Why is the CPU frequency not much higher than that of ten years ago, but the speed has increased? What determines the speed?

Frequency is not the only criterion to measure CPU performance. Intel and AMD will announce the basic frequency when releasing new CPU. In fact, how many GHz this frequency refers to the digital clock signal frequency (clock frequency) inside the CPU. It does not represent the true performance level of CPU, and a 4GHz CPU is not necessarily better than a 3GHz CPU.

Examples of high frequency and low energy

Case 1: In early years, when AMD and Intel were still fiercely competing around 1GHz, Intel was threatened by Athlon based on K7 architecture and rushed to launch Pentium 3 processor of 1. 13GHz. Pentium 3 was forced to recall because of many problems. At that time, people thought that the main frequency represented the performance of CPU. So Intel, which lost one after another, made a big move and launched Pentium 4 with NetBurst architecture, which appeared in 1.4~ 1.5GHz.

However, someone soon discovered the problem. According to the running score, 1.5G of Pentium 4 is only about 1.2~ 1.3GHz of Pentium 3. The reason why the frequency is so high lies in the architecture. Pentium 4 adopts a 20-stage pipeline technology based on NetBust architecture, which was previously 10. The advantage of long pipeline is high frequency but low efficiency.

Case 2: Much like the ultra-long assembly line design of Pentium 4, AMD's bulldozer has also become the same high frequency and low energy as Pentium 4, and its single-core performance is obviously backward compared with its previous generation K 10 architecture. ?

"Insufficient performance, overclocking to make up for it", the ultimate one is FX-9590, with a fundamental frequency of 4.7GHz, a maximum dynamic acceleration of 5.00GHz, and a TDP of 220W W, so that 90% of air-cooled radiators can't hold down at all, so FX-9590 is bundled with a high-end water-cooled radiator. The concept of frequency In order to ensure that all hardware units in the CPU can work together, a set of clock signals are needed to run synchronously with the system.

The clock signal is composed of a series of pulse signals, which always sends out a square wave signal with a certain voltage amplitude and time interval, and periodically changes back and forth between "0" and "1".

The number of repetitive pulses per unit time 1s is the frequency (Hz), and the clock frequency and period are reciprocal (f= 1/T). 1GHz means that 1 will generate 1 billion clock pulse signals. You can imagine how exquisite the internal structure of CPU is, which can handle such a short signal, and the whole system can run in a coordinated and orderly way. Why does the CPU frequency change?

Such as turbocharging

Boost technology allows each CPU core to have its own PLL circuit, so that the voltage and frequency of each core can be controlled independently. The power control unit will monitor the core parameters such as temperature, current and power consumption in real time at the speed of 1ms (1000 times per second), so that the CPU can adjust the CPU frequency according to the load. At the same time, because the more cores involved in the operation, the more complicated it is to control, so the more cores, the lower the highest frequency can be.

The external frequency is on the computer motherboard, mainly CPU, supplemented by memory and various peripheral devices, and many devices need to work together. The communication and data exchange between these devices must be correct every minute. Therefore, they must have a fixed clock for time correction, coordination or reference. This clock is generated by the clock generator on the motherboard, which is called external frequency. Although the frequency doubling CPU has been replaced, the external motherboard chipset, memory and external interface are still the old standards.

The working frequency of these devices is fixed for a long time, which is far lower than the working frequency of CPU, and it is impossible to communicate well with CPU. If CPU wants to get faster operation speed, it needs to get ultra-high-speed frequency to support faster operation speed. There is usually a PLL frequency generator inside the CPU, which divides the input clock signal and increases the input external frequency according to a certain proportion, thus obtaining the actual working frequency of the CPU. This ratio is called frequency doubling coefficient (frequency doubling for short).

Overclocking is calculated according to the main frequency of CPU: main frequency = external frequency.

X

Frequency doubling, overclocking is nothing more than overclocking, frequency doubling. Generally, we choose overclocking because it is much higher than the external frequency and easy to get. At present, many motherboards have their own one-button overclocking function. Motherboard manufacturers help you adjust the overclocking parameters in the BIOS, and overclock with one button.

In fact, the frequency doubling of CPU is extremely high, and the transmission speed of CPU and other devices in the system is still the same. The limit speed of CPU getting data from the system can't meet the running speed of CPU. So sometimes in order to meet the requirements of external transmission, we need to exceed the external frequency appropriately. The high temperature caused by overclocking will lead to the phenomenon of "electron migration".

However, the phenomenon of "electron migration" will damage the precision-designed transistors inside the CPU, so the heat dissipation of the CPU must be done well, and the overclocking of liquid nitrogen is also the reason.

There are many factors that affect the frequency, such as CPU architecture, pipeline design, internal register design, supported instructions, power consumption, temperature and so on. Therefore, the factory frequency of CPU is considered comprehensively, and the minimum value is taken as the highest frequency of CPU. Why is the current CPU frequency still around 4GHz? We must first understand how the power consumption of the transistor is calculated: the static power consumption is equal to the voltage multiplied by the current, w = v * i. When the transistor switches between "1" and "0", it will generate dynamic power consumption according to the switching frequency, w = v2 * f. Obviously, the higher the frequency, the greater the power consumption. Why didn't chip manufacturers give up making higher-frequency CPU? Because semiconductor technology has been progressing, 10nm, 7nm, 5nm, 3nm. The reduction of transistor area will reduce the voltage and current consumed by it by almost the same proportion. The improvement of technology can make the transistor smaller and the turn-on voltage lower, which obviously makes up for the increase of power consumption caused by frequency improvement. However, the process can not be improved endlessly, and the road will be difficult after 7nm.

Obviously, the higher the frequency, the greater the power consumption. Why didn't chip manufacturers give up making higher-frequency CPU? Because semiconductor technology has been progressing, 10nm, 7nm, 5nm, 3nm. The reduction of transistor area will reduce the voltage and current consumed by it by almost the same proportion. The improvement of technology can make the transistor smaller and the turn-on voltage lower, which obviously makes up for the increase of power consumption caused by frequency improvement. However, the process can not be improved endlessly, and the road will be difficult after 7nm.

After the transistor size is reduced, the static power consumption will increase instead of decrease, which will bring great heat energy conversion and the heat accumulation between transistors will be very serious. The heat dissipation of CPU has become an urgent problem to be solved. If the heat dissipation is not done well, the CPU life will be greatly reduced.

At present, the ubiquitous dynamic frequency technology in CPU, overheating will make CPU at the lowest working frequency, and high frequency is just a decoration and joke. Simply increasing the clock frequency of CPU will become unrealistic because of the subsequent heat dissipation problem. After all, liquid nitrogen will not always be used to cool the CPU, so both Intel and AMD are interested in stopping the research and development of high-frequency chips and turning to low-frequency multi-core architecture.