Intel Museum Tour: 38 years of x86 history and more
During the IDF conference, which was held a couple of weeks ago, we visited the Intel Museum in the central office in Santa Clara. The exposition shows how silicon from simple sand turns into a microchip crystal, and contains the company's main products, starting with the legendary Intel 4004 CPU. We offer to take a look at each of these chips and refresh the memory of how they are born.
Intel is known as the creator of the world's first single-chip microprocessor (as they would say now, CPU on Сhip - similar to the term SoC, System on Chip) and the x86 instruction set, but this is not the only area in which the California giant gained fame as a pioneer or at least Locomotive development of certain technologies. In the recent past, Intel has shown that SSD is destined to become an indispensable component of the mass PC, releasing the Intel X25-M drive - the first truly successful drive based on MLC memory, which has long served as a benchmark in this category.And last month, we saw what SSDs with memory chips are really capable of — a new type — XPoint. At the same time, few people know or remember that the first DRAM chip, without which any modern computer is unthinkable, also released Intel.
We want to show readers the devices that lie at the base of the empire, which Intel has become, and tell you exactly how each of them helped make a personal computer the way we know it today — a fast and convenient, compact laptop or a powerful server. We have no doubt that many of the visitors of 3DNews remember this story and were holding in their hands Intel processors of the past. But there are newcomers, in whom the interest in computer technology woke up not so long ago by the standards of this rapidly growing industry. Our brief tour of the Intel Museum is intended primarily for them.
From grit to processor: how to make chips
The space of the Intel Museum is divided into two parts - an exhibition of devices and exhibits illustrating the process of manufacturing integrated circuits. Connoisseurs reading this article need not explain how processors are made.Some of the less experienced readers, on the contrary, saw in our reviews only large semiconductor wafers containing dozens and hundreds of chips, which then find a place in processors, video cards and memory modules. However, the plate is only the final stage of a long and complex process.
The stand of the museum, which you see in the photo, talks about what happens at each stage of the production of the microchip. We briefly repeat this information, using additional illustrations.
It all starts with pieces of highly purified silicon (no more than 1 billionth of impurities), which the factory extracts from quartz sand. Then from silicon melted at a temperature of 1425 ° C according to the so-called. the Czochralski method forms a huge ingot weighing several hundred kilograms, which, unlike crystals found in nature, has a single crystal lattice, almost devoid of defects. To do this, a seed crystal is dipped into a crucible with liquid silicon on a long rod and removed, rotating at a precisely controlled rate.
Silicon Containing Minerals
Alloying additives in the form of elements such as boron, phosphorus, etc.,used to change the electrical properties of a material — to make it a p-type semiconductor (the predominant method in the production of plates) or n-type. Here, many of the readers will need to remember the course of physics or to replenish knowledge independently. Alas, this article is just a story about the Intel Museum, it does not pretend to be a complete introduction to the principles of semiconductors and transistor logic.
In order to get round plates on which future processors will be formed, silicon crystal is cut with the finest wire saw. The first Intel chips were made on the basis of plates with a diameter of 2 inches (51 mm). The most common technology of today produces plates with a diameter of 300 mm, which have a thickness of 0.775 mm. The next step, while limited to experimental use, will increase the diameter of the plates to 450 mm.
By the way, the finished plates that the chip makers demonstrate to the public are in fact too thin to hold in their hands, so they are covered with glass on both sides. On production for carrying plates used special tools and containers.
The structure of the chip on the semiconductor wafer is formed using photolithography. Here are the basic steps of this extremely sophisticated process.
A layer of insulator silicon oxide (SiO) is formed on the plate.
Apply a layer of so-called. photoresist. A positive photoresist (the most common type) becomes soluble in a particular developer (developer) as a result of exposure to light.
The light is projected onto the plate through a photo mask, the drawing of which in an enlarged form determines the structure of those elements of the transistors that are formed at this stage. Notice that on the layout of a semiconductor factory with one of the photographs of the so-called. the clean rooms where the photolithographic process takes place are flooded with yellow-orange light. This reduces the risk of parasitic illumination of the photoresist.
Exposed portions of the photoresist are washed away by the developer.
Areas of SiO on the surface of the plate that are no longer protected by the photoresist are etched with caustic matter.
Source of illustration - Wikipedia
The procedure is repeated several times with the addition of new operations. One of the iterations includes the so-called. ion implantation - the bombardment of a silicon base (representing a p-type semiconductor) with ions, which as a result creates areas of a reverse semiconductor type (n),which are the two components of the transistor - the source (source) and the drain (drain). Applying a layer of polycrystalline silicon followed by etching creates a gate (gate) of the transistor.
When transistors are ready, according to the same principle, the most numerous layers in a microcircuit are created (there are up to 12 of them) containing metal compounds. A modern semiconductor plate undergoes about 50 iterations of the photolithographic process, the final surface treatment (as a result of which the plate acquires the usual thickness for ready-made microcircuits), not to mention the multiple stages of intermediate testing. In the end, the chips are cut out of the plate mechanically and packaged in suitable enclosures.
From Intel 4004 to Intel Core
The manufacture of modern microcircuits and the prospects for its development in future years is an extremely extensive topic worthy of a more detailed story. We continue our tour of the Intel Museum with a brief overview of the samples that mark the main milestones of the company's history.
The main product for Intel in the early years of its existence (from its foundation in 1968 to the early 1980s) was memory chips. The first two products released by Intel in 1969 were the Static Random Access Memory (SRAM) Intel 3101 and Intel 1101 chips.
Among the two devices, the development of which Intel led in parallel, the first was completed and hit the market faster. Intel 3101 used cells based on Schottky transistors: the actual transistor in combination with a Schottky diode. The latter component is an integral part of the vast majority of switching power supplies, which you can read about in our guidebook, and in the SRAM memory, the Schottky transistor technology has given way to MOSFET transistors.
The pioneering product of this type was the Intel 1101 chip. Then it was a discrete chip in a computer, and later SRAM was integrated directly into the microprocessor core, where it is used as a cache due to the high data transfer rate and the absence of the need to constantly update the stored information. However, discrete SRAM chips have been used for quite a long time - for example, in the Intel CPU slot-form factor up to the Pentium III.
Intel 1101 and 3101 - the first SRAM memory chips from Intel
Intel also owns the laurels of the creator of the world's first random access dynamic memory chip (DRAM) - Intel 1103 (1970). RAM modules, on-board memory of video adapters, controller cache in SSD - all these and many other computer components are based on DRAM memory,not so high, but more dense and cheaper than SRAM.
Another type of memory, in the production of which Intel was a pioneer, was the EPROM (Intel 1702, 1971). For a device equipped with an EPROM chip, this non-volatile memory is not modifiable, because the programmer used for recording uses a higher voltage for the cells. The contents of the EPROM are erased under the influence of ultraviolet radiation, for which the sealed window is usually provided for on the chips. EPROM was typically used to store microcontroller firmware and BIOS in a PC. Subsequently, the place of EPROM was taken by the EEPROM technology, which allows you to overwrite the contents of the chip by means of the finished device. It was also first introduced on the Intel market (Intel chip 2816, 1980)
Intel 1702 and 2816 - the first EPROM and EEPROM memory chips
Layout of Intel 4004 on a larger scale
As for the central processors, this year Intel celebrated the 10th anniversary of the Core architecture. Since 2006, these processors remain unsurpassed in performance among all CPUs compatible with the x86 instruction set. It is no exaggeration to say that Intel is currently making the fastest general-purpose processors for desktops, laptops and servers.This situation persists for so long that all the peripeties of the previous history of the CPU are finally a thing of the past and have lost touch with modern computers in the memory of the majority of their owners. Recall the contribution that each of the processors stored in the Intel Museum made to the success of modern CPUs.
The first masses microprocessor in the world, the Intel 4004, released in 1971, is the most popular among the masses of Intel inventions. His innovation was that it was the Intel 4004 that for the first time combined all the steps of executing instructions on a single chip. No matter how simple the CPU logic at that time was, it was only the invention of a variety of photolithographic process that formed transistors with a polycrystalline silicon gate (which we described in the first part of the article), which took place just three years before the Intel 4004, solved this problem.
Intel 4004 and other Busicom 141-PF chips
Intel 4004 was produced according to the technical process with a step of 10 microns. It contains about 2,300 transistors in a 12 mm2 core. CPU clock speed is 1 MHz. The processor was created by Intel on the order of the Japanese company Busicom for the Busicom 141-PF calculator. In addition to the Intel 4004, the device includes three other chips: the Intel 4001 ROM chip, RAM - the Intel 4002, and the shift register Intel 4003 (which performed the conversion from a serial to a parallel interface).
Further exhibits also do not need a long presentation.Here is the IBM PC 5150, the ancestor of "personal computers" as a trademark, which then swallowed or supplanted many other platforms that embodied the idea of a computer for one person - as opposed to the huge mainframes used collectively. The IBM PC was based on two components that determined the appearance of most computers in the workplace and at home for many years to come: the MS-DOS operating system and the processor with the x86 instruction set, created by Intel (Intel 8086, released in 1978). Backward compatibility with 16-bit Intel 8086 is still preserved in x86 processors: theoretically, the code written for 8086 can be run on a modern system.
IBM PC 5150
The following photo shows Intel processors starting at 80286 and ending with the Pentium II.
Recall in brief what is remarkable for each of them.
80286 - the first x86 CPU in the family, capable of operating in the so-called. protected mode that provides virtual memory support. As a result, preemptive multitasking became possible - the execution of the code of one program, not threatening the data of another program.
80386, or i386, is the first 32-bit Intel processor.
The i486 is the first x86 family of CPUs that has broken the ceiling of 1 million transistors.Intel also for the first time integrated a floating-point coprocessor into the core. Previously, it was a discrete component.
Pentium is an innovative chip that for the first time for Intel embodies a superscalar architecture (the ability to execute several instructions simultaneously).
Pentium Pro is one of the key stages in the evolution of Intel processors, presenting the function of an extraordinary execution of instructions that is critical for the speed of modern CPUs. Another fundamental fact: from this point on, Intel CPUs no longer execute x86 dial-in instructions, but convert them into short micro-operations of fixed length (like a set of RISC instructions).
P6, microarchitecture Pentium Pro, formed the basis of processors Pentium II and Pentium III. And then comes the stage of the story, much better known to most 3DNews readers. Intel chose a completely new design, NetBurst, as the architecture for the future Pentium 4 CPUs (2000). The future of NetBurst, relying on the clock frequency as the main source of speed, the creators viewed up to the mark of 10 GHz.
However, faced with the problem of increased heat generation in the final years of the Pentium 4 and Pentium D, Intel was forced to revert to the P6 architecture (which, with a number of changes,was still used in Pentium M mobile CPUs) and based on its principles, the super-successful Core architecture (Core 2 Duo / Quad).
Virtual tour of the museum Intel has come to an end. Our story does not claim a comprehensive biography of the products of this company. In the end, the exposition itself includes only the most glorious or significant Intel devices. In contrast, specialized products (such as chips with the MIC architecture for Xeon Phi accelerators) are not available here, as well as representatives of areas that Intel stopped developing (for example, processors with ARM or Itanium instruction sets) or did not bring to commercial implementation (Larrabee graphics processor) .
Nevertheless, the exhibits of the small Intel museum reveal their charm only when the story behind each of them is known, so we hope that the condensed comment in this article was interesting for our readers.