In everyday installed, the CPU processor time we choose, take a look at this piece is nothing more than the number of physical CPU core, supports Hyper-Threading technology, for example to intel will determine your choice i7, i5 or i3 series; again Some segments we need to consider the issue of frequency of the CPU, this would directly determine the purchase of specific models. So we usually always said that this intel Core i3-8350k default frequency is 4GHz, the AMD Dacentrurus 3 1300X default frequency of 3.5GHz, then this really is the CPU frequency CPU running speed yet? It is determined by what it?
Intel, AMD released when the new CPU will always publish its fundamental frequency, in fact, the frequency of how many GHz, in fact, refers to the internal digital clock signal frequency of CPU, also known as clock frequency, so it does not represent the true CPU performance levels, 4GHz is not necessarily stronger than 3GHz CPU, and at least we can not generalize. But the level of the clock frequency is indeed related to the operation speed of a CPU, the higher frequency means faster operation? So 4GHz frequency is how fast is it?
The concept of frequency
In this CPU complex digital system, in order to ensure that all internal hardware units can collaborate to work fast, CPU architecture engineers tend to design a clock signal synchronized with the system to operate. Clock signal is a signal composed of a series of pulses, and the amplitude of the voltage is always a certain time interval a continuous square wave signal emitted, which changes periodically back and forth between 0 and 1 . As shown below.
In the time between the first pulse and the second pulse interval period is called, it is in seconds (s). However, the number of pulses generated per unit time 1s called frequency, the basic unit of measurement is the hertz frequency Hz.
Clock frequency (f) and the period (T) of both the reciprocal: f = 1 / T
This formula shows that represent the clock frequency is repeated in a second frequency, and the current level GHz CPU is already in general, that generate one billion per second pulse signal.
CPU processor speeds
In Intel Core i3-8350k example, its frequency is 4GHz default, meaning that the internal clock frequency of 4GHz, may generate a second pulse signal 4000000000, 0.25ns (in other words every clock cycle when a pulse signal is only used ). What a shocking clock, CPU internal structure can imagine how subtle it is possible to deal with such a short signal, the entire system to run concurrently orderly, so when he says that CPU is the crystallization of the wisdom of mankind, which greatly enhance the progress of our technological level.
Minimum time unit as the CPU operation clock period, all operations are based on the internal clock as the reference cycle. In general, all rising edge of the CPU clock as a reference executing instructions, the higher the frequency, the more the number of instructions executed by the CPU, work faster.
The CPU frequency is what determines it? In fact, this is a very complex issue, because the decision of this frequency is something of a learning system, factors that affect the frequency of height very much, such as CPU architecture, pipeline design, internal registers designed to support instruction and even power, the temperature of these physical factors, so that the CPU frequency is factory integrated a variety of reasons, to lower the minimum barrel effect as the highest frequency of the CPU .
That's why we now CPU clock speed will become of it?
And have the Turbo Boost technology, each CPU core has its own PLL (Phase Locked Loop, phase-locked loop) circuit, so that each core voltage and frequency can be controlled independently, for which Intel CPU inside specially designed PCU ( power control unit, power control) unit, the PCU at a rate of 1ms (1000 times per second) of the real-time monitoring of the four core temperature, current, and power consumption parameters, so there Turbo Boost frequency may be adjusted according to the CPU load Frequency of. And because the more the number of cores involved in the operation, control them is more complex, so the more general core number, the lower the maximum frequency can be achieved.
FSB
CPU在诞生后不久，各大CPU巨头为了追求高性能，开启了频率大战（有没有效果这个我们先不提），但是这样一来，CPU虽然跑得更换了（主频更高），但是外部的主板芯片组、内存、外部接口（PCIe、Sata）可还是处于旧有标准，而且这些设备的运行频率早就固定下来了，并且远低于CPU工作频率。这样一来CPU就无法很好与之交流，Intel就机智地提出了倍频的概念（下面讲述），并且提出了影响至今的一个CPU主频计算公式：主频=外频 X 倍频，外频的提出可以让主板外部的设备工作在较低的频率下，并且还能正确地CPU进行交流。
但总是有很多网友将前端总线频率和外频混为一谈，其实他们不太一样。在以前有北桥的时代，前端总线是CPU总线接口单元和北桥芯片之间的数据交换通道，曾经在AMD雷鸟系列、Intel奔腾 4处理器以前，前端总线与外频是一致的，但后来有了四倍数据传输率技术或者是八倍数据传输率技术，前端总线频率就极大地提高了。举个例子，如果一个处理器的频率是2GHz，外频为100MHz，使用四倍数据传输率技术时，前端总线频率就变成400MHz；如果是八倍，那么就是800MHz。前端总线频率越大, 代表着CPU与北桥芯片之间的数据传输能力越大, 更能充分发挥出CPU的功能。
目前处理器的默认外频基本上都是100MHz
倍频
目前的CPU设计的外频都相当低，只有100MHz，CPU要获得更快运算速度，我们就需要获得一个超高速的频率来支撑更快运算速度。而CPU通常就是在内部设计有一个锁相环频率发生器，对于输入的时钟信号进行分频处理，按照一定比例提高输入的外频频率，从而得到CPU的实际工作频率，这个比例就称之为倍频系数（简称倍频）。
利用倍频技术, 较为完美地解决了CPU和内存等数据中转站的异步运行问题。为CPU后来向更高频率方向发展打下了扎实的基础。
超频
超频作为经久不衰的话题，一直都是PC DIY界中的常青树，一般是指强迫设备运行在高于其默认频率的主频以获得更高的性能，诸如CPU、显卡、内存等都可以超频, 其中以CPU为最为流行，可以在最大程度压榨CPU的性能，提高产品的性价比。而Intel、AMD一直将CPU可超频作为一大卖点宣传，那么我们超频超的是哪个频率？
根据CPU主频计算公式：主频=外频 X 倍频，我们无非就是超外频、倍频。
其实在不同时代，超得的是不同的频。在Pentium MMX时，Intel为了让CPU稳定工作，在主板BIOS中锁定了倍频， 不能随意更改倍频，那时候的我们只能从提高外频着手，一些旧式主板可通过设置主板上的跳线来变更计算机系统的外频（还记得跳线怎么插吗？），而在后期的主板BIOS中一般都会有SoftMenu技术，我们只需要在BIOS界面中动动手指头，调整外频频率就可以顺利提高CPU的主频。
而现在大家为了创造更高的主频，一般都是选择超倍频，因为超倍频提升幅度远比外频要高，而且来得容易。同样只需要在主板BIOS上调整倍频即可，目前很多主板都自带一键超频功能，主板厂商都BIOS中帮你调整好超频参数，只需要一键点击皆可以超倍频。
在相同的外频下, 倍频越高 CPU的主频也越高。但实际上，CPU倍频太高了，但是CPU与系统其他设备传输速度还是一样，之间的数据交换有限，造成了高主频的CPU就会出现明显的“瓶颈”效应-----CPU 从系统中得到的数 据的极限速度不能满足 CPU运算的速度。因此有时候为了满足外部传输需求，我们要适当超外频。
需要注意的是，超频会导致CPU发热量远大于正常工作温度，甚至是降低 CPU性能、寿命（缩缸）或造成系统的不稳定（蓝屏）。 降低 CPU 寿命是因为超频产生的高温会导致“电子迁移”现象， 而“电子迁移”现象会损坏CPU内部精密设计的晶体管，所以一定要必须做好CPU的散热工作，液氮超频也是出于这样的考虑。
但有时候CPU体质不佳（内部晶体管在制造的时候存在某些缺陷、瑕疵），导致超频困难，需要对CPU核心施加更高的工作电压，以我们拿到的地雷级intel Core i3-8350K为例，为了上5GHz频率，电压已经加到1.5V（默认1.34V），而之前的Core i7-7700K分分钟上5GHz。
那为什么我们现在CPU频率基本还停留在4GH平台呢？
CPU处理器中有一条金科玉律，那就大名鼎鼎的摩尔定律，它阐述了晶体管数目与性能提升的关系，之于它究竟是还活着，还是像死了般活着还很难说。但是我们今天要讲的是另一条不太出名的定律——登纳德缩放比例（Dennard Scaling）。
1974年内存之父罗伯特登纳德在其论文中表示，晶体管面积的缩小使得其所消耗的电压以及电流会以差不多相同的比例缩小，这个就是登纳德缩放比例定律。很多人摸不着头脑，这个和CPU频率提升有半毛钱关系吗？
确实密切相关。我们先了解晶体管功耗是如何计算的，静态功耗的就是常规的电压乘以电流，W=V x I。而晶体管在做 1和 0的相互转换时会根据转换频率的高低产生动态功耗，W=V2 x F。显然，频率越高，功耗就越大，但我们在此后的30年都没有放弃做频率更高的CPU呢？
答案是以为我们的半导体工艺一直在进步，目前甚至已经做到了10nm，7nm量产在即。根据登纳德缩放比例，工艺的提升，可以让晶体管们做的更小，导通电压更低，显然就弥补了频率提升带来功耗增加问题。但是我们的工艺并不是无休止境地提升，很快就会进入了一个长期的技术平台期，7nm以后路将会十分艰辛。
而且晶体管尺寸缩小以后，静态功耗不减反增，带来了很大的热能转换，加之晶体管之间的积热十分严重，让CPU散热问题成为亟待解决的问题。散热做不好，CPU寿命大大下降，而且目前普遍存在的动态频率技术，过热会让CPU处于最低工作频率，高频只是个装饰、是个笑话。单纯提高CPU时钟频率因为随之而来的散热问题而变得不再现实，毕竟我们不会无时无刻地使用液氮为CPU降温，所以Intel、AMD都很识趣地停止了高频芯片的研发，转而向低频多核的架构开始研究。
极限超频一般都是需要液氮、液氦来辅助散热
也因此我们目前才会看到多核CPU的大爆发，这是提升性能更好的方法。看完这篇干货，还想看其他同样这么干的超能课堂吗，找小超哥（微信：9501417）拿我们的百篇超能课堂吧~同时看完这篇文章对CPU超频还有其他感兴趣的问题不妨找我们小超哥提出，大家一起来探讨啊~
