After talking about the CPU frequency for so many years, what exactly is it?
Central processing unit
n daily installation, when you choose a CPU processor, you must first see how many physical cores the CPU has and whether it supports hyper-threading technology. Take Intel as an example to determine whether you choose i7, i5 or i3 series; To subdivide some, you need to consider the frequency of the CPU , which directly determines the specific model to purchase. So we usually say that the silent frequency of this intel Core i3-8350k is 4Ghz , and the silent frequency of this AMD Ryzen 3 1300X is 3.5Ghz , so is this CPU frequency really the operating speed of the CPU ? What determines it?
Intel andAMD always announce their basic frequency when they release new CPU . In fact, how many Ghz is this frequency, actually refers to the frequency of the digital clock signal inside the CPU , also known as the clock frequency, so it does not represent the true CPU . Performance level, 4Ghz CPU is not necessarily better than 3Ghz , at least we can not generalize. But the clock frequency is really related to the computing speed of a CPU . The higher the frequency, the faster the computing speed? So how fast is the 4Ghz frequency?
concept of frequency.
In the complex digital system of the CPU , in order to ensure that all internal hardware units can work together quickly, CPU architecture engineers often design a set of clock signals to operate in synchronization with the system. The clock signal is composed of a series of pulse signals, and is always a square wave signal that is continuously sent out at a certain voltage amplitude and time interval, and it periodically changes back and forth between 0 and 1. As shown below.
The time interval between the first pulse and the second pulse is called the period, and its unit is seconds (s). But the number of pulses generated within 1s per unit time is called frequency, and the most basic unit of measurement for frequency is Hz.
Clock frequency (f) and period (T) are the reciprocal of each other: f=1/T
This formula shows that the frequency represents the number of times the clock repeats in 1 second, and current CPU s are generally at the Ghz level, which means that 1 billion pulse signals are generated every second.
CPU frequency
Take Intel Core i3-8350k as an example. Its silent frequency is 4Ghz , which means its internal clock frequency is 4Ghz , which can generate 4 billion pulse signals per second. In other words, each pulse signal only takes 0.25ns (clock cycle ). This is a shocking clock. You can imagine how sophisticated the internal structure of the CPU is, capable of processing such short signals, and the entire system runs in an orderly manner. That's why the CPU is the crystallization of the wisdom of all mankind, which is greatly improved. It has improved our technological level.
The clock cycle is the minimum time unit for CPU operation, and all internal operations are based on this clock cycle. Generally speaking, the CPU uses the rising edge of the clock pulse as the benchmark for executing instructions. The higher the frequency, the more instructions the CPU executes and the faster the work speed.
So what determines the CPU frequency? In fact, this is a very complicated issue, because determining this frequency is a systemic thing, and there are many factors that affect the frequency height, such as CPU architecture, pipeline design, internal register design, supported instructions and even power consumption, Because of the physical factors of temperature, the CPU factory frequency is a combination of multiple considerations, and the minimum value under the barrel effect is regarded as the highest frequency of the CPU .
Then why does our current CPU frequency change?
Each core of a CPU with Turbo Boost technology has its own PLL (Phase Locked Loop) circuit, so that the voltage and frequency of each core can be independently controlled. For this reason, Intel specially designed the PCU ( Power Control Unit, PCU will monitor the temperature, current, power consumption and other parameters of these four cores in real time at a speed of 1ms (1000 times per second), so there is Turbo Boost frequency that can adjust the CPU according to the load. Frequency of. At the same time, because the more cores involved in the operation, the more complicated the control is. Therefore, the more cores, the lower the maximum frequency that can be achieved.
FSB
Soon after the birth of the CPU , in order to pursue high performance, the major CPU giants started a frequency war (we will not mention whether there is any effect), but in this way, although the CPU has been replaced (higher frequency), but the external The motherboard chipset, memory, and external interfaces (PCIe, Sata) of the motherboard are still in the old standard, and the operating frequency of these devices has long been fixed, and is far lower than the CPU operating frequency. As a result, the CPU cannot communicate well with it. Intel wittily proposed the concept of frequency multiplication (described below), and proposed a calculation formula for CPU main frequency that has affected so far: main frequency = external frequency X multiplier, external frequency The introduction of frequency allows the external devices of the motherboard to work at a lower frequency, and the CPU can communicate correctly.
But there are always many netizens who confuse the FSB frequency with the FSB, but they are not the same. In the era when there was a North Bridge, the front side bus was the data exchange channel between the CPU bus interface unit and the North Bridge chip. Before the AMD Thunderbird series and Intel Pentium 4 processor, the front side bus and the FSB were the same, but later there was With quadruple data transfer rate technology or eight times data transfer rate technology, the front-side bus frequency is greatly improved. For example, if the frequency of a processor is 2Ghz and the external frequency is 100MHz, when using the quadruple data transmission rate technology, the front-side bus frequency becomes 400MHz; if it is eight times, then it is 800MHz. The greater the frequency of the front-side bus, the greater the data transmission capacity between the CPU and the Northbridge chip, and the more capable of fully exerting the functions of the CPU .
The default FSB of the current processor is basically 100MHz
Frequency multiplication
The FSB of the current CPU design is quite low, only 100MHz. For the CPU to obtain faster computing speed, we need to obtain an ultra-high-speed frequency to support faster computing speed. The CPU usually has a phase-locked loop frequency generator internally designed to divide the input clock signal and increase the input FSB frequency according to a certain ratio, so as to obtain the actual operating frequency of the CPU. This ratio is called Frequency multiplication coefficient (referred to as frequency multiplication).
The use of frequency multiplication technology perfectly solves the problem of asynchronous operation of data transfer stations such as CPU and memory. It laid a solid foundation for the later development of the CPU in the direction of higher frequencies.
Overclocking
As an enduring topic, overclocking has always been an evergreen tree in the PC DIY world. It generally refers to forcing the device to run at a frequency higher than its default frequency to obtain higher performance, such as CPU , graphics card, memory, etc. It can be overclocked, among which CPU is the most popular, which can squeeze the performance of the CPU to the greatest extent and improve the cost-effectiveness of the product. Intel and AMD have always promoted CPU overclocking as a big selling point, so which frequency are we overclocking?
According to the calculation formula of CPU main frequency: main frequency = FSB X multiplier, we are nothing more than super FSB and multiplier.
In fact, in different eras, different frequencies are superb. In Pentium MMX, Intel locked the multiplier in the motherboard BIOS in order to make the CPU work stably. We could not change the multiplier at will. At that time, we could only start by increasing the FSB. Some old motherboards can set jumpers on the motherboard. To change the FSB of the computer system (remember how to plug in the jumper?), and in the later motherboard BIOS, there will usually be SoftMenu technology. We only need to move our fingers in the BIOS interface and adjust the FSB frequency to increase smoothly. The main frequency of the CPU.
Now, in order to create a higher main frequency, everyone generally chooses super-doubled frequency, because the boosting range of super-doubled frequency is much higher than that of FSB, and it is easy to come by. Similarly, you only need to adjust the frequency multiplier in the motherboard BIOS. At present, many motherboards have built-in one-key overclocking function. Motherboard manufacturers will help you adjust the overclocking parameters in the BIOS. You can overclock with just one click.
Under the same FSB, the higher the multiplier, the higher the CPU frequency. But in fact, the CPU multiplier is too high, but the transmission speed between the CPU and other devices in the system is still the same, and the data exchange between them is limited, resulting in an obvious "bottleneck" effect on the high-frequency CPU ----- The limit speed of the data obtained by the CPU from the system cannot meet the speed of the CPU operation. Therefore, sometimes in order to meet the external transmission requirements, we have to properly super FSB.
It should be noted that overclocking will cause the CPU to heat up much higher than the normal operating temperature, and even reduce the performance and life of the CPU (cylinder reduction) or cause system instability (blue screen). The life span of the CPU is reduced because the high temperature generated by overclocking will cause the phenomenon of "electromigration", and the phenomenon of "electromigration" will damage the precision-designed transistors inside the CPU, so the heat dissipation of the CPU must be done well. Liquid nitrogen overclocking is also due to this Considerations.
But sometimes the CPU is poor (there are some defects and flaws in the internal transistors during manufacturing), which makes it difficult to overclock. It is necessary to apply a higher working voltage to the CPU core. Take the mine-level intel Core i3-8350K we got. For example, in order to use the 5GHz frequency, the voltage has been increased to 1.5V (default 1.34V), while the previous Core i7-7700K uses 5GHz every minute.
Then why do we still stay on the 4GH platform for the CPU frequency?
There is a golden rule in CPU processors, that is, the famous Moore's Law, which explains the relationship between the number of transistors and performance improvement. It is difficult to say whether it is still alive or alive like dead. But what we are going to talk about today is another less well-known law-Dennard Scaling.
In 1974, Robert Dennard, the father of memory, stated in his paper that the reduction in the area of the transistor causes the voltage and current consumed by it to be reduced by almost the same ratio. This is the Dennard scaling law. Many people can't figure it out. Does this have half a dime relationship with the increase in CPU frequency?
It is indeed closely related. Let us first understand how the transistor power consumption is calculated. The static power consumption is the conventional voltage multiplied by the current, W=V x I. When the transistor performs the mutual conversion between 1 and 0, it will generate dynamic power consumption according to the switching frequency, W=V2 x F. Obviously, the higher the frequency, the greater the power consumption, but have we not given up on making higher-frequency CPUs for the next 30 years?
The answer is that our semiconductor process has been improving, and even 10nm has been achieved, and 7nm mass production is imminent. According to Dennard's scaling ratio, the improvement of the process can make the transistors smaller and the turn-on voltage lower, which obviously compensates for the increase in power consumption caused by the increase in frequency. But our process is not endlessly improved. It will soon enter a long-term technology platform period, and the road after 7nm will be very difficult.
Moreover, after the transistor size is reduced, the static power consumption does not decrease but increases, which brings about a large thermal energy conversion. In addition, the heat accumulation between the transistors is very serious, making the CPU heat dissipation problem an urgent problem to be solved. If the heat dissipation is not done well, the life of the CPU will be greatly reduced, and the dynamic frequency technology currently prevailing, overheating will make the CPU at the lowest operating frequency, the high frequency is just a decoration and a joke. It is no longer realistic to simply increase the CPU clock frequency due to the ensuing heat dissipation problem. After all, we will not use liquid nitrogen to cool the CPU all the time, so Intel and AMD are very interesting to stop the research and development of high-frequency chips. And began to study the low-frequency multi-core architecture.
The answer is that our semiconductor process has been improving, and even 10nm has been achieved, and 7nm mass production is imminent. According to Dennard's scaling ratio, the improvement of the process can make the transistors smaller and the turn-on voltage lower, which obviously compensates for the increase in power consumption caused by the increase in frequency. But our process is not endlessly improved. It will soon enter a long-term technology platform period, and the road after 7nm will be very difficult.
Moreover, after the transistor size is reduced, the static power consumption does not decrease but increases, which brings about a large thermal energy conversion. In addition, the heat accumulation between the transistors is very serious, making the CPU heat dissipation problem an urgent problem to be solved. If the heat dissipation is not done well, the life of the CPU will be greatly reduced, and the dynamic frequency technology currently prevailing, overheating will make the CPU at the lowest operating frequency, the high frequency is just a decoration and a joke. It is no longer realistic to simply increase the CPU clock frequency due to the ensuing heat dissipation problem. After all, we will not use liquid nitrogen to cool the CPU all the time, so Intel and AMD are very interesting to stop the research and development of high-frequency chips. And began to study the low-frequency multi-core architecture.
Extreme overclocking generally requires liquid nitrogen and liquid helium to assist in heat dissipation
Therefore, we are only now seeing a major explosion of multi-core CPUs, which is a better way to improve performance.
What is processor, how a processor work?
No comments:
Post a Comment