What is the power of a computer measured in? Processor performance and how it is measured

Good afternoon, dear guests and regular visitors of my blog. Today we will talk about one sore topic, namely speed, or how processor performance is measured.

I want to say right away that this is not a frequency per core, as was previously accepted, but a combination of several mathematical quantities at once, referred to as FLOPS (FLoating-point Operations Per Second) - an off-system unit of performance.

What determines the computing power of a computer, and is it worth paying attention to the frequency indicator? In all this we will try to figure it out.

Where do legs grow from?

Quite often on the Internet you can find disputes that "Intel is being dragged due to the higher frequency of the cores." In other words, the frequency parameter is put at the head of the table, and the rest of the nuances (the number of threads, cache size, work with certain instructions and the technical process) are forgotten for some reason.

Until about the beginning of the 2000s, such a comparison took place, since the characteristics of the central chip and its speed rested precisely on the frequency. Suffice it to recall the following names:

Pentium 133 and 333;
Pentium 800 etc.

And then the situation changed dramatically, as developers began to devote more time to building the internal architecture of the chips, adding cache memory, support for new instructions, calculation methods, and other elements that increase performance without increasing the same frequency.
New speed criteria have appeared in the arena:

cache memory;
data bus frequency;
bit depth.

Those. it became almost impossible to determine the capabilities of a chip based on the frequency potential alone.

What affects the performance of modern processors?

So, let's get acquainted with the concepts that characterize the operation of the processor, the speed of calculations and all other parameters.

Bit depth– determines the amount of data processing per cycle. On this moment there are both 32-bit and 64-bit variants. Imagine that the data size is 1 byte (8 bits). If the chip calculates 4 bytes of information per run, it is 32-bit, if 8 bytes it is 64-bit.

The logic is elementary to disgrace: when comparing 2 CPUs with identical frequencies and different bit depths, the one with a 64-bit logic set will win (the difference ranges from 10 to 20%).

Good day and my respect, dear readers, visitors, passing personalities and .. in general, everyone who reads these lines. Today we'll talk about which processor to choose and how to do it.

Many of us want to always have an adequate computer piece of hardware at hand. good quality and powerful power, and even at an affordable price.

However, despite our wishes, not everyone (I would even say, a few) are able to immediately name all the main criteria for choosing one or another computer component. And if they somehow manage with the video card and something else, then when it comes to the brain of everything and everything, namely, CPU, then this is where the absolute ambush begins.

Therefore, we once again (because, as many remember, there were already articles on choice, and much more) decided to lend a helping hand to all those in need and talk about how to choose the right processor, namely, what you need to know, what to look for attention to what characteristics are there and so on.

In general, today we are waiting for an article from the series: “I want to buy a processor, but I don’t know what to look for .. Tell me?”.

In short, sit back and.. Let's go!

Which processor to choose - main characteristics

As I said, the article will be as practical as possible, so we will not rant for a long time about what a CPU is and what it is for, but we will immediately rush off the bat.

We have already somehow touched on processor topics in articles such as and, however, questions are constantly pouring in from readers, they say, give a clear guide on what and how to buy.

And since the project is, so to speak, social (we take into account the "Wishlist" of visitors), without thinking twice, we decided to consecrate this issue in as much detail as possible.

Note:
Very often one has to deal with a situation where users buy various fancy and expensive ones in the hope that everything will fly and run right away, but the processor is not given due attention, after which it slows down the entire system, because it simply cannot provide all the necessary agility and nimbleness to everyone other operating subsystems and components.

Therefore, knowledge of the main parameters is necessary first of all in order to assess the realistically possible computational performance of the future system. It turns out that focusing on the characteristics of the processor, you will be able to maximize the full potential of all the components of your computer counterpart.

Actually, here's what you have to decide when choosing a processor:

Manufacturer's brand (Intel or AMD);
Technical process of production;
Marking and architecture;
CPU platform or socket type (socket);
Processor clock speed;
Bit depth;
Number of Cores;
Multithreading;
cache memory;
Power consumption and cooling;
Branded bells and whistles of technology.

Conclusion . Which processor to choose based on this? If you are a supporter of all laptops and the like portable devices, then TDP and all sorts of fans there should not be turned special attention- there and so everything is already calculated and installed for you. If you want to assemble a high-performance desktop system, then you need to take a serious "cooler".

Integrated graphics core

With the development of the process technology for the production of processors, it became possible to place various microcircuits inside the CPU, in particular the graphics core.

This solution is convenient because you do not need to buy a separate video card. It is focused mainly on the budget sector (office environment), where the graphical capabilities of the system are secondary. AMD embeds in its computing processors Radeon HD video chips, such a single element was called APU (accelerated processing unit).

Conclusion . Which processor to choose based on this? If your goal is a budget computer in which graphics do not play an important role (well, you do not play powerful games, do not do 3D design, etc., etc., but just watch movies, surf the Internet, etc., etc.), then a hybrid processor with an integrated video core is what the doctor ordered, so to speak cheap and cheerful. If you need video power, then, of course, there is no point in spending on a processor with a video core - better.

All sorts of branded technologies

For such a long time of the existence of processors, their manufacturers have acquired their own "gadgets" - additional features, accelerating and expanding the computing power of the CPU. For example, here are some of them.

From AMD:

3DNow!, SSE (instructions) - acceleration of work in multimedia computing;
AMD64 - work with 64-bit instructions, as well as with 32-bit architectures;
AMD Turbo Core - analogue of Intel turbo boost ;
Cool'n "Quiet - reducing power consumption by reducing the multiplier and core voltage.

From Intel:

Hyper Threading (hyperthreading) - the creation of two virtual (logical), computing cores for each physical core;
Intel Turbo Boost - increasing the frequency of the CPU, depending on the workload of the cores;
Intel Virtualization Technology- run multiple operating systems at the same time without loss of performance.

Conclusion . Which processor to choose based on this? Of course, additional “goodies” in the form of proprietary technologies are not something to be based on when choosing a CPU, but no one bothers you to get them for free as a pleasant bonus, the main thing is to decide what is needed.

So the last one for today is...

Processor marking

It is very important to be able to read and correctly interpret the processor markings, because stores are different, sellers are not always honest, but it’s unlikely that anyone wants to pay extra N-thousand rubles for an incomprehensible “stone”, and therefore it is important to be able to read processor markings. Let's break it down into specific example, say, for the manufacturer AMD .

IN general view marking from AMD (for generation Family 10h) can be represented as follows (see image):

The decryption will be as follows:

Processor brand (1). The following characters are possible:

A- AMD Athlon;
H - AMD Phenom;
S - AMD Sempron;
O - AMD Optheron.

Processor assignment (2). Options:

D - desktop - for workstations or desktop PCs;
E - embedded server - for dedicated servers;
S - server - for servers.

Processor model (3). Possible designations:

E - energy efficient processors;
X – locked multiplier;
Z is the unlocked multiplier.

Thermal package and cooling system class (4). The data is taken from the table (see image):

Processor case (5). The data is taken from the table (see image).

Number of cores (6). Values from 2 to C (12).

The amount of cache memory (7). Data from table (see image).

Processor revision or stepping (8). Data from table (see image).

So, based on the data in the table, we can easily determine what kind of processor we have in front of us, for example, judging by the model below (see image), we have ..

AMD processor labeled HDZ560WFK2DGM , which stands for:

H - AMD Phenom family CPU;
D - purpose: workstations / desktop PCs;
Z560 - processor model number 560 (Z - with a free multiplier);
WF - TDP up to 95 W;
K - the processor is packaged in a 938 pin OµPGA package (Socket AM3);
2 is the total number of active nuclei;
D - 512 KB L2 cache and 6144 KB L3 cache;
GM is the core of the C3 stepping processor.

So, knowing the credentials of the tables, you can easily figure out what kind of instance is in front of you.

Actually, this is all I would like to tell. I think that the information will be useful for you and will come in handy more than once.

Where is the best place to buy a processor?

Days to change the goods without any questions, and in case of warranty problems, the store will take your side and help solve any problems. The author of the site has been using it for at least 10 years (since the time when they were part of Ultra Electoronics), which he advises you;

, - one of the oldest stores on the market, as a company has been around for about 20 years. Decent selection, average prices and one of the most user friendly sites. In general, a pleasure to work with.

The choice is traditionally yours. Of course, no one canceled Yandex.Market there, but from good shops I would recommend these, and not some MVideos and others large chains(which are often not just expensive, but flawed in terms of quality of service, warranty work, etc.).

Afterword

Today we found out in as much detail as possible which processor to choose and how to do it correctly, i.e. what you can pay attention to when buying it.

The information is quite specific and technically, perhaps difficult and unusual for some, so if you haven’t learned something, then re-read it again, and then again, then open the price list and try to make several options for choosing processors for different needs.

Then read again, then select again. In general, and so on in a circle until you fill your hand :)

We have fulfilled our good mission, so it's time to say goodbye for a while.
As always, if you have any questions, additions, thanks and all that, feel free to write comments.

P.S. For the existence of this article, thanks to a member of the team 25 FRAME

I have been repeatedly asked the question - how much power does a computer consume? Such a question is usually interesting from two points of view: firstly, to choose a suitable power supply so that, on the one hand, you don’t overpay for excess power, but, on the other hand, you don’t end up with a computer barely working on a weak PSU; secondly, this question is not so rarely asked in order to calculate the impact of a 24-hour computer on the family budget.

This article presents the results of measurements of the energy consumption of several enough typical configurations computers, and at the same time, the properties of power supplies associated with their power consumption from the mains are also investigated.

Theoretical Introduction

In chains alternating current It is customary to distinguish four types of power. First, this instant power(instantaneous power) - the product of current and voltage at a given time. Secondly, this is the so-called active power(active power, average power) - the power released on a purely resistive load, it is measured in watts - W. Active power goes entirely to useful work(heating, mechanical movement), and usually it is she who is understood as the power consumption. The active power is calculated through the integral over one period from the instantaneous power:

Since the real load usually also has inductive and capacitive components, then the active power is added reactive(reactive power), measured in reactive volt-amperes - VAR. The reactive power is not consumed by the load - received during one half-cycle of the mains voltage, it is completely returned to the network during the next half-cycle, only loading the supply wires in vain. Thus, the reactive power is completely useless, and, if possible, they fight with it using various corrective devices.

The vector sum of active and reactive powers gives full power(apparent power) - accordingly, the square of the apparent power is equal to the sum of the squares of the active Pact and reactive Q capacities:

In practice, however, the apparent power is not calculated in terms of reactive and active, but as the product of Root Mean Squared (RMS) current and voltage:

In turn, the rms values are calculated as the square root of the integral over one period of the square of the value:

Everyone is familiar with the voltage of 220V in the lighting network - this is just the root mean square value. However, it should be noted here that most measuring instruments shows RMS values only if the voltage or current waveform is sinusoidal. In other words, let's say, a pointer voltmeter is simply graduated so that on a sinusoidal voltage the value shown by it something equal to the root mean square value; if the voltage differs from sinusoidal, then the voltmeter will show exactly something. And since in impulse blocks power supply, not equipped with power factor correction circuits (Power Factor Correction - PFC), the current consumption is very far from sinusoidal, then to measure the RMS current it is necessary to use the so-called TrueRMS devices that honestly integrate the measured value - otherwise the measurement error will be very large. For example, we used a UT-70D multimeter from Uni-Trend to control voltage and current:

However, full power is not enough to complete the picture; active power is also needed. To measure it, we used an ETC M-221 digital oscilloscope, which, being connected to a shunt through which the power supply under study was powered, took voltage and current oscillograms. Thus, we get the functions U(t) And I(t). More precisely, not the functions themselves, but a table of their values - therefore, we move from integration to summation:

Here N- the number of readings per one period of the mains voltage. To facilitate the calculations, a simple program was written that reads the data files saved by the oscilloscope from the disk (it saves them in its own format, so it was obviously impossible to process the data, say, in Excel) and calculates all the values \u200b\u200bthat might be of interest to us - the total and active power, rms current and voltage, unit efficiency (for this, of course, the load on the unit must be known) and power factor - the ratio of active power to apparent power.

Power supplies

The first part of the experiment on measuring the power consumed by computers is the study of the operation of power supplies with an artificial load. The same setting was used as the load as when testing the power supplies - this made it possible to load the unit under study to any allowable power, from zero to the maximum possible for this unit.

The experiment involved three different power supplies - 250W FSP250-60GTA from Fortron/Source Technology Inc. (FSP Group), 300W DPS-300TB-1 from Delta Electronics Group and 460W HP2-6460P from Emacs / Zippy Technology Corp.. If the first two blocks are undoubtedly already familiar to readers, then I will briefly tell you about the last one - this block is supplied as part of the Chenbro Group server cases and is powerful block food very High Quality, designed for entry-level servers. It differs from the first two blocks not only in maximum power, but also in the presence of active PFC.

During the experiment, a load with power from 25W to 250, 300 or 400W (depending on the power supply) was connected to the blocks, and oscillograms of the mains voltage and current consumed by the PSU were taken. Further, based on the oscillograms, the total and active powers, the efficiency of the power supply and the power factor were calculated.

It can be seen that the efficiency of all three blocks at minimum power is about 60%, but it quickly grows with increasing load (especially for the HP2-6460P block) and already at a load of 50-60W it reaches 68% set by the ATX / ATX12V Power Supply Design Guide (section 3.2.5.1 of the document). For the first two blocks - FSP250-60GTA and DPS-300TB-1 - the efficiency is approximately the same and at the maximum it is about 80%, while for the HP2-6460P it is noticeably higher and reaches a record 94% at a power of 200W.

The definition of efficiency was not an end in itself - in the future, when measuring the power consumed by real computers, knowledge of the efficiency will be required to recalculate the power consumed from the network to the power consumed by the actual computer stuffing.

The power factor is the ratio of active power to apparent power. Since the difference between these two powers is due to reactive power, which does not carry any benefit, then ideally the active power should be equal to the apparent power and, accordingly, the power factor should be equal to one. The practical benefit of this will primarily be felt by the owners of UPS, whose maximum output power is measured exactly in volt-amperes, not watts - the total power consumed by the same system can be reduced by a quarter only through the use of power factor correction circuits.

The graph above shows that for units that are not equipped with any correction circuits, the power factor is in the range of 0.65-0.7, weakly depending on the load; passive PFC, used in the DPS-300TB-1 block, helps rather weakly - the power factor increases to 0.7-0.75, but no more. For a power supply with active PFC - HP2-6460P - everything looks different: if at low powers the power factor for it is 0.75, then already at a power of 200W it reaches 0.97, and at a power of 400W - up to 0.99 .

On the oscillograms, it looks like this: the power supply without correction consumes current in short and high pulses, approximately coinciding with the peak of the mains voltage sine wave (green line is voltage, yellow line is current):

This waveform was taken at 200W on a Fortron/Source unit; as the load decreases, the current peaks become narrower and lower. For the unit from Delta Electronics, the picture looks a little different, but in principle nothing changes - all the same current surges at the maximum voltage, only slightly smoothed by the passive PFC choke, and zero current at a voltage less than two-thirds of the maximum:

This picture is explained by the features of the circuitry of pulsed power supplies: at the input of such a power supply there is a rectifier followed by a capacitor (or, to be precise, usually two capacitors), from which the supply voltage is already removed for the inverter of the pulsed DC-DC converter. When you turn on the power supply to the network with the first quarter-wave of the mains voltage, the capacitor is charged up to three hundred and a few volts. Then the mains voltage begins to drop rapidly (the second quarter-wave), while the capacitor discharges into the load much more slowly - as a result, at the moment the mains voltage begins to rise (the third quarter-wave), the voltage on the capacitor that has not had time to discharge will be about 250V, and while the voltage in the network is less - the charge current will be equal to zero (the rectifier diodes are blocked by the reverse voltage applied to them, equal to the difference between the voltages on the capacitor and in the network). In the last third of the quarter-wave (of course, I give all numerical estimates very approximately - in reality they depend on the magnitude of the load and the capacitance of the capacitor), the voltage in the network will exceed the voltage on the capacitor - and the charge current will flow. The charge will stop as soon as the voltage in the network again becomes less than on the capacitor - this will happen in the first half of the fourth quarter-wave.

For a block with active PFC, the picture changes completely. Here, the current is already proportional to the voltage, as in a conventional resistive load:

As a result, the power taken from the network is evenly distributed over the half-cycle of the mains voltage, and the current amplitude is much less than that of power supplies without power factor correction or with passive correction.

So, everything is clear with power supplies, now you can move from laboratory load to real computers.

Computers

Four computers of different power took part in this test, from the currently relatively slow Pentium III 800MHz to a dual-processor computer on AMD Athlon and a single-processor computer on Pentium 4 3.06GHz.

Computer configurations:

It can be said that an office computer is not a fast processor for today, a relatively simple video card, nothing more.

Processor Pentium III 800EB
Motherboard on Intel chipset i815EPT
256MB SDRAM
Winchester Quantum Fireball AS 30GB
Video card GeForce2 MX400, 64MB
Network card 3Com 3C905C-TX
CD-ROM LG CRD-8521B
A mid-range home computer is a good but relatively inexpensive processor and graphics card that can handle most modern games.

Processor AMD Athlon XP 2100+
Motherboard based on VIA KT400 chipset
256MB DDR SDRAM
Winchester IBM ICL35 80GB
Video card ATI Radeon 8500
Sound creative card Audigy
CD-RW Teac CD-W540E
DVD-ROM ASUS E616
Powerful workstation - two processors, RAID, lots of memory.

Two AMD Athlon 1200 processors based on the Thunderbird core
512MB DDR SDRAM
Four Maxtor D740X 20GB hard drives in a RAID array
Video card Matrox Millennium
Computer top level- the fastest processor, the fastest graphics card.

Processor Intel Pentium 4 3.06GHz
Motherboard based on Intel i850E chipset
Two 512MB RDRAM modules
Two hard drives western digital WD400JB in RAID1
Video card NVIDIA Quadro4 900XGL
DVD-RW Pioneer DVR-104

Connected to computers optical mouse MS IntelliMouse and PS/2 keyboard. The power consumption of the monitor (NEC LCD 1525V) was not taken into account - it was powered from a separate outlet.

Energy consumption was measured in three modes - when idle (Windows is loaded, nothing else happens), when the hard drive is defragmented, and when the computer is booted using ZD 3D Winbench 2000 and 3D Mark 2001SE (tests were chosen, of course, not to evaluate performance, but only to create load on the processor and video card). In each case, up to ten oscillograms were taken, but only the maximum measured values were included in the final results.

So, the results. The table below shows the power consumption of the “stuffing” of the computer itself - that is, the measured power consumption from the network has already been multiplied by the efficiency of the used power supply.

The power ratio for each individual computer is, in principle, quite predictable - for example, on systems with Athlon XP 2100+ and Pentium 4 3.06GHz, a powerful video card made its contribution in 3D tests. The relatively high consumption of systems for AMD processors when idle, this is due to the fact that these processors require a bus disconnect to switch to power-saving mode, which is not implemented on the vast majority of motherboards. Work station on two Athlon "s, thanks to four hard drives, it showed a good increase in power consumption during defragmentation, but on 3D tests the power increased by only 17W - firstly, the Matrox Millennium video card does not have any 3D accelerator, so its consumption changes slightly, firstly secondly, since processors do not go into low power mode without the system bus being disabled, a noticeable increase in load has little effect on power consumption.

The absolute power values are quite interesting. Maximum fixed power consumption - 154W for the most powerful computer on the P4 3.06GHz, with a gigabyte of memory and a Quadro4 900XGL video card. And even if we add to this power, say, a DVD drive and the active use of hard drives (although I personally can hardly imagine a situation where all computer components are used at full power at the same time), the total power consumption will obviously not exceed 200W. However, this is the average power consumption, and there is also an instantaneous one, which cannot be measured using the applied technique - it is due to consumption surges, for example, when moving the hard drive heads (the current consumed in this case is approximately 1-2A along the + 12V line). But even taking into account such surges (which, by the way, are partly extinguished by the output capacitors of the power supply), the instantaneous power will not exceed 250W.

However, very often there are cases when powerful computers either refuse to work at all on power supplies with a power of 250-300W, or work unstable (the most common sign of a lack of PSU power is reboots or freezes when starting 3D tests, games and similar programs ). The point here is that for many manufacturers of power supplies, the concept of power is becoming more and more conditional - if we have long ceased to be surprised by the so-called peak power (PMPO - Peak Maximum Power Output) of cheap computer speakers, reaching completely unrealistic values of hundreds of watts, it seems that soon you will have to get used to the same power designations on cheap power supplies. I'm not even talking about the real currents issued by the power supplies - but the power written on the label often does not agree with the load currents written right there.

For example, let's compare two blocks that were considered in the fifth series of testing ATX power supplies - Fortron / Source FSP300-60BTV and PowerMini PM-300W. Both units are listed as 300W, but the first one is medium. price category, and the second - to the bottom. If you look at the labels, it turns out that the FSP300 is capable of delivering up to 15A current via the +12V bus, while the PM-300 can only deliver up to 12A.

What does this lead to? In modern computers, a lot is powered by the +12V bus - here is a DC-DC converter for powering the processor (in systems based on Pentium 4; in systems based on AMD processors, +5V is usually used), and a video card with its own on-board stabilizer, and a solenoid head drive hard drive, and the DVD-ROM engine "a ... Obviously, a situation can easily arise when the instantaneous consumption on this bus will block the capabilities of the PM-300W block, but at the same time will be within acceptable limits for the FSP300-60BTV and even for many 250W blocks capable of supplying up to 13A on this bus for an unlimited time, and at the peak - up to 16A (for example, blocks from the same Fortron / Source company).If we add to this the small capacitance of the capacitors at the output of the PM-300W (and capacitors can significantly smooth out consumption surges short duration), the absence of any power reserve ... The result is obvious - at the first current surge in a cheap unit, either protection will work (and in many such PSUs it is not even set up on the application this power, and the power is 20-30W less), or the voltage will drop - for a short time, but by such an amount that the computer freezes or reboots.

Moreover, cases and power supplies from Microlab with the marking “M-ATX-350W” have recently appeared on sale. Of course, the buyer thinks that these blocks are designed for a power of 350W, however... + 5V - 20A. If you open the ATX/ATX12V Power Supply Design Guide and look at the tables with the recommended load capacity for power supplies of various capacities (Section 3.2.3.2), it turns out that such output currents can only be considered normal for a 200W ATX12V power supply. However, formally there is nothing to complain about - as I said, the output power is not indicated anywhere on the block, and the name of the model ... “at least call it a pot, just don’t put it in the stove,” as folk wisdom says.

However, there are also blocks that directly violate the requirements of the Design Guide. For example, Codegen 250X1. This unit is sold as designed for Pentium 4 processors, in other words, conforming to the ATX12V standard. Of course, there is also a 4-pin ATX12V connector. At the same time, the maximum allowable current on the +12V bus is 9A, while the Design Guide explicitly says that units with a current of less than 10A should not have this connector (Section 3.2.3.2), and, accordingly, such a unit cannot comply with ATX12V standard (Section 1.2.1).

Conclusion

Several interesting conclusions can be drawn from the studies carried out.

First, not everyone modern computer a power supply unit with a power of more than 300W is required, and often 250W is enough. The average consumption of even a very sophisticated computer is only about 150W, that is, a 300W power supply provides it with a good margin. Even on video cards based on the GeForce FX chip, the consumption of which can reach up to 70W (the used Quadro4 900XGL had about 20W), average power, consumed from the power supply, will not exceed 200W.

Secondly, really as a rule, there is no problem with a lack of power of a 300W power supply - in fact, many cheap power supplies are simply not able to deliver the power indicated on them, so the problem should rather be formulated as “lack of power of 150W, more than which some PSUs are not capable of delivering, despite labeled 300W”. When buying a power supply, I would advise you to pay attention not only to the total power, but also to individual currents on different buses - as you can see, units with the same declared power can differ significantly in declared currents, not to mention real currents. Besides a good criterion is the mass of the block - the heavier it is, the better, as a rule.

Thirdly, not all power factor correction schemes give a noticeable effect. Passive correction, which is very widely used in blocks of the middle price category, improves the power factor by only 0.05-0.1 and makes it less dependent on the load, while active correction schemes can bring the power factor up to 0.95-0.99. Accordingly, when buying a power supply, you should pay attention not only to the very fact of the presence of PFC, but also to its implementation - blocks with passive PFC are easy to distinguish by an additional choke of impressive size in them, which is usually fixed on the top cover of the PSU.

), as well as its derivatives. At the moment, it is customary to rank as supercomputers systems with a computing power of more than 10 Teraflops (10 * 10 12 or ten trillion flops; for comparison, the average modern desktop computer has a performance of about 0.1 Teraflops). One of the most powerful computer systems in the Linpack test - the Japanese K computer - has a performance in excess of 10.5 Petaflops.

Definition ambiguity

There are several difficulties in determining the computing power of a supercomputer. First, keep in mind that system performance can be highly dependent on the type of task being performed. In particular, the need for frequent data exchange between components negatively affects the computing power. computer system, as well as frequent memory access. In this regard, peak computing power is allocated - hypothetically, the maximum possible number of operations on floating point numbers per second that a given supercomputer is capable of producing.

performance measurement

Real computing power is estimated by passing special tests (benchmarks) - a set of programs specially designed for performing calculations and measuring their execution time. Usually, the speed of solving a large system of linear algebraic equations by the system is estimated, which is due, first of all, to the good scalability of this problem.

The most popular benchmark is the Linpack benchmark. In particular, HPL (an alternative implementation of Linpack) is used in compiling the list of TOP500 supercomputers in the world.

Other popular testing programs are NAMD (molecular dynamics problem solving), HPCC (HPC Challenge Benchmark), NAS Parallel Benchmarks.

The most powerful supercomputers

As of June 2011, the most powerful supercomputers are:

The highest place taken by Russia is 12th in November 2009, with the Lomonosov supercomputer. For November, 2011 after updating the Lomonosov supercomputer takes the 18th place.

Notes

Wikimedia Foundation. 2010 .

See what "computing power of a computer" is in other dictionaries:

Power: Power (in physics and engineering) is the ratio of work done in a given period of time to that period of time. The cardinality of a set (in mathematics) is the number of elements in a set. The computing power of a computer is the number of operations, ... ... Wikipedia

Screenshot from a test program demonstrating processor overclocking. Contents 1 ... Wikipedia

Scheme personal computer: 1. Monitor 2. Motherboard 3 ... Wikipedia

The style of this article is not encyclopedic or violates the norms of the Russian language. The article should be corrected according to the stylistic rules of Wikipedia ... Wikipedia

"PC" is redirected here. See also other meanings. This article is about all types of personal computers, for the most common platform, see: IBM PC compatible computer. The main components of a personal computer Personal ... ... Wikipedia

Overclocking, overclocking (from the English. Overclocking) increase the performance of computer components due to their operation in forced (abnormal) modes of operation. Contents 1 Criteria for normal computer operation ... Wikipedia

In order to measure the performance of a computer using tests, it is not necessary to download any third-party applications and utilities.

It is enough to use the resources already built into the operating system.

Although for more detailed information the user will have to find a suitable program.

Based on the test results, one can draw conclusions which of the parts of a PC or laptop need to be replaced before the rest - and sometimes just understand the need to buy a new computer.

The need for verification

Performing a computer speed test is available to any user. The check does not require any specialized knowledge or experience with certain versions of Windows OS. And the process itself is unlikely to take more than an hour.

Reasons why you should use the built-in utility or third-party application include:

unreasonable slowdown of the computer. Moreover, not necessarily the old one - a check is needed to identify problems with new PCs. So, for example, the minimum results and indicators of a good video card indicate incorrect installed drivers;
checking the device when choosing several similar configurations in a computer store. Usually they do this before buying laptops - running a test on 2-3 devices that are almost identical in parameters helps to find out which one suits the buyer better;

the need to compare the capabilities of the various components of a progressively upgraded computer. So, if the HDD has the lowest performance value, then it should be replaced first (for example, with an SSD).

According to the results of testing, which revealed the speed of the computer performing various tasks, you can detect driver problems and incompatibility of installed devices. And sometimes even poorly functioning and failed parts - for this, however, you will need more functional utilities than those that are built into Windows by default. Standard Tests show minimal information.

Checking by means of the system

You can check the performance of individual computer components using the built-in capabilities operating system Windows. Their principle of operation and information content are approximately the same for all versions of the platform from Microsoft. And the differences are only in the way of launching and reading information.

Windows Vista, 7 and 8

In the 7th and 8th versions of the platform, as well as Windows Vista, the performance counter of computer elements can be found in the list of basic information about the operating system. Just click to display them. right click click on the "My Computer" icon and select properties.

If testing has already been conducted, information about its results will be available immediately. If the test is running for the first time, it will have to be launched by going to the performance test menu.

The maximum score that Windows 7 and 8 allow you to get is 7.9. It’s worth thinking about the need to replace parts if at least one of the indicators is below 4. For a gamer, values above 6 are more suitable. Windows Vista has a better indicator of 5.9, and a “critical” one of about 3.

Important: To speed up performance calculations, you should turn off almost all programs during the test. When testing a laptop, it is advisable to turn it on to the network - the process noticeably consumes battery power.

Windows 8.1 and 10

For more modern operating systems, finding information about computer performance and starting its calculation is no longer so easy. To run a utility that evaluates system parameters, should take these steps:

1Go to command line operating system(cmd via menu "Run", called by pressing the keys simultaneously Win + R);

2Enable evaluation process, leading the team winsat formal –restart clean;

3Wait for completion of work;

4Go to folder Performance\WinSAT\DataStore located in the Windows system directory on system drive computer;

5Find and open in text editor file "Formal.Assessment(Recent).WinSAT.xml".

Among a lot of text, the user must find WinSPR block, where approximately the same data is located that is displayed on the screen and Windows 7 and 8 systems - only in a different form.

Yes, under the name SystemScore the general index calculated by the minimum value is hidden, and MemoryScore, CPUScore And GraphicsScore stand for memory, processor, and graphics card metrics, respectively. GamingScore And DiskScore- performance for the game and for reading / writing the hard disk.

The maximum value for Windows 10 and version 8.1 is 9.9. This means that the owner office computer you can still afford to have a system with numbers less than 6, but for the full operation of a PC and a laptop, it must reach at least 7. And for a gaming device - at least 8.

Universal way

There is a way that is the same for any operating system. It consists in launching the task manager after pressing the Ctrl + Alt + Delete keys. A similar effect can be achieved by right-clicking on the taskbar - there you can find an item that launches the same utility.

On the screen you can see several graphs - for the processor (for each thread separately) and random access memory. For more details, go to the "Resource Monitor" menu.

From this information, you can determine how heavily loaded individual PC components. First of all, this can be done by the percentage of loading, secondly - by the color of the line ( green means normal work component, yellow- moderate red- the need to replace the component).

Third Party Programs

By using third party applications It's even easier to test your computer's performance.

Some of them are paid or shareware (that is, requiring payment after the end of the trial period or to increase functionality).

However, these applications also conduct testing in more detail - and often give out a lot of other useful information for the user.

1.AIDA64

As part of AIDA64, you can find tests for memory, cache, HDDs, SSDs, and flash drives. And when testing the processor, 32 threads can be checked at once. Among all these pluses, there is a small drawback - you can use the program for free only during the "trial period" of 30 days. And then you have to either switch to another application, or pay 2265 rubles. for a license.

2. SiSoftware Sandra Lite

3.3DMark

4. PC Mark 10

The application allows not only to test the operation of computer elements, but also to save the results of tests for further use. The only drawback of the application is the relatively high cost. You will have to pay $30 for it.

5. CINEBENCH

The test images consist of 300,000 polygonal images that add up to more than 2,000 objects. And the results are given in the form PTS indicator - the higher it is, the more powerful computer . The program is distributed free of charge, which makes it easy to find and download it on the network.

6.ExperienceIndexOK

Information is displayed on the screen in points. Maximum amount– 9.9, as for latest versions Windows. It is for them that the work of ExperienceIndexOK is intended. It is much easier to use such a program than to enter commands and search the system directory for files with results.

7. Crystal DiskMark

To test a disk, select a disk and set the test parameters. That is, the number of runs and the size of the file that will be used for diagnostics. After a few minutes, the average read and write speed for the HDD will appear on the screen.

8 PC Benchmark

After receiving the test results, the program offers to optimize the system. And after improving the work in the browser, a page opens where you can compare the performance of your PC with other systems. On the same page, you can check if the computer can run some modern games.

9.Metro Experience Index

10. Pass Mark Performance Test

conclusions

Usage various ways Computer Performance Checker allows you to check how your system is performing. And, if necessary, compare the speed of individual elements with the performance of other models. For a preliminary assessment, such a test can also be carried out using the built-in utilities. Although it is much more convenient to download for this special applications- especially since among them you can find several quite functional and free.

Video: