Hello!. I want to ask you for advice.
So here's my problem. I just can’t choose a processor, but this is the most important component among or. After all, just by one processor you can determine whether your computer is modern and productive or whether it is an old one, suitable only for working in office applications.
When buying a computer, the first thing the seller always asks is: “What tasks do you need a computer for?”
Second: “How much do you expect?”
Third: "Which processor should you choose?"
Then, depending on the specified purposes for using the computer and the indicated amount, as well as the selected processor, the seller will select the motherboard and all other components.
I just can’t decide on the choice of processor? Why? I'll answer you. Despite the large volume random access memory(8GB) and a good video card, on the previous computer I bought, all the games that were current at that time ran without problems, but the FPS was always low and video processing was done in Adobe Premiere Pro took much longer than for a friend of mine who has a similar computer, but only with a processor from a different manufacturer.
In the end, I concluded that it was all because of the processor!
I’m ready to allocate the necessary amount to purchase a processor, but I don’t want to overpay. I have a great desire to choose exactly the processor that I need. I use the computer to the fullest, I can play games, and also digitize videos, burn discs, communicate on the Internet, and so on.
I hope to learn on your website not only how to choose a processor, but also how to choose a motherboard, RAM, video card, HDD, power supply, case and monitor!
In the meantime, your answers to the questions listed below would help me a lot!
- Which processor manufacturer came before Intel or AMD?
- Why are Intel processors always more expensive, is it just because of the Intel brand? Could it be that Intel processors are exactly the same in quality and performance as AMD and we are simply overpaying for the Intel name?
- When should you buy only an Intel processor? And when can you afford to save money and buy an AMD processor?
- If I ultimately choose Intel, is it worth spending money on a branded 4-core processor? Intel Core i7, maybe limited to Intel Core i5 or a completely 2-core Intel Core 2 Duo processor?
- And if I choose an AMD processor, then which model should I pay attention to: the very expensive AMD FX-9590 or just the highly productive 8-core AMD FX-8350 processor?
- Why do I have computer games oh, low FPS (the number of frames changed per unit of time), what does it even depend on?
- What better than AMD FX-8350 or Intel Core i7-3770K?
- How do processors with a "K" at the end differ from processors without this letter, such as the Intel Core i7-3770K and Intel Core i7-3770?
- Which processor would you choose yourself and please indicate approximate prices for current processor models?
How to choose a processor
Hello friends, Alexey is with you again! There are a lot of questions, but I can handle it, although the article will be long, but also interesting. After reading it, you will know everything about processors!
In fact, when assembling a computer, usually, first of all, the processor is selected and then everything else is made for it.
Choosing a processor is one of the easiest tasks when determining the future configuration of a computer. Here often decisive factor is the amount we are willing to spend on it, or high specifications, if the processor is planned to be used for professional or highly specialized activities.
This article can be used as a guide to choosing a processor for a new computer, or for upgrading an old one.
general information
I don't want to go into history and talk about how processors have evolved, suffice it to say that processors are the highest modern achievement. They are produced in only a few factories in the world that have truly space technology. Therefore, the processor is one of the most reliable system components today.
Historically, the entire market for central processing units (CPUs) personal computers divided between two large corporations, well-known: Intel and AMD.
Who is the leader, Intel or AMD?
Two-way licensing
In 1968, three outstanding physicists Gordon Moore, Andrew Grove, and Robert Noyce founded the world-famous INTegrated ELectronics Corporation, which we all know as INTEL.
It is INTEL that is a recognized pioneer in the field of technologies included today in modern lines of all processors. This is often the subject of debate among supporters of a particular company. They say Intel is better, but AMD is cheaper, with sometimes a slight difference in performance.
Perhaps not everyone knows, but Intel and AMD have had an official agreement on free two-way licensing since 1976. This means that each company can use any technology developed by a competitor without receiving any additional license. And AMD has always used this, which cannot be said about such a proud bird as Intel.
As a result, almost all technologies developed by Intel exist on AMD processors, otherwise they simply would not be able to support modern applications, whose developers focus primarily on architecture Intel processors.
Note: Many users will find this strange. Why on earth would that be Intel share development secrets with AMD. Friends, do not forget, both companies are located in the USA, and there is antitrust legislation there, in addition, both companies Intel and AMD are official suppliers of their products to the US Army.
What types of processors are there?
Appearance
Externally CPU looks like a monolithic metal case covering a board with a so-called crystal (a piece of silicon with microscopic electronic elements) and big amount contact legs (or pads) on the other side.
Intel processor (has modern pads)
AMD processor (with classic legs)
We will not delve into the jungle of processor microarchitecture, such as exclusive and inclusive cache, branch prediction unit, data prefetch unit, etc. I will tell you only about the most important characteristics of processors, which distinguish them and are of greatest importance to us.
How do Intel and AMD processors differ from each other, or how to choose a processor and not regret it later!
First of all, Intel and AMD processors differ in the length of the so-called computing pipeline, which determines the main differences in the areas of their use.
Note: A pipeline is a method of organizing calculations that is used in modern processors to increase their performance. http://ru.wikipedia.org
Intel processors have historically been aimed at the industrial sector, which is often dominated by stream processing operations, i.e. when data flows in a large continuous stream. Classic examples of streaming information processing include video encoding and archiving of large volumes of data. Therefore, Intel processors have a fairly long pipeline, which allows them to process more information in one pass and, accordingly, do it faster.
AMD processors have seriously made a name for themselves on the market When computer systems went to the masses and were initially positioned as multimedia (gaming) processors, which emphasizes the name of the company’s own technologies 3DNow!
AMD processors, compared to Intel processors, have a shorter computing pipeline, as a result of which these processors are slightly worse at processing streaming data, since less information is processed in one pass, but this does not prevent them from doing an excellent job, for example, with computer games, in which data cannot be predicted in advance, since it depends on user actions and is therefore transmitted in small portions, which are quickly processed on a short pipeline of the AMD processor.
This suggests a simple conclusion.
If you plan to constantly process video or create archives and information processing time is critical for you, then there is only one way out - an Intel processor. If you are a simple home user or you need a computer for the office, then you can significantly save your budget by purchasing an AMD processor, which will also cope with its tasks perfectly, but will cost $100 less...
Many admirers AMD processors may notice:“Well, it’s just like that, all AMD processors are only suitable for the office!”
Of course not, friends! If you take the most current modern 4- and 8-core processors from AMD, for example CPU AMD FX-8350 4.0 GHz /8 cores/ 8+8Mb/125W/5200 MHz Socket AM3 (price 6,500 rubles), then you can do absolutely everything, play all modern games, process videos and so on, but according to all kinds of tests, this processor will be inferior in performance by about 10-15% to a similar 4-core processor from Intel, for example this Intel Core i7-3770K 3.5 GHz (price 11,000 rubles).
I want to say that if you are a gamer, then Intel processors are what you need. In almost all modern games, computers with Intel processors will produce 30% more FPS (frames per second) compared to their AMD counterparts. If you are engaged in video processing, then again you will have to look towards Intel for the same reason.
I'll even say this the only advantage of AMD processors over Intel processors this is a lower cost. A modern processor from AMD will cost less than a processor from Intel by about $100. Agree, that kind of money is not lying around on the road either.
We must give AMD credit for its fighting spirit; having such a serious opponent as Intel, the company never gives up! Realizing that it is losing in technology, AMD is trying to win with its pricing policy.
The most modern processor from AMD - FX-9590
Isn't any special achievement this processor is the same FX-8350 processor, but only overclocked by the manufacturer itself to a frequency of 4.7 GHz and in turbo mode 5.0 GHz, which also has excessive power consumption and heat dissipation. Again, if you give the results all kinds of tests, then this processor has no advantage over the Intel Core i7-3770K 3.5 GHz and Intel Core i7-4770K 3.5 GHz, and the AMD FX-9590 (price 12,000 rubles) is a little more expensive than the Intel processors I mentioned. In addition, I forgot to tell you that with modern games, the AMD FX-9590 processor gets seriously hot, and this is not surprising with such an increase in supply voltage and frequency, and you will have to buy a serious cooling system, and that’s still money.
How do you still choose a processor? In my opinion, the most reasonable choice for a computer enthusiast who can play games, digitize videos, archive various data, communicate on the Internet, and so on, on this moment Intel Core i7-3770 3.4 GHz processor. The absence of the letter “K” at the end indicates that this processor has a locked multiplier, that is, you will not be able to overclock it, but I want to say that even without overclocking this processor works like an airplane, I don’t know where to overclock it, and you will save money 1,000 rubles. He's had enough already acceptable price 10,000 rubles. This processor is the “Editor's Choice” of many computer publications, and in general has long proven itself to be a good product.
Do you want an Intel processor, but Core i7 is a bit expensive for you?
20% percent, that is, the Intel Core i7-3770 processor is not at all inferior in power to its younger brother Intel Core i5-3570K 3.4 GHz (price 8,000 rubles). It turns out that this is a direct competitor to the AMD FX-8350 4.0 GHz processor we have already reviewed (price 6,500 rubles). The Intel Core i5-3570K processor is in no way inferior to it, but the price, as we see, is again slightly more expensive than that of an AMD processor.
If you are an enthusiast and a fan of overclocking the processor Having squeezed out of it exorbitant frequencies, pay attention to the Intel Core i7-3770K 3.5 GHz and Intel Core i7-4770K 3.5 GHz processors (price 12,000 rubles) with an unlocked multiplier. For example, the Intel Core i7-4770K processor can be overclocked to 4.5 GHz.
What else are good about Intel processors? They have a built-in graphics core, that is, a built-in video card. If you bought a computer with an Intel processor, then you may not have to buy an expensive video card for a while. Of course, you won’t be able to play the latest games with it, but you can play games that are two or three years old, but for office tasks such a video card will be a good choice.
If you want to find out the prices for modern processors, go to the end of the article, there is a price list for an average computer store. After familiarizing yourself with it, you will go to the computer store already prepared and will know the approximate layout.
How else do processors differ from each other?
Friends, what we have just discussed with you is a little superficial. After all, in addition to the manufacturer (Intel and AMD), processors differ from each other in the number of cores, frequency, cache, socket, presence or absence of a video core, energy consumption and heat generation, and much more. Let's look at this issue in more detail, I'm sure this secret knowledge will be useful to you.
Main characteristics of processors
All processors, regardless of manufacturer, differ in such basic indicators as the number of cores, core frequency, cache memory size, and support for different RAM frequencies. So, first things first.
Increasing the number of computing cores has the greatest impact on processor performance and, accordingly, on price too. Modern computer must have at least a 2-core processor, and preferably a 4-core processor. Options with 6, 8 or more cores can be considered as a purchase for the future.
Also, processor performance directly depends on the core frequency. Today, the normal frequency of a modern processor is considered to be between 3 and 4 GHz. The higher the core frequency, the higher the performance, but also the higher the power consumption, temperature, requirements for the motherboard, power supply, and the actual price.
Processor cache
Cache size also affects processor performance, but not to the same extent as multi-core or core frequency. Additionally, this impact will vary from application to application. In some programs the increase can be 15%, in some 5... But this has a significant effect on the price, because cache memory, being incredibly fast (an order of magnitude faster than RAM), is also very expensive...
There are 3 levels of processor cache.
L1 cache. Level 1 cache has the highest operating speed, but also the smallest size at 64 KB per core. It contains the basic instructions (algorithms) necessary for the operation of the processor and is usually not emphasized.
L2 cache. Level 3 cache is slower than level 2 and is not available on all processors. Processors positioned as powerful multimedia processors have about 3-6 MB of total level 3 cache (for all cores). Top expensive processors may have 8 MB or more of total Level 3 cache.
And finally, the memory controllers built into the processor determine how fast the RAM it can support (1333, 1600, 2000 MHz). In this regard, Intel processors often outperformed the clumsy AMD. But the increase in real applications as with cache memory, it may not always be noticeable. The amount of RAM has always played a big role here. If there is enough RAM, the computer works normally, if not, it slows down. That's all science) Information about what memory the processor supports can be found on the manufacturer's website. It is also necessary that the motherboard supports the same frequency.
Additional processor characteristics
Other, but also important differences between processors are process technology, power consumption, and operating temperature.
Characteristics such as power consumption and operating temperature greatly depend on the manufacturing process technology of the processor. As it improved, processors became faster, cooler, and at the same time more economical. This miracle of technological progress has no negative sides - the finer the technical process, the better. What does this even mean? In the process of improving production technologies, it is possible to make microscopic transistors that make up the computing cores, capacitors that make up the cache and the conductors between them of smaller and smaller sizes. As a result, it is possible to fit many more of these elements onto a piece of silicon of the same size, which allows for increased performance, while at the same time the conductors heat less and consume less energy, since they have also become thinner and their resistance has become lower. That's all physics, friends)
Today, the most modern processors are manufactured using a 22 nm (nanomicron) process technology, which is what we should strive to acquire.
CPU power consumption depends on the number of cores, their frequency and technological process. Here you need to take into account that a powerful processor cannot be installed on the cheapest motherboard and powered by the same power supply. Since they were not originally designed for such a load and can quickly fail. Energy consumption modern processors ranges from 65-125 watts, indicated on their packaging and on the manufacturer’s website. Similar data is indicated in the documentation and on motherboard websites. Read about how to choose the right power supply in the previous article.
Temperature is equated to the maximum power consumption of the processor and is characterized by such an indicator as the maximum temperature package “Thermal Design Power” or “TDP”. For modern processors it is also 65-125 Watts. Here you need to take into account that for a processor with a TDP of 65 Watt the simplest and cheapest cooler is enough, with a TDP of 100 Watt a more powerful cooler is needed, preferably with 2-4 heat pipes, with a TDP of 125 Watt a cooler with 4 heat pipes or more is needed. Cooler literally translated from English is a cooler, which is usually an aluminum, sometimes with a copper base, radiator with a fan attached to it to remove heat from the processor. The most advanced models have a design with so-called heat pipes, which on one side are in close contact with the processor, and on the other with the fins of the radiator, blown by a fan. Usually, the processor comes with a cooler specifically designed for it, but there are processors on sale without a cooler, so it is advisable to have this information.
The photo shows a cooler with heat pipes.
Please note that when installing or replacing a processor, you will need thermal paste, which is applied in a thin layer to the processor before installing the cooler. It is necessary for better heat transfer, otherwise the processor will overheat. If the processor is new and comes with a cooler, then thermal paste will already be applied to it.
Processor sockets
The processor socket, or as it is also called Socket, is the connection point between the processor and the motherboard. Processor sockets for each manufacturer and line of processors are different and they are marked either by the number of pins in the socket or by the marking of the line of processors.
Technological process at present time is running very quickly, processors change, processor sockets change. Well, what can I say... If you collect new computer, do not use motherboards and processors with outdated sockets, because if problems arise or you want to improve these components in a year or two, it will be difficult for you to find a replacement for them.
Intel Pentium- old line of 1 and 2 nuclear processors, with average performance, suitable for an office computer
Intel Core 2 Duo - an old line of 2 and 4 core processors, with high performance, suitable for replacement on older computers
Modern Intel processors
Intel Core i3 - the youngest, most affordable line of 2-core Intel processors
Intel Core i5 is an average, fairly productive line of processors, has both 4-core and some 2-core models
Intel Core i7 - senior, high-performance line of 4 and 6 core processors
More detailed labeling of these processors primarily depends on their frequency and cache size.
All Core series processors have a built-in video core, i.e. don't require additional installation video cards into the computer. This can be a beneficial solution if the PC will be used primarily for non-gaming purposes. But we must pay tribute to Intel engineers, who made it an order of magnitude more powerful than previous solutions that were integrated onto motherboards. Such a built-in video core can easily handle games from bygone years like Half Life 2 or Underground.
3. If the processor is not listed in the compatibility list, you can still try it by first updating the BIOS and agreeing with the seller about a return if it doesn’t work. Or give the system unit to the seller, let him try to install it himself. The only requirement here is that the processor fits into the permissible thermal package (TDP) of the motherboard, otherwise it may not withstand it (burn out).
I once observed how one of my clients, due to the installation, had too powerful processor on a weak motherboard it burned through!
4. If the processor is quite power hungry, you may need a more powerful and reliable power supply. Also, don’t forget about a sufficient cooler and thermal paste for cooling.
I wish you a good choice and Have a good mood! And if something doesn’t work out the first time, don’t forget that there are things in life more important than the processor, for example video card)
Approximate prices in central Russia
Our regular readers may remember a series of articles that was published in 2009 under the general title “The influence of various characteristics on the performance of processors of modern architectures.” In it, we examined a certain number of spherical processors in a vacuum in order to, based on an analysis of their performance, form a general impression of the speed of real processors and the factors influencing it. In the new year, after the release of the next version of the methodology, we decided to creatively rework the previously tested method with an emphasis on greater realism of the issues being studied, that is, modeling situations as real as possible. Like last time, we decided to start with AMD products, namely with its newest platform: Socket AM3. Fortunately, the manufacturer promises this platform a fairly long life, its popularity among the user environment is great, and the company chose a name that is more successful than its competitor - in terms of alphabetical sorting. :)
The current AMD line seems somewhat chaotic at first glance (we would say that all subsequent ones too...), but the manufacturer’s logic can be understood: of course, it’s much nicer to sell a defective processor than to throw it away. And since this company produces quite a lot of modifications with different volumes and types of caches and the number of cores, accordingly, there is a great temptation to come up with a name for a copy with a “defective” core or cache, disable the core or part of the cache, and disable the entire processor - still sell. :) Thanks to this wonderful, innovative policy of AMD, in the line of AM3 processors it produces there are as many as three varieties of dual-core ones - with different sizes of L2 cache, and even with the presence of L3; two modifications of tri-core ones - with and without L3; and again three modifications of quad-core ones - with and without L3, as well as with different L3 volumes. In addition, a single-core Sempron is also available for the AM3 platform. Having summarized the main technical characteristics of the CPU for the AM3 platform in one small table, we finally have a chance to understand that a certain kind of logic in model range AMD has:
| Sempron | Athlon II X2 | Phenom II X2 | Athlon II X3 | Phenom II X3 | Athlon II X4 | Phenom II X4 | Phenom II X6 | |
| cores | 1 | 2 | 2 | 3 | 3 | 4 | 4 | 6 |
| L2 cache, KB | 1024 | 2x512/1024 | 2×512 | 3×512 | 3×512 | 4×512 | 4×512 | 6×512 |
| L3 cache, KB | − | − | 6144 | − | 6144 | − | 4096/6144 | 6144 |
So, we are seeing a fairly logical “journey” from 1 core to 6, accompanied by variations on the size of the L2 cache, as well as the presence or absence of L3 and its size. At the same time, AMD “plays” with L2 capacity on relatively weak processors (dual-core), and then the introduction of L3 is used as a universal “accelerator of everything.” You can also note two equally strange-looking processors: the Phenom II X2, which with only 2 cores has a gigantic L3 cache, and, conversely, the Athlon II X4 - which, with 4 cores, lacks it completely. In theory, the first should be an ideal option for old software without multi-threaded optimization (although then it doesn’t really need a second core...), and the second should be a processor for optimists who hope that a 4-core CPU will defeat all processors with a smaller the number of cores, regardless of the size of the cache. So it will be or not - let's look at the results...
Accordingly, the most interesting comparisons emerge from the point of view of performance analysis:
- Increasing the number of cores with the same cache size:
- from 1 core to 2;
- from 2 cores to 3;
- from 3 cores to 4;
- from 4 cores to 6.
- Increasing the cache size with the same number of cores:
- on 2-core processors (different L2 sizes, adding L3);
- on 3-core processors (adding L3);
- on 4-core processors (adding L3, different L3 sizes).
- Variations on the theme “fewer cores, but more cache*”:
- 1-core processor versus 2-core;
- 2-core processor versus 3-core.
* - implies: for a single core.
As you can see, the soil for research is an unplowed field. True, in order for us to be able to fix our attention precisely on the influence of the above factors, removing all the interfering ones, we still needed to make one nod towards “synthetics” - regardless of whether such a CPU model exists in reality, all test participants worked at one core frequency: 2.6 GHz. However, everything is not so bad: Athlon II X3/X4, Phenom II X3/X4 with such a frequency really exist, only 2600 MHz Sempron, Athlon/Phenom II X2 and Phenom II X6 do not exist.
Testing
As mentioned above, testing was carried out in accordance with the latest 2010 methodology, with some minor modifications:
- Since the task before us was quite large-scale and interesting, and all test participants behaved very decently, and practically did not demonstrate any oddities that were inexplicable from a logical point of view, we made a voluntaristic decision to declare all optional tests permanent - thus, they are present in the main section. and participate on a general basis in GPA.
- Since a number of the considered processors are, so to speak, “virtual”, and are not actually produced, for of this cycle, for ease of comparison, we selected our own reference (100-point) processor from among those that took part in this particular series of tests: AMD Phenom II X4 810.
Also, some may find the first topic that we decided to explore unexpected: it is obvious that it is by no means in the first place on the list of questions, no matter how you look at it. Here you just have to forgive us for some chaoticness in the sequence of release of the series: it is caused by a simple “working moment” - the series will be released in the sequence in which the results considered in them become available. Unfortunately, the extensiveness of our testing methodology causes one inevitable drawback: tests take a very long time. Accordingly, if we decided to sacrifice efficiency for the sake of beauty, the first series (logically, we should start with comparisons with the participation of Sempron), you would have to wait about another month, while this one is ready now. We decided, on the contrary, to sacrifice beauty for the sake of efficiency, and we hope you will understand us. In addition, the format of the current testing: “one article - one answer to a specific question” is quite conducive to such an approach: after all, there are no “important” and “unimportant” questions, each of them is interesting in its own way, and each will certainly find its reader .
So let's get started. In this series we will look, as promised, one simple and specific question: Does a 3-core processor, in which each core has 512 kilobytes of L2 cache, have an advantage over a dual-core CPU, in which each core has 2 times more L2 cache - 1024 kilobytes? The first has the advantage of an additional core. On the other hand, each core of the second can work with double the amount of cached data. The situation, by the way, is not at all as obvious as it might seem at first glance...
3D visualization
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 79 | 92 | ||
| 94 | 91 | ||
| 94 | 90 | ||
| 98 | 95 | ||
| 95 | 90 | ||
| 98 | 94 | −4% |
|
| Group Score | 92 | 91 |
Yes, yes, to the question of the non-obviousness of the situation. Amazingly, when visualizing a three-dimensional image, only one package out of six was able to get some benefit from the additional core, but the other 5 reacted very critically to the reduction in L2 volume. Of course, it’s clear what this is connected with: most likely, they simply could not use the third core, and it was idle. Well, let's praise the 3ds max developers for good optimization, but at the same time we state: they are still in a clear minority.
Rendering 3D scenes
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 63 | 83 | ||
| 51 | 74 | ||
| 48 | 71 | ||
| Group Score | 54 | 76 |
In this group, the performance gain from adding another core is close to ideal, but regarding rendering this fact does not cause any surprise: 512 kilobytes of L2 cache is quite enough for the cores, because the scene is divided into fairly small pieces that can be calculated in parallel.
Scientific and engineering calculations
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 89 | 95 | ||
| 96 | 93 | ||
| 94 | 91 | ||
| 92 | 87 | ||
| 98 | 94 | ||
| 65 | 73 | ||
| 74 | 84 | ||
| Group Score | 87 | 88 |
The situation is more complicated: engineering CADs, apparently, operate with fairly large volumes of information when calculating, but do not know how to use the third core (to be fair, they often ignore the second one too...). The multi-threaded optimized Maya, Mathematica (remember that since 2010 we have been using a multi-threaded optimized version of the MMA test for this package) and MATLAB performed well, due to which the overall score for the group brought the 3-core CPU to the lead.
Raster graphics
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 101 | 97 | ||
| 96 | 97 | ||
Corel PhotoImpact | 99 | 98 | |
| 73 | 86 | ||
| Group Score | 92 | 95 |
The difference of ±1 percent is well within the measurement error, so all we can do is highlight the cache-loving ACDSee and the well-threaded optimized Photoshop. And again, due to a more tangible advantage in a well-optimized application, the 3-core is in the lead in the overall score of the group.
Data compression
Our compilation test (at least it should in theory...) currently supports up to 16 threads, so the advantage of a processor with more cores is not surprising.
Java
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| Group Score | 55 | 74 |
|---|
...Well, we know well the excellent parallelizability of the SPECjvm test from the 2009 method.
Browsers
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 90 | 94 | ||
| 93 | 97 | ||
| 97 | 97 | ||
| Group Score | 93 | 96 |
A completely new, unknown group of tests, for which there are no statistics yet, but a rather banal result: two benchmarks gave a slight advantage to the third core, and the third did not notice any difference at all.
Audio encoding
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 50 | 67 | ||
| 50 | 66 | ||
Monkey's Audio | 50 | 67 | |
| 50 | 67 | ||
| 51 | 67 | ||
| 50 | 67 | ||
| Group Score | 50 | 67 |
Audio encoding speed tests since 2009 have received excellent multi-threaded optimization through the use of the dbPoweramp package, which can run as many encoding processes as it detects in the processor system. In this situation, the victory of the 3-core player was a foregone conclusion.
Video encoding
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 77 | 93 | ||
Mainconcept (VC-1) | 64 | 81 | |
| 49 | 72 | ||
| 55 | 76 | ||
| 50 | 65 | ||
| 72 | 85 | ||
| Group Score | 61 | 79 |
Video encoding packages also demonstrate very decent multi-processor optimization, including the previously unused Adobe Premiere and Sony Vegas. Moreover, note: the two above-mentioned packages have one of the best in the group.
Playing video
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 50 | 70 | ||
| 104 | 49 | ||
| 53 | 75 | ||
| 48 | 72 | ||
| Group Score | 64 | 67 |
The new group of tests presented one of the few surprises, reacting sharply negatively to the 3-core. Looking ahead, we note: it seems that we are talking specifically about the reaction to 3 cores, and not to a decrease in the L2 volume, since the 4-core does not show such a large drop in performance. Perhaps there is a phenomenon of categorical “indigestion” by specific software of a number of cores other than a power of two; we have already encountered this before.
Virtual machine
The vast majority of games used the third core quite successfully, only Borderlands, S.T.A.L.K.E.R., Crysis and World in Conflict are not particularly optimistic (less than 10% increase). It’s not that the trend is so clear (UT3, for example, contradicts it), but you can notice that 3 games out of the four listed are not very new.
Total score
| 2 cores + 2×1024 L2 | 3 cores + 3×512 L2 | %% | |
| 71 | 80 |
The overall score is quite in the spirit of the times: even with a trimmed cache, multi-core is still in favor. However, not without some juicy details: 16 out of 57 tests preferred a processor with fewer cores, but a larger L2 volume on the core. There is a temptation to declare this fact the machinations of retrogrades and the laziness of programmers who are not good enough at using the resources of modern processors... and, probably, this is so. Still, for normal multi-core support you need to do certain work(sometimes quite large), and a large L2 sometimes causes an increase in performance “by itself,” without additional programmer effort. In this case, we should end on an optimistic note: judging by the overall score, there are fewer and fewer lazy people among software developers. As for practical recommendations, they are obvious: in general, in the case of the Athlon II, 3 cores are still definitely better than 2.
The search module is not installed.
Single core or dual core?
Victor Kuts
The most significant recent event in the field of microprocessors has been the widespread availability of CPUs equipped with two computing cores. The transition to a dual-core architecture is due to the fact that traditional methods for increasing processor performance have completely exhausted themselves - the process of increasing their clock frequencies has recently stalled.
For example, in the last year before the advent of dual-core processors, Intel was able to increase the frequencies of its CPUs by 400 MHz, and AMD even less - by only 200 MHz. Other methods of improving performance, such as increasing bus speed and cache size, have also lost their former effectiveness. Thus, the introduction of dual-core processors, which have two processor cores in one chip and share the load, has now turned out to be the most logical step on the complex and thorny path of increasing the performance of modern computers.
What is a dual-core processor? In principle, a dual-core processor is an SMP system (Symmetric MultiProcessing; a term denoting a system with several equal processors) and is essentially no different from an ordinary dual-processor system consisting of two independent processors. This way we get all the benefits of dual-processor systems without the need for complex and very expensive dual-processor motherboards.
Before this, Intel had already made an attempt to parallelize the instructions being executed - we are talking about HyperThreading technology, which ensures the division of the resources of one “physical” processor (cache, pipeline, execution units) between two “virtual” processors. The performance gain (in individual applications optimized for HyperThreading) was approximately 10-20%. Whereas a full-fledged dual-core processor, which includes two “honest” physical cores, provides an increase in system performance by 80-90% and even more (naturally, with full use of the capabilities of both of its cores).
The main initiator in the promotion of dual-core processors was AMD, which at the beginning of 2005 released the first dual-core Opteron server processor. As for desktop processors, Intel took the initiative here, announcing the Intel Pentium D and Intel Extreme Edition processors around the same time. True, the announcement of a similar line of Athlon64 X2 processors produced by AMD was only a few days late.
Dual-core Intel processors
The first dual-core Intel Pentium D 8xx processors were based on the Smithfield core, which is nothing more than two Prescott cores combined on one semiconductor chip. An arbitrator is also located there, who monitors the status system bus and helps share access to it between cores, each of which has its own 1 MB L2 cache. The size of such a crystal, made using a 90-nm process technology, reached 206 square meters. mm, and the number of transistors is approaching 230 million.
For advanced users and enthusiasts, Intel offers Pentium Extreme Edition processors, which differ from the Pentium D by supporting HyperThreading technology (and an unlocked multiplier), thanks to which they are defined operating system as four logical processors. All other functions and technologies of both processors are completely the same. Among them are support for the 64-bit EM64T instruction set (x86-64), energy saving technologies EIST (Enhanced Intel SpeedStep), C1E (Enhanced Halt State) and TM2 (Thermal Monitor 2), as well as the NX-bit information protection function. Thus, the considerable price difference between the Pentium D and Pentium EE processors is largely artificial.
As for compatibility, processors based on the Smithfield core can potentially be installed in any LGA775 motherboard, as long as it meets Intel's requirements for the board's power module.
But the first pancake, as usual, was a disaster - in many applications (most of which are not optimized for multi-threading), dual-core Pentium D processors not only did not outperform single-core Prescott processors running at the same clock frequency, but sometimes even lost to them. Obviously, the problem lies in the interaction of the cores via the Quad Pumped Bus processor bus (when developing the Prescott core, no provision was made for scaling its performance by increasing the number of cores).
To eliminate the shortcomings of the first generation of dual-core Intel processors, processors based on the 65-nm Presler core (two separate Cedar Mill cores located on the same substrate), which appeared at the very beginning of this year, were called upon. A more “fine” technical process made it possible to reduce the area of the cores and their power consumption, as well as increase clock frequencies. Dual-core processors based on the Presler core were called Pentium D with indexes 9xx. If we compare the Pentium D 800 and 900 series processors, in addition to a noticeable reduction in power consumption, the new processors have doubled the second level cache (2 MB per core instead of 1 MB) and support for the promising Vanderpool virtualization technology (Intel Virtualization Technology). In addition, it was released Pentium processor Extreme Edition 955 with HyperThreading enabled and running at a 1066 MHz FSB.
Officially, processors based on the Presler core with a bus frequency of 1066 MHz are compatible only with motherboards based on the i965 and i975X series chipsets, while the 800 MHz Pentium D in most cases will work on all motherboards that support this bus. But, again, the question arises about the power supply of these processors: the thermal package of the Pentium EE and Pentium D, with the exception of the younger model, is 130 W, which is almost a third more than that of the Pentium 4. According to Intel itself, stable operation of a dual-core system is only possible when using power supplies with a power of at least 400 W.
The most efficient modern desktop dual-core processors from Intel are, without a doubt, Intel Core 2 Duo and Core 2 eXtreme (Conroe core). Their architecture develops the basic principles of the P6 family architecture, however, the number of fundamental innovations is so large that it is time to talk about the new, 8th generation of processor architecture (P8) from Intel. Despite the lower clock frequency, they noticeably outperform the P7 family of processors (NetBurst) in performance in the vast majority of applications - primarily due to an increase in the number of operations performed in each clock cycle, as well as by reducing losses due to the large length of the P7 pipeline.
Desktop processors of the Core 2 Duo line are available in several versions:
- E4xxx series - FSB 800 MHz, 2 MB L2 cache common to both cores;
- E6xxx series - FSB 1066 MHz, cache size 2 or 4 MB;
- X6xxx series (eXtreme Edition) - FSB 1066 MHz, cache size 4 MB.
The letter code "E" indicates the power consumption range from 55 to 75 watts, "X" - above 75 watts. Core 2 eXtreme differs from Core 2 Duo only in its increased clock frequency.
All Conroe processors use a well-developed Quad Pumped Bus and LGA775 socket. Which, however, does not mean compatibility with older motherboards. In addition to support clock frequency 1067 MHz, motherboards for new processors must contain a new voltage regulation module (VRM 11). These requirements are mainly met by updated versions of motherboards based on Intel 975 and 965 series chipsets, as well as NVIDIA nForce 5xx Intel Edition and ATI Xpress 3200 Intel Edition.
In the next two years, Intel processors of all classes (mobile, desktop and server) will be based on the Intel Core architecture, and the main development will be in the direction of increasing the number of cores on a chip and improving their external interfaces. In particular, for the desktop market, this processor will be Kentsfield - Intel's first quad-core processor for the high-performance desktop PC segment.
Dual-core AMD processors
The AMD Athlon 64 X2 line of dual-core processors uses two cores (Toledo and Manchester) inside a single die, manufactured using a 90 nm process technology using SOI technology. Each of the Athlon 64 X2 cores has its own set of actuators and a dedicated second-level cache; they share a memory controller and a HyperTransport bus controller. The differences between the cores are in the size of the second level cache: Toledo has an L2 cache of 1 MB per core, while Manchester has half that figure (512 KB each). All processors have a 128 KB L1 cache, and their maximum heat dissipation does not exceed 110 W. The Toledo core consists of approximately 233.2 million transistors and has an area of approximately 199 square meters. mm. The Manchester core area is noticeably smaller - 147 square meters. mm., the number of transistors is 157 million.
Dual-core Athlon64 X2 processors inherited from Athlon64 support for Cool`n`Quiet energy saving technology, a set of 64-bit AMD64 extensions, SSE - SSE3, and NX-bit information protection function.
Unlike dual-core Intel processors that work only with DDR2 memory, the Athlon64 X2 is capable of working with both DDR400 memory (Socket 939), which provides a maximum bandwidth of 6.4 GB/s, and DDR2-800 (Socket AM2), whose peak throughput is 12.8 GB/s.
There are enough modern ones for everyone motherboards Athlon64 X2 processors work without any problems - unlike the Intel Pentium D, they do not impose any specific requirements on the design of the motherboard power module.
Until very recently, the AMD Athlon64 X2 was considered the most productive among desktop processors, but with the release of Intel Core 2 Duo the situation has changed radically - the latter have become the undisputed leaders, especially in gaming and multimedia applications. In addition, new Intel processors have lower power consumption and much more efficient power management mechanisms.
AMD was not satisfied with this state of affairs, and in response, it announced the release in mid-2007 of a new 4-core processor with an improved microarchitecture, known as the K8L. All its cores will have separate L2 caches of 512 KB each and one common level 3 cache of 2 MB in size (in subsequent versions of the processor the L3 cache can be increased). More detailed perspective AMD architecture K8L will be reviewed in one of the upcoming issues of our magazine.
One core or two?
Even a cursory glance at the current state of the desktop processor market indicates that the era of single-core processors is gradually becoming a thing of the past - both of the world's leading manufacturers have switched to producing mainly multi-core processors. However, software, as has happened more than once before, still lags behind the level of hardware development. Indeed, in order to fully utilize the capabilities of several processor cores, the software must be able to “split” into several parallel threads processed simultaneously. Only with this approach does it become possible to distribute the load across all available computing cores, reducing computation time more than could be done by increasing the clock frequency. Whereas the vast majority of modern programs are not able to use all the capabilities provided by dual-core or, especially, multi-core processors.
What types of user applications can be most effectively parallelized, that is, without special processing of the program code, they allow you to select several tasks (program threads) that can be executed in parallel and, thus, load several processor cores with work at once? After all, only such applications provide any noticeable increase in performance from the introduction of multi-core processors.
The greatest benefits from multiprocessing come from applications that initially allow natural parallelization of calculations with data sharing, for example, realistic computer rendering packages - 3DMax and the like. You can also expect good performance gains from multiprocessing in applications that encode multimedia files (audio and video) from one format to another. In addition, 2D image editing tasks lend themselves well to parallelization in graphic editors like the popular Photoshop.
It is not without reason that applications of all the categories listed above are widely used in tests when they want to show the advantages of Hyper-Threading virtual multiprocessing. And there’s nothing to say about real multiprocessing.
But in modern 3D gaming applications one should not expect any significant speed increase from multiple processors. Why? Because a typical computer game cannot be easily parallelized into two or more processes. Therefore, the second logical processor, at best, will only perform auxiliary tasks, which will provide virtually no performance gain. And developing a multi-threaded version of a game from the very beginning is quite complex and requires considerable labor - sometimes much more than creating a single-threaded version. These labor costs, by the way, may not yet pay off from an economic point of view. After all, computer game manufacturers traditionally focus on the most widespread part of users and begin to use new capabilities of computer hardware only if it is widespread. This is clearly seen in the use of video card capabilities by game developers. For example, after new video chips with support for shader technologies appeared, game developers ignored them for a long time, focusing on the capabilities of stripped-down mass solutions. So even advanced players who bought the most “sophisticated” video cards of those years never received normal games that used all their capabilities. A roughly similar situation with dual-core processors is observed today. Today there are not many games that even really use HyperThreading technology, despite the fact that mass processors with its support have been in full production for several years now.
In office applications the situation is not so clear. First of all, programs of this class rarely work alone - a much more common situation is when several office applications running in parallel are running on the computer. For example, the user is working with text editor, and at the same time the website is loaded into the browser, and scanning for viruses is carried out in the background. Obviously, having multiple applications running allows you to easily use multiple processors and get a performance boost. Moreover, everything Windows versions XP, including Home Edition (which was originally denied support for multi-core processors), is already able to take advantage of dual-core processors by distributing program threads between them. Thus ensuring high efficiency in the execution of numerous background programs.
Thus, we can expect some effect even from unoptimized office applications if they are run in parallel, but it is difficult to understand whether such a performance increase is worth the significant increase in the cost of a dual-core processor. In addition, a certain disadvantage of dual-core processors (especially the Intel Pentium D processors) is that applications whose performance is limited not by the processing power of the processor itself, but by the speed of memory access, may not benefit as much from having multiple cores.
Conclusion
There is no doubt that multi-core processors are definitely the future, but today, when most of the existing software is not optimized for new processors, their advantages are not as obvious as manufacturers try to show in their advertising materials. Yes, a little later, when there is a sharp increase in the number of applications that support multi-core processors(primarily this applies to 3D games, in which new generation CPUs will help to significantly relieve the load on the graphics system), purchasing them would be advisable, but now... It has long been known that buying processors “for growth” is far from the most effective investment.
On the other hand, progress is rapid, and for a normal person changing a computer every year is, perhaps, overkill. Thus, all owners have enough modern systems based on single-core processors, there is no need to worry too much in the near future - your systems will still be “at par” for some time, while for those who are planning to purchase a new computer, we would still recommend turning their attention to the relatively inexpensive junior models of dual-core processors.
|