grown in the resulting figure, and then everything goes along the trodden path - another layer of photoresist, the photolithography process (already using a new mask), etching, implantation ...
The logical elements that are formed during the photolithography process must be connected to each other. To do this, the plates are placed in a solution of copper sulfate, in which, under the influence of an electric current, the metal atoms "settle" in the remaining "passages" - as a result of this galvanic process, conductive regions are formed that create connections between the individual parts of the processor "logic". Excess conductive coating is removed by polishing.
Hurray - the most difficult behind. It remains a tricky way to connect the "remains" of transistors - the principle and sequence of all these connections (buses) is called the processor architecture. For each processor, these connections are different - although the circuits seem completely flat, in some cases up to 30 levels of such "wires" can be used.
When the processing of the plates is completed, the plates are transferred from production to the assembly and testing shop. There, the crystals are first tested, and those that pass the test (and this is the vast majority) are cut out of the substrate with a special device.
At the next stage, the processor is packaged in a substrate (in the figure - an Intel Core i5 processor, consisting of a CPU and an HD graphics chip).
The substrate, die and heat distribution cover are connected together - this is the product we will mean when we say the word "processor". The green substrate creates an electrical and mechanical interface (gold is used to electrically connect the silicon chip to the case), thanks to which it will be possible to install the processor in the motherboard socket - in fact, this is just a platform on which the contacts from a small chip are separated. The heat-distributing cover is a thermal interface that cools the processor during operation - it is to this cover that the cooling system will adjoin, whether it be a cooler radiator or a healthy water block.
Now imagine that a company announces, for example, 20 new processors. All of them are different from each other - the number of cores, cache volumes, supported technologies ... Each processor model uses a certain number of transistors (calculated in millions and even billions), its own principle of connecting elements ... And all this needs to be designed and created / automated - templates, lenses, lithographs, hundreds of parameters for each process, testing... And all this should work around the clock, at several factories at once... As a result, devices should appear that have no right to make mistakes in operation... And the cost of these technological masterpieces should be within the bounds of decency...
