How to choose a central processing unit

The key advantage of the personal computer as a platform has been and remains versatility. It is thanks to the ability to solve completely different tasks on one device that desktop computers familiar to us for decades without the slightest loss are experiencing the emergence of new and “revolutionary” platforms that threaten to replace them.

However, to be fully versatile, a computer needs adequate performance. And, if in games it is primarily determined by the video card, then for work tasks the capabilities of the central processor are most often important (although, of course, RAM and the disk subsystem also matter).

In this guide, answers to the main questions that arise when choosing a CPU will be given, and the parameters of the processors themselves will be divided into important ones and those that do not have a decisive value when choosing.

What you should NEVER pay attention to
The two CPU manufacturers each have fully formed product lines that cater to different market segments, from HEDT to embedded systems. And it is only logical that a processor for a powerful workstation and a processor for a nettop simply cannot have the same qualities.

Moreover, different processor models, even in the same family, can differ markedly in characteristics. Therefore, to say that the conditional Core i3 is exactly the same as the conditional Core i9, only a little slower is simply to manipulate. Moreover, to manipulate not even facts, but the emotions of a potential buyer.

Therefore, when choosing a central processor, clearly understand: you are buying a specific device with specific characteristics. Namely: performance in important tasks for you, the total cost of the platform, the possibility of further upgrades, power consumption, requirements for the cooling system, etc. These characteristics may suit you or seem inappropriate to the cost of the processor, but these will be real parameters related specifically to the product under discussion.

But such mythical criteria as “brand reputation”, “dampness of architecture”, “a ticket to the owners’ club” and other “disclosure percentages” are a direct and short path to buying the worst possible option.
















Frequently asked Questions
Q: I have a motherboard called < socket_name >. Can I put a processor in it under < socket_name_plus_one_digit >?

A: You can’t.

The motherboard socket is nothing more than a mate for the processor pins. Simply put – for legs or pads, which everyone can see by turning the processor cover (or crystal) down.

For each platform, the location of these contacts is unique, and often not only the number and location of contacts differ, but also the dimensions of the socket. As a result, the processor cannot be physically installed in a socket that is alien to it, and if you succeed, most likely, the processor and the motherboard will receive irreversible damage as a result. As, in fact, your wallet – after all, you will have to buy two more new devices for replacement.

Moreover, sometimes even constructively similar platforms are incompatible with each other. For example, a processor for an LGA 1151 socket will not work in a motherboard with an LGA 1151_v2 socket, and a processor for an LGA 1151_v2 socket, accordingly, will not start in a motherboard for an LGA 1151.

If you already have a motherboard, just check the list of compatible processors, which is always available on the motherboard manufacturer’s website. It’s just a couple of minutes, but you will save much more. And both time and money.

Q: But I bought a processor under < socket_name > and a motherboard with the same < socket_name >, but the system still does not start, the screen is black and only the fans are spinning. What does it mean?

A: Two options.

Option “1” – You have connected a monitor to the outputs on the motherboard, while the processor lacks integrated graphics. Surprisingly, this happens quite often. There is only one way out – to connect the monitor to a discrete video card.
Option “2” – you bought a processor that is compatible with your motherboard, but not supported on the BIOS version that is currently recorded on the board.
Such a situation is possible on absolutely all platforms – remember at least Intel Kaby Lake processors (7000 series) and motherboards based on 100 series chipsets … well, or Coffee Lake Refresh (9000 series) and motherboards based on 300 series chipsets, released before the CPUs themselves. In any case, this is not a flaw in the platform itself. If the board was released earlier than the processor, it simply cannot know how to work with it. Here you can draw an analogy with cars: if some option began to be installed at the factory only in 2019, then it cannot appear on a car produced in 2017 – unless, of course, you buy and install it.

If you are just going to buy a motherboard and are in doubt if it supports the processor, check the factory firmware number. This is not at all difficult to do, and you do not need to buy a board and carry it home for this:

If the CPU is purchased for an upgrade, and you already have a working system with another processor, you can upgrade yourself, and for this there is a standard toolkit.

In the end, you can contact the service center to update the BIOS of the motherboard. The firmware service is available in the CS of the CSN company, but no one will prohibit contacting any other company that repairs and maintains a PC.

Q: But I have a 450 watt power supply unit, I want to replace the processor with ” model name “. Will my block suffice, or will it also need to be changed?

A: Depends on the actual characteristics of your PSU and the power consumed by the entire system together.

Suppose that your power supply is of high quality, modern and actually delivers the declared power, and most of it is through a 12 volt line, and even does not drain voltage at peak load. Then there are no complaints about it, and you just have to find out how the power of the power supply corresponds to the appetites of the system.

In addition to the processor, the graphics card actively consumes electricity under 3D load, so it’s worth examining its real characteristics. The consumption of the motherboard, RAM, hard drives and expansion cards is not so significant, but, depending on the number of devices indicated above, add another 50–70 watts to the power consumption of the video card.

How to determine the consumption of the processor and video card under load? The easiest way is to use ready-made tests from authoritative sources that use adequate measurement techniques.

A less accurate way is to measure yourself. If you are sure of the quality and technical condition of your components, run the utility for monitoring system parameters HWinfo64, and then measure the power consumption of the processor and video card under stress tests. FurMark or MSI Kombustor for a graphics card, OCCT Linpack or Prime95 for a processor. Then add up the data obtained, add the above 50-70 watts to the rest of the components, and you will find out how much your system consumes at peak load.

Accordingly, if this figure is significantly less than the declared power of your power supply unit, you can upgrade. If the current configuration consumes almost the maximum power for the unit, the power supply unit should definitely be replaced.

However, remember that any stress tests you conduct are entirely at your own risk. If your system has problems with cooling, or the power supply unit is going through hard times and cannot provide high-quality power, one or more components may fail. However, in this case, replacing the processor can lead to a similar result …

Q: But a 95 watt cooler is enough for cooling ” processor_name “? There is no possibility to change the cooler either …

A: But it depends on the actual power consumption of the processor.

In the characteristics of the CPU, such a value is always given as TDP, it stands for Thermal Design Power, and represents the requirements for the thermal power that the cooler is able to dissipate. However, these requirements are often specified only for the normal operation of the CPU. Moreover, the nominal mode here means the base frequency of the processor, and not the frequency at which it actually works due to the nominal dynamic overclocking. For example, the same Core i9-9900K with a declared TDP of 95 watts can really consume no more than 95 watts at peak load, but only if it operates at a base 3600 MHz with an appropriate voltage. In reality, due to the MCE technology, the processor operates at 4700 MHz even when all its cores are fully loaded. And energy consumption, and hence heat dissipation, in this case,

However, the opposite is also true. If the processor has a software limit for 90 watts of power consumption, then it cannot physically release 130, 150, 200 or 250 watts of thermal energy: if this were so, we would be talking about a revolution in the field of household heaters that produce more power than consumed. And even there it would be close to interplanetary flights.

But jokes aside: in reality, 90-100 watts of power consumption only means that it will be much easier to cool the processor than it is commonly thought:

As far as coolers are concerned, by the way, it is also worth focusing on TDP with great caution. Each manufacturer uses a different measurement methodology, as a result of which coolers with a TDP declared at around 130 watts can have completely different designs and completely different efficiency. But, nevertheless, if 130 watts of dissipated power is declared for a cooler, then, most likely, it will cope with a processor with an energy consumption of 70–80 watts.

Q: Or maybe just take the processor in the box? There will also be a complete cooler!

A: In fact, it is far from the fact that he will be there.

For example, Intel processors with an unlocked multiplier are shipped without a stock cooler, the same applies to several AMD Ryzen 1000 series processors and Ryzen 3000 processors with the XT suffix. In this case, the manufacturer believes that the product will certainly be used with a better and more efficient cooling system than a boxed cooler.

Whether to take a processor in BOX or OEM-configuration is everyone’s personal choice. But you shouldn’t choose BOX just for the sake of a cooler: it will not always be distinguished by high efficiency at a comfortable noise level. And most often, for the difference in price between BOX and OEM, you can buy a cooler much better than the standard one.

Q: Do I need to overclock the processor?

A: Everyone’s personal file.

If you are purchasing a processor for a long time, it is better to consider the option with an unlocked multiplier. You may not need overclocking at the time of purchase, but you will need it in the future when the demands of games and working software increase. The stock, as they say, does not hold a pocket. Although the price of a processor with an unlocked multiplier and a motherboard with the ability to overclock may turn out to be much higher than a platform without overclocking.

However, this is true only for the Intel desktop platform, where there is a division into overclockable and non-overclockable components.

On AMD socket AM4 desktop platform, overclocking is available for all CPU and APU models. Only motherboards based on low-end A320 and A520 chipsets lack the ability to overclock the processor. However, these motherboards are intended for the office segment, so everything is logical here.

Q: I’ve chosen a processor, but I looked at the characteristics – and it says that the processor has built-in graphics. But I already have a video card – why pay extra for something I won’t be using?

A: Should I overpay?

Paradoxically, even if your computer has a powerful discrete graphics card, integrated graphics are a useful bonus, which in some situations can, on the contrary, save you money, time and nerves. Any video card, no matter how reliable it is, can fail over time and go to the service center. Any video card may require an upgrade over time, and it is not at all a fact that you will sell the old one and buy a new one on the same day.

Yes, you can think of other reasons when you are temporarily left without a video card. What do you have to do in this case if you don’t have an embedded video? That’s right, sit without a computer or go to a local flea market and buy a cheap used card for a while.

What if you have built-in graphics? Connect the monitor to it and use it further. Of course, you can’t play heavy games, but you can surf the Internet and play your favorite hits of the past. Without wasting time on trips and trips to the markets.

Q: Just about video cards. So I want to buy , which processor will fit it?

A: The processor has no characteristics that would prohibit it from working with certain video cards. As a rule, if the video card uses the PCI-e interface and is supported in the OS installed on the computer, this is all that is required of it.

In other words, if you are assembling a PC on a new platform, but there is not enough money to simultaneously update the video card, you can use the GPU left over from the previous system, or a budget solution of the old generation bought on the secondary market.

But the opposite is also true: in computers assembled on not the newest platforms, you can install video cards of the current generations if you lack the performance of the graphics part, or the budget allows you to replace only the video card.

The built-in benchmark of Assassin’s Creed: Odyssey claims that the test system combines the Intel Core i7-4930K processor, released in 2013 for the long-dated LGA 2011 platform, and the GeForce RTX 2080 Ti graphics card, released in late 2018, which is still relevant.