How many phases are needed to power the processor. Nutrition Phases - Quantity, Practical Benefits of Quantity

Now let's move on to an equally important part of any PC - motherboard.

1. The color of the motherboard is important, and it is best to take black

A funny myth with a very simple story: big vendors like Apple or Asus started painting their expensive boards black about 10 years ago. Of course, they broke less than simpler "colored" boards from competitors, hence the belief that "black goez fasta" began. In fact, the color of the board can be absolutely anything - yellow, green, white, blue, black - because this is a banal painting that in no way affects the internal characteristics of the textolite. So, for example, in the 90s, textolite was often not painted at all, and most of the boards - both expensive and cheap - had a dirty yellow color. So the difference between a black and white board is exactly the same as between a black and white iPhone- only in color and nothing more.

2. Heating the processor power circuits up to 90 degrees is a critical lot

Mosfets are highlighted in red - the hottest elements of the CPU power circuit.

Do not confuse the processor itself and its power circuits - indeed, for silicon CPUs, temperatures above 90-100 degrees are critical and will lead to quick exit him out of order. But this is not true for power circuits: for example, their hottest part - the so-called mosfets (insulated gate field-effect transistors) - have operating temperatures up to 150-175 degrees, so 90 degrees on them, of course, is a lot, but not critical. All other elements of the power circuits, such as capacitors and chokes, heat up much less and are often not covered by radiators because of this.

3. The internal peripherals on boards are always low quality and you need to buy them separately

A myth coming almost from the bearded 90s, when sound and network controllers on boards really left much to be desired. However, this is no longer the case now: 99% of boards are equipped with gigabit LAN controllers from Intel or Realtek, and given that the speeds home internet on average, an order of magnitude lower, there will be no problems with them.

Everything is a bit more serious with sound - now boards are mainly equipped with controllers from Realtek. Calling them audiophile is out of the question, but if you listen to music from streaming services and play games, there will definitely be no problems with sound quality.

4. All sorts of expensive boards with a bunch of ports and heatsinks are not needed, since even the cheapest solutions based on the Z370 chipset support my Core i9 - I will choose from them

Of course, there is always a desire to save money, and you can often take a cheaper board without, for example, built-in Wi-Fi or m.2 slots, saving up to a couple of thousand rubles. But, alas, further savings usually begin to affect the circuitry of the board - namely, manufacturers begin to reduce the number of CPU power phases on the board from 6-10 down to 3-4. Why is it scary? If earlier the energy needed to power the processor passed through 10 phases, heating them not very much, now it will pass through only 3 phases, due to which the heating will increase significantly. Plus here is the fact that cheap boards often don’t even have the simplest heatsinks on the power circuits, they can easily heat up to 120+ degrees with top processors under load, which is already critical for them:

In addition, various negative effects begin: for example, overheating protection may work, which will reduce the voltage on the processor, which means its frequency and performance. Weak power circuits may initially not provide the voltage necessary for the top processor to work under load, which again will negatively affect its frequency. So, alas, cheap boards are better left for simpler processors.

5. For high-end PCs, it is better to take full-size boards

The myth again comes from the beginning of the 2000s, when compact boards began to appear - then manufacturers, in pursuit of size, could really seriously cut down the functionality of such boards. But now there is no such thing - of course, mini-ITX boards have only one PCIe x16 slot and usually two slots for RAM, but all other parameters - even the ability to overclock processors and an m.2 slot with NVMe support - can be present, so there are no problems to build a top PC with a Core i9-9900K and RTX 2080 Ti in a case with dimensions slightly larger than those of consoles.

6. Reinforced PCIe slots and RAM - marketing, they are not needed

In the past few years, various manufacturers have begun to reinforce PCIe slots and even RAM, justifying this by saying that modern top-end video cards often weigh 1.5-2 kg, which can break the slot. However, here you need to understand a couple of things: firstly, this does not answer the question of why reinforcing RAM slots, since even with radiators, the dies hardly weigh more than a couple of hundred grams and definitely won’t break the plastic. Secondly, upon closer inspection, it will be seen that the reinforcement of the slot of the board itself does not touch, that is, the slots are still held only by their own contacts:

I think you are under the impression that I am contradicting myself and proving that reinforcement is really marketing. However, this is not entirely true: in reality, under the weight of a heavy video card, the narrow slot of the plastic PCIe slot can expand slightly, which will cause contact to be lost. Reinforcement will prevent this from happening - but, again, if you have a heavy video card, you should buy a special holder so as not to break the slot out of the board.

7. Mobile (SODIMM) RAM cannot be put into a desktop board (with DIMM slots)

On the face of it, it seems that this is not a myth - DIMM and SODIMM dies differ in size at times, so laptop RAM simply will not physically fit on a desktop board. But remember about SD-cards - they also come in different formats, however, with the help of an adapter, you can take a microSD and put it in a full-size slot, and it will work without any problems.


With RAM, everything is exactly the same: electrically, SODIMM practically does not differ from DIMM, so by purchasing the appropriate adapter, you can easily put laptop RAM into a computer, and it will work without problems. Of course, the question of the advisability of such a solution is in question, but if you have an extra RAM chip for laptops lying around and you have nowhere to put it, you can easily upgrade your PC with it.

8. If the processor power connector on the motherboard is 8 pin, then a 4 pin power supply will not work

It should be understood that the 8 pin power supply on the board is just 4 + 4 pins (this is hinted at by the fact that many power supplies have 8 pins just represented as 4 + 4), which are connected in parallel:


Accordingly, if you connect only 4 out of 8 pins, then in most cases the motherboard will work without problems. Of course, you should understand that you should not seriously load the processor with such a connection - the “extra” 4 pins are just created in order to reduce the heating of the wires from the PSU and the tracks in the textolite. But if, for example, you bought new board and CPU, but new block there wasn’t enough money from the 8 pin supply - it’s quite possible to “sit out” on the 4 pin.

9. If the processor is not supported motherboard, then nothing can be done, you need to change the board

Usually this is still not a myth, but recently there have been enough exceptions: for example, Xeon processors for the LGA771 server socket have become very popular, which are on various trading floors often given for several hundred rubles. And they, with some desire (cutting "ears" in a new place and soldering the conductor), can be put into ordinary desktop boards on LGA775:

Another exception is the LGA1151 and 1151v2 sockets: they differ mainly only in software, so with some “magic” with the BIOS, you can force 8th generation processors to work on officially unsupported motherboards with 100th or 200th chipsets.

10. Updating the BIOS is a complicated ritual that you should not do yourself

For some reason, for many, the phrase "updating the BIOS" causes panic fear and the idea of ​​​​a stern bearded computer scientist who conjures with floppy disks and prints some incomprehensible characters in command line. Fortunately, for the last 5 years this has not been the case for a long time - BIOSes often have a friendly user interface in Russian and support working with a mouse, and updating the BIOS is just a couple of mouse clicks, after which the necessary update will be downloaded from the Internet and installed itself.

There is also an opinion that if everything works, then you should not update the BIOS. Again, this is not the case, because often new BIOS versions have various security fixes (such as patches against Meltdown or Specter), which should not be ignored. And even more so if the board does not work correctly - which happens if you bought it right after the release - often it is the BIOS updates that will solve your problems.

11. All slots of the same type on the board are identical, you can use any

Not entirely true: so, usually only the PCIe slot closest to the processor can operate at maximum x16 speed, the slots below often work only in x8 or x4 mode, so you should not use them with fast video cards:

The same applies to SATA: if you use the m.2 slot at the same time with NVMe drive, then one of the SATA connectors may be disabled (since the number of PCIe lanes in the chipset is limited), so do not be surprised that after installing it in a computer fast SSD For some reason, your hard drive has stopped being detected.

12. XXX motherboards are better than YYY motherboards

In general, such a comparison is incorrect, just like with other types of technology. However, there are always brands that produce very low-quality products: for example, in laptops, these are Digma and iRU. There is a similar division among motherboard manufacturers.

So, MSI, Asus, Gigabyte (as well as Supermicro and Tyan in the server segment) are considered good manufacturers: again, this does not mean that their boards are perfect, but still they usually have the least problems. ASRock, Colorful, Biostar, ECS are considered mid-level manufacturers - perhaps it makes sense to compare them with Xiaomi smartphones: they seem to be cheaper than AAA brand solutions, but require some knowledge to set everything up as it should, and their BIOS at first can be raw .

The rest of the motherboards, usually Chinese (from Xuanan) or from OEMs, are often very problematic: they are whimsical about RAM, react incorrectly to buttons, can turn off during operation, etc. And, alas, you don’t have to wait for software fixes - OEMs don’t post them on the Internet at all, and you can get them only from new board revisions, and Chinese manufacturers usually “forget” about support.

13. Small boards (mATX, mini-ATX) cannot be placed in large cases (Full or Mid Tower)

The myth is again 20 years old, when compact motherboards were just beginning to appear, and there were simply no mounts for them in the cases. However, now even in the simplest "tin boxes" such mounts are present - another question is why take a spacious case and put a miniature board in it.

14. Boards for Intel processors better than AMD

The reason for the emergence of such a myth is quite understandable: usually at the start of sales with new AMD processors there are problems: for example, Ryzen was picky about RAM, and not all dies could work at least 3000 MHz. Processors from Intel are traditionally more stable in this regard, but, in any case, the problem here is precisely the software one: “hardware” boards of the same level for processors from Intel and AMD usually differ only in socket and chipset - they even look very similar.

15. For any manipulation with the board, you need to remove the BIOS battery

Do not confuse de-energizing the board (that is, pulling the power supply cord out of the socket) with removing the BIOS battery - the latter is only needed to save BIOS settings if there is a sudden power failure. Accordingly, the voltage from it goes only to the BIOS chip, so you can safely completely assemble the PC with the battery inserted. The only exception is if you need to reset the BIOS settings: in this case, logically, you need to get the battery.

As you can see, there are plenty of myths about motherboards. Do you know any more? Write about it in the comments.

Methodology and stand

In today's testing, a large number of computer technology to show how much power real-world gaming systems consume. In this regard, I relied on the assembly of the heading "Computer of the Month". A complete list of all components is given in the table below.

Test bench, software and accessories
CPU	Intel Core i9-9900K Intel Core i7-9700K Intel Core i5-9600K Intel Core i5-9500F AMD Ryzen 5 1600 AMD Ryzen 5 2600X AMD Ryzen 7 2700X
Cooling	NZXT KRAKEN X62
Motherboard	ASUS ROG MAXIMUS XI FORMULA ASUS ROG Crosshair VIII Formula ASUS ROG STRIX B450-I GAMING
RAM	G.Skill Trident Z F4-3200C14D-32GTZ, DDR4-3200, 32 GB Samsung M378A1G43EB-CRC, DDR4-2400, 16 GB
video card	2 × ASUS ROG Strix GeForce RTX 2080 Ti OC ASUS Radeon VII ASUS DUAL-RTX2070-O8G NVIDIA GeForce RTX 2060 Founders Edition ASUS ROG-STRIX-RX570-4G-GAMING AMD Radeon RX Vega 64 ASUS PH-GTX1660-6G
Storage device	Samsung 970 PRO MZ-V7P1T0BW
power unit	Corsair CX450 Corsair CX650 Corsair TX650M Corsair RM850x Corsair AX1000
Frame	Open test bench
Monitor	NEC EA244UHD
operating system	Windows 10 Pro x64 1903
Video card software
NVIDIA	431.60
AMD	19.07.2005
Additional software
Uninstalling drivers	display Driver Uninstaller 17.0.6.1
Measuring FPS	Fraps 3.5.99
	FRAFS Bench Viewer
	Action! 2.8.2
Overclocking and Monitoring	GPU-Z 1.19.0
	MSI Afterburner 4.6.0
Optional equipment
Thermal imager	Fluke Ti400
Sound level meter	Mastech MS6708
Wattmeter	watts up? PRO

Test benches were loaded with the following software:

• Prime95 29.8- Small FFT test, which loads the CPU as much as possible. A very resource-intensive application, in most cases, programs that use all the cores are not able to load the chips more.
• AdobePremierPro 2019- 4K video rendering by means of the central processor. An example of resource-intensive software that uses all processor cores, as well as available reserves random access memory and storage.
• The Witcher 3: Wild Hunt- testing was carried out in full screen mode in 4K resolution using maximum settings graphics quality. This game is very heavy not only on the graphics card (even two RTX 2080 Tis in an SLI array are 95% loaded), but also on the CPU. Eventually system unit is loaded more than, for example, with the help of "synthetics" FurMark.
• The Witcher 3: Wild Hunt +Prime95 29.8(Small FFT test) - a test for the maximum power consumption of the system when both the CPU and GPU are 100% loaded. And yet, it should not be ruled out that there are more resource-intensive bundles.

Energy consumption was measured using a watts up? PRO - despite such a comical name, the device can be connected to a computer, and with the help of special software it allows you to track it various options. So, the graphs below will show the average and maximum levels of energy consumption of the entire system.

The period of each power measurement was 10 minutes.

⇡ How much power do modern gaming PCs need

Once again, I note: this article is to a certain extent tied to the heading "Computer of the Month". Therefore, if you jumped into our light for the first time, then I recommend that you familiarize yourself with at least. Each "Computer of the Month" considers six assemblies - mostly gaming ones. I used similar systems for this article. Let's get acquainted:

• A bundle of Ryzen 5 1600 + Radeon RX 570 + 16 GB of RAM is an analogue of the starting assembly (35,000-37,000 rubles per system unit, excluding the cost of software).
• Bundle Ryzen 5 2600X+ GeForce GTX 1660 + 16 GB of RAM is an analogue of the basic assembly (50,000-55,000 rubles).
• A bunch of Core i5-9500F + GeForce RTX 2060 + 16 GB of RAM is an analogue of the optimal assembly (70,000-75,000 rubles).
• A bunch of Core i5-9600K + GeForce RTX 2060 + 16 GB of RAM is another optimal build option.
• A bundle of Ryzen 7 2700X + GeForce RTX 2070 + 16 GB of RAM is an analogue of an advanced assembly (100,000 rubles).
• A bunch of Ryzen 7 2700X + Radeon VII + 32 GB of RAM is an analogue of the maximum assembly (130,000-140,000 rubles).
• A bunch of Core i7-9700K + Radeon VII + 32 GB of RAM is another option for the maximum assembly.
• A bundle of Core i9-9900K + GeForce RTX 2080 Ti + 32 GB of RAM is an analogue of an extreme assembly (220,000-235,000 rubles).

Unfortunately, I was not able to get Ryzen 3000 processors at the time of all the tests, but the results obtained from this will not become less useful. The same Ryzen 9 3900X consumes less than the Core i9-9900K - it turns out that as part of an extreme assembly, it will be even more interesting and important to study the power consumption of an 8-core Intel.

And yet, as you may have noticed, only mass platforms are used in the article, namely AMD AM4 and Intel LGA1151-v2. I didn't use HEDT systems like TR4 and LGA2066. First, we have long since abandoned them in Computer of the Month. Secondly, with the advent of the 12-core Ryzen 9 3900X in the mass segment and on the eve of the imminent release of the 16-core Ryzen 9 3950X, such systems have become too highly specialized. Thirdly, because the Core i9-9900K still gives everyone a light in terms of power consumption, once again proving that the calculated thermal power declared by the manufacturer says little to the consumer.

Now let's move on to the test results.

To be honest, the results of testing in programs such as Prime95 and Adobe Premier Pro 2019, I give more for review - for those who do not play and do not use discrete graphics cards. You can safely rely on this data. Basically, here we are interested in the behavior of test systems under loads close to the maximum.

And there are some very interesting things going on here. In general, we see that all the systems considered do not consume very much energy. The most gluttonous, which is quite logical, was the system with Core i9-9900K and GeForce RTX 2080 Ti, but even in stock (read - without overclocking) it consumes 338 watts when it comes to games, and 468 watts - at maximum PC load. It turns out that such a system will have enough power supply for an honest 500 watts. It is so?

⇡ It's not just about watts

It would seem that this article can be completed: recommend to everyone a power supply unit with a capacity of 500 honest watts - and live in peace. However, let's do a few more experiments to get a complete picture of what's going on with your PC.

In the screenshot above, we see that the power supplies operate as efficiently as possible at 50% load, that is, half of the declared power. It may seem to someone that the difference between a device with a basic 80 PLUS certificate with an efficiency of about 85% at the peak in a 230 V network and, say, a “platinum” PSU with an efficiency of about 94% is not so great, but this is a misconception. my colleague Dmitry Vasilyev points out quite accurately: “An energy source with an efficiency of 85% wastes 15% of its power on heating the surrounding air, while for a “breadwinner” with an efficiency of 94%, only 6% of the power goes into heat. It turns out that the difference is not some there"10%, but x2.5". Obviously, in such conditions, a more efficient power supply is also quieter (it makes no sense for the manufacturer to set the device's fan to the maximum speed), and it heats up less.

Here is the evidence for the above.

The graphs above show the efficiency of some power supplies participating in the tests, as well as the rotational speed of their fans at different degrees of load. Unfortunately, the equipment used does not allow us to accurately measure the noise level, but by the number of revolutions per minute of the built-in fans, we can judge how noisy the power supply will be. It must be noted here that this does not mean at all that under load the PSU will stand out “from the crowd”. Yet usually the noisiest components gaming computer are the CPU cooler and the video card.

Practice, as you can see, converges with theory. Power supplies do operate at their maximum efficiency at about 50 percent load. Moreover, in this regard, I would like to note the Corsair AX1000 model - this PSU reaches its peak efficiency at a power of 300 W, and then its efficiency does not fall below 92%. But other Corsair blocks on the charts have the expected “hump”.

While Corsair AX1000 can operate in semi-passive mode. Only at a load of 400 W does its fan start spinning at a frequency of ~ 750 rpm. The RM850x has the same characteristic, but in it the impeller begins to rotate at a power of ~ 200 W.

Now let's look at temperatures. To do this, I dismantled all the power supplies. Fans with top cover were removed and mounted on a homemade tripod so that the distance between it and the rest of the PSU was approximately 10 cm. I am sure that the device did not become worse in terms of cooling, but this design allowed me to take pictures with a thermal imager. In the graph above, "Temperature 1" refers to the maximum temperature of the power supply inside when the fan is running. "Temperature 2" is the maximum heating of the PSU ... without additional cooling. Please do not repeat such experiments at home on your own equipment! However, such a bold move allows you to clearly show how the power supply heats up and how its temperature depends on the rated power, build quality and the component base used.

Heating the CX450 model up to 117 degrees Celsius is quite a logical phenomenon, because this power supply at a load of 400 W works almost at its maximum, and even does not cool in any way. The fact that the power supply passed this test at all is a great sign. Here is a quality budget model.

Comparing the results of other power supplies, we can conclude that they seem quite logical: yes, the Corsair CX450 model heats up the most, and the RM850x the least. At the same time, the difference in maximum heating rates is 42 degrees Celsius.

Here it is important to define the concept of "honest power". Here is a Corsair CX450 model that can transmit 449 watts of energy over a 12-volt line. It is this parameter that you need to look at when choosing a device, because there are models that do not work as efficiently. In cheaper units of similar power, noticeably fewer watts can be transmitted over a 12-volt line. It comes to the point that the manufacturer claims to support 450 W, but in fact we are talking only about 320-360 W. So let's write it down: when choosing a power supply, you need to look, among other things, at how many watts the device produces on a 12-volt line.

Let's compare the Corsair TX650M and CX650 models, which have the same declared power, but are certified according to different 80PLUS standards: "gold" and "bronze", respectively. I think the thermal images attached above speak louder than any words. Really, support for a specific standard 80PLUS indirectly speaks about the quality of the element base of the power supply. The higher the certificate class, the better the power supply.

It is important to note here that the Corsair TX650M model transmits up to 612 watts over the 12-volt line, and up to 648 watts on the CX650.

Above in the pictures you can compare the heating of the RM850x and AX1000 models, but already at a load of 600 watts. Here, too, there is an obvious difference in temperature. In general, we see that Corsair PSUs cope well with the load assigned to them - and even in stressful situations. At the same time, I think it is now clear why there were no temperature indicators for the AX1000 on the graph above - it does not heat up much, even if you remove the fan cover from it.

Pondering the results obtained, you can see that it will be completely shameless to use a power supply unit with a power twice the maximum power of the PC itself in the system. In this mode of operation, the PSU heats up and makes less noise - these are facts that we have just proven once again. It turns out that a PSU with an honest power of 450 W is suitable for the starting assembly, 500 W for the basic one, 500 W for the optimal one, 600 W for the advanced one, 800 W for the maximum one, and 1000 W for the extreme one. Plus, in the first part of the article, we found out that there is not such a big difference in price between power supplies, the declared power of which differs by 100-200 watts.

However, let's not rush to final conclusions.

⇡ A few words about the upgrade

Builds in the "Computer of the Month" are designed not only to work in the default mode. In each issue, I talk about the overclocking capabilities of some components (or the pointlessness of overclocking in the case of some processors, memory and video cards), as well as the possibilities for subsequent upgrades. There is an axiom: the cheaper the system unit - the more compromises it has. Compromises that will allow you to use a PC here and now, but the desire to get something more productive, quiet, efficient, beautiful or comfortable (underline as necessary) will not leave you anyway. Captain Evidence suggests that in such situations, a power supply with a good margin of watts is very useful.

I will give a clear example of upgrading the starting assembly.

I took the AM4 platform. Recommended 6-core Ryzen 5 1600, Radeon RX 570 and 16GB DDR4-3000 RAM. Even when using a stock cooler (the cooling system that comes with the CPU), our chip can easily be overclocked to 3.8 GHz. Let's say I took a radical step and changed the CO to a much more efficient model that allowed me to increase the frequency from 3.3 to 4.0 GHz with all six cores loaded. To do this, I needed to raise the voltage to 1.39 V, and also set the fourth level Load-Line Calibration of the motherboard. This overclock essentially turned my Ryzen 5 1600 into a Ryzen 5 2600X.

Let's say I bought a Radeon RX Vega 64 video card - on the Computeruniverse website a month ago it could be taken for 17,000 rubles (excluding shipping), and even cheaper by hand. And in the comments to the "Computer of the Month" they talk so sweetly about the used GeForce GTX 1080 Ti, sold for 25-30 thousand rubles ...

Finally, instead of the Ryzen 5 1600, you can take the Ryzen 2700X, which, after the release of the third generation AMD chip family, has noticeably fallen in price. There is no particular need to overclock it. As a result, we see that in both cases of the upgrade I suggested, the system power consumption more than doubled!

This is just an example, and the actors in the situation described may be completely different. However, this example, in my opinion, clearly shows that even in the starting assembly, a power supply unit with an honest power of 500 W, and even better, even 600 W, will not interfere at all.

⇡ Overclocking and everything connected with it

Since we are talking about overclocking, I will give an example of the power consumption of stands before and after overclocking. Frequencies have been increased for the following systems:

• Ryzen 5 1600 (@4.0GHz, 1.39V, LLC 4) + Radeon RX 570 (1457/2000MHz) + 16GB RAM (DDR4-3200, 1.35V).
• Ryzen 5 2600X (@4.3GHz, 1.4V, LLC 4) + GeForce GTX 1660 (1670/2375MHz) + 16GB RAM (DDR4-3200, 1.35V).
• Core i5-9600K (@4.8/5.0GHz, 1.3V, LLC 4) + GeForce RTX 2060 (1530/2000MHz) + 16GB RAM (DDR4-3200, 1.35V).
• Ryzen 7 2700X (@4.3GHz, 1.4V, LLC 4) + GeForce RTX 2070 (1500/2000MHz) + 16GB RAM (DDR4-3200, 1.35V).
• Ryzen 7 2700X (@4.3GHz, 1.4V, LLC 4) + Radeon VII (2000/1200MHz) + 32GB RAM (DDR4-3400, 1.4V).
• Core i7-9700K (@5.0/5.2GHz, 1.35V, LLC 5) + Radeon VII (2000/1200MHz) + 32GB RAM (DDR4-3400, 1.4V).
• Core i9-9900K (@5.0/5.2GHz, 1.345V, LLC 5) + GeForce RTX 2080 Ti (1470/1980MHz) + 32GB RAM (DDR4-3400, 1.4V).

“Gaming PCs don’t need 1 kW blocks” - commentators under articles on the site

Comments like this are common when it comes to gaming PCs. In the vast majority of cases - and we found this out in practice - this is the case. However, in 2019 there is a system that can impress with its power consumption.

This, of course, is about extreme assembly in its, so to speak, maximum combat form. Not so long ago, an article "" was published on our website - in it we talked in detail about the performance of a pair of the fastest GeForce video cards in 4K and 8K resolution. The system is fast, but the components are chosen in such a way that it is very easy to make it even faster. In addition, it turned out that overclocking the Core i9-9900K to 5.2 GHz is not at all superfluous in the case of the GeForce RTX 2080 Ti SLI array and games in Ultra HD. Only at the peak, as we can see, such an overclocked configuration consumes more than 800 watts. Therefore, for such a system in such conditions, a kilowatt power supply will definitely not be superfluous.

⇡ Conclusions

If you carefully read the article, you have identified for yourself a few main points that you need to keep in mind when choosing a power supply. Let's list them again:

• Unfortunately, it is impossible to focus on the TDP indicators declared by the manufacturer of the video card or processor;
• the power consumption of computer equipment does not change much from year to year and is within certain limits - therefore, a high-quality power supply purchased now will serve for a long time and will definitely come in handy during the assembly of the next system;
• the needs for cable management of the system unit also affect the choice of a PSU of a certain power;
• not all power connectors on the motherboard need to be used;
• not always a power supply of lower power is more profitable (in terms of price) of a more powerful model;
• when choosing a power supply, you need to look, among other things, at how many watts the device produces on a 12-volt line;
• support for a certain 80 PLUS standard indirectly indicates the quality of the element base of the power supply;
• It is completely shameless to use a power supply whose honest power is twice (or even more) the maximum power consumption of the computer.

Quite often you can hear the phrase: More is not less". This very laconic aphorism perfectly describes the situation when choosing a power supply. Take a model with a good supply of power for your new PC - it will definitely not be worse, and in most cases it will only be better. Even for an inexpensive gaming system unit that consumes about 220-250 W at maximum load, it still makes sense to take good model with honest 600-650 watts. Because this block:

• will work quieter, and in the case of some models - absolutely silently;
• it will be colder;
• will be more efficient;
• will allow you to safely overclock the system, increasing the performance of the central processor, video card and RAM;
• allows you to easily upgrade the main components of the system;
• will survive several upgrades, and also (if the power supply is really good) will settle in the second or third system unit;
• will also save money during the subsequent assembly of the system unit.

I think few readers will refuse a good power supply. It is clear that it is not always possible to immediately buy a high-quality device with a big reserve for the future. Sometimes, when buying a new system unit and on a limited budget, you want to take a more powerful processor, faster video cards, and a higher capacity SSD - all this is understandable. But if there is an opportunity to buy a good power supply with a margin, there is no need to save on it.

We are grateful to companiesASUS andCorsair, as well as to the computer store "Regard" for the equipment provided for testing.

#Number_of_lines_+12V

You can independently identify how many lines in a particular PSU can be on its label - if there are more than one lines, then the maximum load in amperes is separately indicated for each + 12V circuit, which are designated as "+ 12V1, + 12V2, etc." Actually, the output lines are called "rails" in English, and, accordingly, a power supply unit with one output line will be called "single rail PSU", and with several - "multiple rails PSU".

PSU with one line + 12V

PSU with multiple +12V lines

There are several PSU models that actually have two +12V voltage sources, but these are usually very high power PSUs (from 1000W). And in most cases, these two exits are again divided into four, five or six lines for security reasons. (But, for example, they don’t share, and this is not so bad, which will be discussed further)

In some even rarer cases, the two original +12V lines can be combined into one powerful output.

So why do you really need to separate the +12V lines?

Safety. For the same reason, houses tend to have more than one fuse switch (popularly known as "bags"). The ultimate goal is to limit the current in one circuit to a value of 20A so that the temperature of the conductor carrying it does not become dangerous.

Short-circuit protection only works when there is practically no resistance in the shorted circuit (i.e., for example, when a bare wire hits ground), and in more complex cases, when a short circuit occurs on printed circuit board or in the motor, the resistance in the circuit remains sufficient so that the short-circuit protection does not trip. In this case, there is a very huge pressure on the circuit and a rapid increase in the current strength in the conductors leads, first of all, to the melting of the insulation and, subsequently, to a fire. Current limitation in each line eliminates this problem, i.e. this explains the need to divide the outputs into separate lines with individual limiters.

Is it true that in some PSUs with multiple +12V lines claimed there is no line separation at all?

Yes it is. Fortunately, this is the exception to the rule, not the norm. This is done to reduce development and production costs. Why is it stated that there are several lines - in order to fully comply with the ATX12V specification, because in other characteristics it is observed.

Why do such PSUs remain on the market, and manufacturers have no problems with their certification?

Yes, because Intel recently removed the requirement to separate +12V lines from the specification, but did not publicize this fact widely. They just changed "required" to "recommended", leaving manufacturers slightly puzzled.

Does splitting the +12V lines give "cleaner and more stable voltages"?

The truth is that marketers constantly emphasize this fact, but usually it is not, it just seems more euphonious than "This PSU is unlikely to cause a fire." And since, as mentioned above, all lines in most cases originate from the same source, and no additional filtering is performed, the voltages remain the same even if there were no division.

Why do some people take it upon themselves to argue that a PSU with a single +12V output is better?(just a great example - )

There have been several companies that have made 4-line 12V PSUs that in theory should be able to supply more than enough current for a high-end gaming station and have run into a lot of problems. When making the PSU according to the EPS12V server specification, all PCI-E 6-pin connectors were taken out of the common +12V lines with a load capacity of 18A, instead of a separate one. This line was easily overloaded with two powerful video cards along with other possible consumers, which led to a PC shutdown. Instead of a "civilized" solution to the problem, these manufacturers generally abandoned the division of + 12V outputs.

Now "enthusiast" PSUs with several +12V lines either have an overestimated maximum line load capacity for PCI-E connectors (and nothing else is connected to it), or two such lines are distributed over four or even six connectors. And PSU certification for SLI in any case requires a minimum of a separate + 12V line for PCI-E connectors.

It costs $1.5 - $3 more to make a split-line PSU for the manufacturer, and in most cases this amount is not passed on to the buyer, which already forces marketers to put forward theories that a +12V PSU without line split is no worse and even better .

But nevertheless, there are statements that, for example, a PSU with one + 12V line is better suited for overclocking, etc. But this is more like a placebo effect, which arose due to the fact that, for example, their previous PSU was faulty, was not powerful enough, or the load was not properly distributed along the lines.

So it turns out that a PSU with +12V load distribution over several lines does not have any specific drawbacks?

No, actually, it's not. Consider two examples:

Example #1:

One PSU model with a nominal value of 700W formally has enough power for any SLI system of two single-chip video cards. But this PSU has only two PCI-E connectors, each of which hangs on its own + 12V line. The problem is that these lines are capable of delivering 18 amps of current, which is almost three times the maximum current that a 6-pin PCI-E video card connector is designed to handle. Accordingly, when you try to install two video cards that require two of these connectors, problems begin.

It would be ideal if two connectors were soldered on each of the lines, but instead you have to use adapters from the "regular" 4-pin Molex to PCI-E 6-pin, which leads to an overload of the circuits that power the rest of the system block, while the actual "video card" circuits remain heavily underloaded. The problem could be solved by a 6-pin PCI-E -> 2x 6-pin PCI-E adapter in duplicate, but it cannot be called common. So in a situation like this best solution Problems (in addition to replacing the PSU) remain independent soldering of two PCI-E connectors to two corresponding lines.

Example #2:

Thermoelectric coolers (also called Peltier coolers) consume a lot of power and are usually powered by Molex connectors. Some models generally use their own separate PSU.

So, if you use a PSU with line separation and power your Peltier element from one of the molexes, then it ends up on the same line with drives, fans, etc., then this line can also be overloaded, since it can be transplanted to other lines, designed to power video cards is impossible without significant tweaks. Naturally, a PSU with one +12V line would be devoid of any problems in such a situation.

Typical configurations for multiple +12V lines:

• 2 x 12V lines, example -
  This is the original ATX12V specification for dividing +12V lines. One is for the processor, the other is for everything else. It is highly unlikely that a modern high-end video card with high power consumption can fit into the number of "everything else". Such a division could only be seen on a PSU with a power of less than 600W.
• 3 x 12V lines, example -
  Modifications to the ATX12V specification, taking into account the use of PCI-E connectors for powering video cards. One line per processor, one for PCI-E connectors and a third for everything else. Works fine even with some SLI configurations, but not recommended for dual graphics cards requiring four PCI-E connectors in total.
• 4 x 12V lines (EPS12V), example -
  In the original, this configuration was required by the EPS12V specification. Since typical applications of such PSUs involve their use in dual-processor systems, the two +12V lines are dedicated exclusively to powering processors through 8-pin connectors. Everything else, including drives and video cards, falls on the remaining two lines. Currently, nVidia does not certify such PSUs for SLI, since there is no separate +12V line for video cards in such PSUs. There will no longer be such PSUs in the PSU segment not intended for servers, several 700-850W models made according to this architecture for the gaming PC market have already been discontinued.
• 4 x 12V lines (Most popular layout in the "enthusiast PC" segment), example -
  An "upgraded" ATX12V, similar to 3 x 12V, except for the fact that two to six PCI-E connectors are shared between two additional +12V lines. Such a scheme is most often found in power supplies from 700 to 1000 watts, although with a power of 800 watts or more, some of the lines may have much more than 20 amps, which is not quite standard, but seems to have already become common practice, for example -
• 5 x 12V lines e.g. -
  Such PSUs can be called an EPS12V/ATX12V hybrid. Two processors with their own power lines, also two lines go to PCI-E connectors. The power of such power supplies is usually from 850 to 1000 watts.
• 6 x 12V lines, example -
  most attractive and universal option, since it, meeting the requirements of the EPS12V specification, can have four or six PCI-E connectors without exceeding the current of 20A on any of the lines (although in practice this limitation, as you have already seen, is interpreted quite freely). Two lines go to processors, two to video cards, two to everything else. This configuration can be seen in a PSU with a power of 1000W or more.

As a conclusion, you can see the fact that 99% of users will never think about whether their power supply has a common or separate + 12V line. Perhaps marketers will continue to praise the merits of both options, but the criteria for buying a PSU will still remain the same:

• Sufficient power for the selected configuration.
• A sufficient number of suitable connectors for the selected configuration.
• SLI or CrossFire certified when using the appropriate MultiGPU configuration.

Processor power phases (processor power phase) - a quantitative characteristic indicating the number of supply phases on the motherboard intended for the processor (this also affects, but in that case the printed circuit board is not the motherboard).

For what?

In theory, the more quantity per phase, less heating and more stable nutrition in load surges, as well as higher durability. That is for overclocking processor, a large number of phases - just necessary. After all, the load on the phases increases significantly and high stability is required to achieve maximum results.

How to visually determine the number of phases?

The number of power phases for a processor or video card can be determined by the inscription on the product box or printed circuit board, or by the number on the board.

Chokes look like they are wrapped around ferrite or just copper wires, twisted into coils having a rather thick section. More often they are packed in small boxes in the form of a rectangular box. parallelepiped For decrease quantities losses, interference And AMY. Two or one of these boxes should be a little apart - these are phases for power, they do not need to be counted. Coils are either in groups or together.

Trick

Not always number of chokes and the words on the box are real quantity real phases. It happens that the manufacturer uses doublers and forms half virtual phases(best case scenario).

To accurately determine the number of phases, you need to look at characteristics of the VRM module And -controller. Phases virtual, at best provide 30% the characteristics that the real ones give. It often happens that the power phases, for example 24 but really real 12 or 6 , but using doublers and triplers. That is, they can be considered as "improved" 12 or 6 phases, but not 24.

What number of power phases per MP can be considered optimal?

Motherboard Division Intel claims to work 4 core processor without overclocking is enough 4-phase. Also, according to them, properly designed 4 phases nutrition with quality ingredients, often win for food stability at improperly designed 16 phases nutrition. For an overclocked multi-core processor, it is enough 8 full phases processor power, or 16 phases, which uses the division-by-2 method, resulting in 8 full-fledged improved phases. It also follows that the number outages, polyphase circuits in the lead due to design complexity And a large number used components.

Phase switching technologies

(power phase switching)

These technologies, built on special controllers, measure how much power does the processor need this moment, and enable or disable blocks with phases. This allows increase the durability of work equipment, reduce power consumption And AMY. Implemented very often indication included phases on the motherboard and even the degree of load on them (as in the image above).