Almost no one knows about CPU when using computers, and everyone can talk about CPU. So have you seen what the CPU looks like inside? The following describes the internal CPU secrets.
(1) Basic CPU Structure:
CPU generally includes three parts: substrate, core, and pin
For example, the current CPU generally consists of three parts: substrate, core, and pin. The substrate is generally PCB, which is the carrier of the core and pins. The core and pins are fixed through the substrate, and the substrate connects the core and pins into a whole. The core is a circuit consisting of many transistors. For example, you can see different colors in the core enlarged image. The same color represents a type of hardware unit designed to implement a function, this hardware unit is composed of a large number of transistors. Different colors represent different hardware units. It should be noted that there is no color distinction in the actual chip. Here, we only want to be intuitive. Different colors represent different hardware units.
(2) The basic unit of the CPU core-transistor:
We often talk about the early Palomino core of AMD's mainstream CPU and the thoroughbred-B core adopted a 37.5 million transistor, the Barton core adopted a 54 million transistor, And the opteron core adopted a 0.106 billion transistor; intel's P4 Northwood core uses a 55 million transistor, Prescott core uses a 0.125 billion transistor, and so on, actually refers to the number of transistors that constitute the most basic unit of the CPU core. What does such a huge number of transistors look like? How does it work? See.
The most basic unit transistor in the CPU Core
Then, the transistor is connected to a whole through a circuit and divided into different execution units to process different data. In this way, collaborative work forms a CPU with powerful processing capabilities. How are these circuits connected together. This is the copper interconnect technology described below.
The CPU is a transistor made from silicon.
For example, the CPU is made of silicon as the raw material to form a transistor, coated with silica as the insulation layer, and then coated with metal wires (now copper) on the insulation layer, independent transistors are connected to work units. Now we use multi-layer copper interconnect technology. The signal transmitted in this way has less mutual interference and better quality. The result is that the CPU overclocking capability is stronger. The current CPU has adopted layer-7 copper interconnect technology, and more layers of copper interconnect technology will be used in the future.
The above copper interconnect technology line design, there is another problem, such a responsible line, how is the middle insulation?
The current in the CPU core flows through the copper interconnect wire.
This is now a popular term-low-K (low dielectric constant insulator) technology. In fact, this low-K material is used for insulation in the copper interconnect layer. From the figure, we can see that in the CPU core, the current flows through the copper interconnect wire, and low-K is used for insulation. So why do we need to adopt the low-k technology? The reason is actually very simple: the use of low-K as the insulation material, so that the leakage between lines is reduced, so that the chip's heat is low. At present, most 0.13 micron process products use FSG fluorine glass medium with low performance as the insulation layer. The chip's heat is far higher than the low-K process.
However, because the chip of Low-k technology is brittle, it requires a high process in chip encapsulation. It was originally planned to adopt the low dielectric constant insulator low-k technology in 0.18 micron manufacturing, but it was not until the 0.13 micron process began to mature. The chip quality of Low-k technology is brittle and requires high technology in chip encapsulation, so it has been plagued by the quality rate. However, now the low-K process is basically mature. With the advanced process, the CPU with a higher frequency can be made to reduce the fever.
(3) Two Problems in CPU usage:
For CPU, many users have two problems: CPU overspeed and CPU life.
A. CPU overclocking: why can I overclock the CPU? What affects the overclocking capability?
The manufacturing process of the CPU is to use a laser to etching the circuit on the wafer. The so-called etching uses a certain wavelength of UV through the mask (mask, equivalent to the baseline used for photo washing) and then shines on the silicon wafer, completely copies the circuit images on the mask to the silicon wafer. The key to the etching process is the use of UV wavelength and wafer quality. The shorter the wavelength, the less obvious the ultraviolet interference and diffraction, and the smaller the width of the transistor.
The higher the wafer purity, the larger the diameter, the higher the chip performance, the higher the yield rate. In the process of silicon wafer production, the farther away from the wafer center, the more prone to bad points. Therefore, the number of bad points is on the rise as the number of points expands from the silicon wafer center. At the same time, the higher the purity, the less defects produced by the chip, the higher the frequency.
Crystal forming, slicing, and etching circuits are essential for manufacturing processors.
It can be seen that there is no frequency in the CPU laser etching process, that is, they are all produced according to the same process. However, in the production process, there will be perfect chips and defective chips.
Then, you need to filter the tags. The product is labeled as high-frequency product based on the quality of different chips. the defective product is labeled as a low-frequency product for stability.
The CPU labeling process is also a strict detection process. Due to the influence of wafer purity and uncertainty in the processing process, it is impossible for all CPUs to be perfect products produced according to the design. In this case, the product needs to be differentiated according to the actual performance. This requires strict detection. The CPU stability test is very strict, and its test environment is quite harsh. The CPU stability is fully tested through a large number of complex multi-task operations. In this way, select a conservative program that can pass all the strictly tested programs stably, and then conduct the product logo. This is the CPU frequency. In daily use, there may not be such a heavy multi-task environment in years, so even if the frequency is too high, it is still stable. This is why almost all CPUs can overclock. Because the vendor has reserved some space for Frequency improvement for stability considerations.
In addition, some CPUs in the same batch of high-frequency versions are marked as low-frequency versions. Most of these occur in the early stages of the adoption of new processes. If the production capacity is insufficient, the amount is not too large. This is the so-called Super frequency.
In fact, the manufacturing process affects the overclocking capability. As long as the process is improved, the overclocking capability naturally increases.
B. Impact of overclocking on CPU life:
Super frequency has an impact on the life of the CPU. Now let's analyze the effect of overclocking on the CPU frequency. In the actual overclocking process, we will increase the voltage and appropriately increase the voltage so that the signal processed in the transistor can be clearer and the transistor switching speed can be faster. However, a high voltage will penetrate the transistor, which we often say is burned out. At the same time, high voltage will lead to an increase in fever. The long-term adverse effect of fever is often referred to as electronic migration.
Influence of overclock on CPU
Electron migration is the migration of metal atoms caused by electronic flows. Inside the chip, on a metal wire with high current strength, the electronic flow gives the metal atoms a momentum. When it accumulates to a certain extent, the metal atoms will be separated from the gravity between the metal atoms, flow around with the current. In this way, the surface of the core line is formed into holes or cuts. This is an irreversible permanent injury, even if it is a slow process. When the accumulation reaches a certain level, a short circuit of the core internal circuit will be formed, and the chip will be scrapped. This process is similar to the flow of water to wash the bed, a little bit of carrying stones, the last is likely to block the river.
It can be said that within the CPU chip, electronic migration has always existed, but in normal use, because it is a long process of impact, so in the normal use process, we can ignore the impact of electronic migration on the chip life. However, the high temperature caused by overclocking greatly enhances the electronic migration of metal atoms.
The increase in temperature will greatly increase the kinetic energy of free electrons, and the collision with metal atoms will be more frequent and intense. At the same time, with the increase of temperature, metal atoms will become active, and over-active will bring two results, one is that the original narrow lead will be broken because of the departure of metal atoms, or a short line is formed when the insulation of the online room is accumulated.
The higher the temperature, the more likely the electron migration phenomenon will occur. The higher the temperature will increase the activity of the electronic motion and the metal atomic thermal motion, so that even the system is still stable under the overclock, however, a processor that works at the limit is likely to be wiped out in a slightly fluctuating voltage, because either internal short circuit or open circuit affects the processor's logical operation function.
Therefore, the high temperature caused by overclocking not only affects the stability of CPU operations, but also gradually shortens the CPU life. So advocate: According to the chip constitution, moderate overfrequency. The limit frequency cannot be pursued, which is very harmful to the CPU. At the same time, do a good job of heat dissipation.