home > Internet > Comparison of intel and amd processors. What is the best processor for a gaming computer

Comparison of intel and amd processors. What is the best processor for a gaming computer

The processor is an essential part of a personal computer. This is a small element of the system that does most of the main calculations of the system. In fact, this is the basis of the computer, thanks to which it functions. And, of course, that system performance is highly dependent on processor performance.

Due to the possibility of simple replacement of components, the processor can be easily replaced. However, often the problem is not the difficulties of replacement, but the choice of a good processor, which will significantly speed up the system and increase the overall performance of the computer. And in order to help you in choosing this element, this rating of the best processors of 2014 will help.

Where is the best place to buy processors

Regardless of the type of components, they must be purchased only in stores. There are several reasons for this:

the store will always answer your questions if you do not understand something;
often stores sell original components;
they provide a guarantee for the goods;
It is often possible to return the goods.

It is worth noting that online stores have the same advantages as real retail chains. However, the choice there is wider, which makes it possible to purchase a more suitable option.

It is recommended not to purchase components from private individuals at all. The same processor can be easily counterfeited so that it looks indistinguishable from the leading models in the market. Also, when buying from a private person, you should not even hope for a guarantee, because primitive market relations are used in this case.

Intel processor rating

Today, the undisputed leader in the processor market is Intel. It produces models with a unique architecture, developing them and achieving the maximum performance of this element of the system. Today, their main line of processors is i7, but even the previous generation, i5, is still actively used and is one of the best views processor.

Intel products have gained immense popularity due to their main advantages:

high performance;
reliability;
durability;
wide functionality;
and interaction with many other component manufacturers.

This makes them the best in their field, although they still have worthy competitors.

One of the best i7 based processors available. Good frequency, 4 cores, integrated graphics that have good performance, as well as an 8 megabyte level 3 cache. Uses socket 1150, which is quite beneficial for those who want to upgrade their computer without purchasing a new motherboard.

Generation: fourth;
Connector type: Socket 1150;
Clock frequency: 4000 MHz;
Number of cores: 4;
Level 3 cache: 8 MB
Advantages: performance, low temperature under load, common socket, overclocking capability;
Cons: The same old socket will force you to buy a new processor when you get a new motherboard.

The most powerful i7-based processor to date. Huge performance, 8 cores, 20 megabytes level 3 cache, with a frequency of 3 GHz. Unfortunately, the price is extremely high, and it also uses a new socket, which makes it necessary to purchase with the motherboard.

Generation: fourth;
Connector type: Socket 2011-3;
Internal clock frequency: 3000 MHz;
Number of cores: 8;
Integrated Graphics: No;
Level 3 cache: 20 MB;
Advantages: performance, overclockability, low operating temperature;
Cons: price.

Another i7-based processor that also uses the new socket. However, the processor belongs to the third generation, so this performance is relatively high. Six cores, 3.4GHz, 12MB L3 cache. The nice thing is that the performance is not too inferior to the previous model, and the price is half as much.

Generation: third;
Connector Type: Socket 2011;
Number of cores: 6;
Integrated Graphics: No;
Level 3 cache: 12 MB;
Advantages: price, performance, overclocking capability;
Cons: None noted.

The i5 based processor however is still a great option. The price here is much lower than previous models, and the performance differs slightly. There is an integrated graphics. The processor is great for upgrading your computer, but when building, it is recommended to think about more promising options.

Generation: fourth;
Connector type: Socket 1150;
Internal clock frequency: 3400 MHz;
Number of cores: 4;
Integrated Graphics: Intel HD Graphics 4600;
Level 3 cache: 6 MB total cache;
Pros: price, performance, old socket;
Disadvantages: poor overclocking capability, operating temperature slightly above average.

AMD processor rating

AMD is Intel's main competitor. Unlike Intel, they are focused on a different segment of society, creating low-cost and high-performance processors. It is due to the fact that their pricing policy is completely different that this company has become a leader in this market.

Unfortunately, the quality of this product often matches the price. Such processors are not designed for overloads or overclocking, however, their basic performance allows you to carry out various heavy operations (programming, graphics processing, games) without much difficulty. Therefore, for home computer AMD justifies itself quite well, which allows you to save a lot of money on collecting system block.

Not all models AMD processors can be called successful. However, there are those among them who are ready to compete with the best developments of Intel, if not surpass them. This rating of AMD processors will determine best models and on their basis to assemble your system unit.

Today, this AMD model is one of the best processors of the company. High frequency, 8 cores, good cache size and other features make this copy an excellent choice for a build. good computer with high performance.

Connector Type: Socket AM3+;
Internal clock frequency: 4000 MHz;
Number of cores: 8;
Integrated Graphics: No;
Advantages: overclocking is possible, high performance, price;
Flaws: given processor gets quite hot, which the stock cooler can't handle.

This processor stands out not only with its excellent performance, but also with the presence of high-quality integrated graphics from Radeon. Its performance allows you to watch movies in high resolution, as well as perform other graphical processes without a discrete graphics card.

Connector Type: Socket FM2+;
Internal clock frequency: 3700 MHz;
Number of cores: 4;
Integrated Graphics: AMD Radeon R7 Series;
Level 3 cache: No;
Pros: integrated graphics, performance;
Disadvantages: without the necessary cooling, the processor heats up quite high, which can make it noisy.

Despite the rather low price, this processor has a very high performance and potential. 3.9 GHz, 6 cores and 8 megabytes of L3 cache allow it to carry huge loads.

Connector Type: Socket AM3+;
Internal clock frequency: 3900 MHz;
Number of cores: 6;
Integrated Graphics: No;
Level 3 cache: 8 MB;
Pros: performance, overclockability, price;
Disadvantages: the cooler sent in the kit is often useless.

The best processor of 2015 - results

Today, the choice of processors is quite wide. Almost any kind of this component is present, so it is rather difficult to determine the best processor of 2015. Each model has its own advantages and disadvantages, but there are clear market favorites. And today it Intel Core i7-4790K and AMD FX-8350. Both models have high performance, but the performance of the simple i7-4790K is slightly ahead of its opponent. However, those who want to save as much as possible and get good performance can also use AMD. After all, the price difference is almost doubled.

The best processor of 2015 - detailed description

The i7-4790k processor is an extremely profitable option for a home computer. This is an excellent processor that will make you forget about the need to upgrade your computer for more than one year. high work efficiency, good price, as well as good integrated graphics make this model the best.

Generation: fourth;
Connector type: Socket 1150;
Clock frequency: 4000 MHz;
Number of cores: 4;
Integrated Graphics: Intel HD Graphics 4600;
Level 3 cache: 8 MB.

Where to buy a processor at competitive prices

If you are interested in purchasing a quality processor, then this store will help you with this. There is a wide selection of processors and components for the computer, so you will be able to choose the best option for myself. The product here is exclusively original, so you should not worry about its quality. Be sure that the processors presented here will allow your computer to work in a way that was previously inaccessible to it.

Highlights of 2015

Despite the fact that we started this material with complaints about the fact that there are not so many significant events in the PC processor market, this does not mean at all that the industry is stagnant. In 2015, Intel released two generations of 14nm desktop processors, one of which, however, received an unprecedentedly short life cycle. At that time, AMD was slowly developing the Kaveri line of hybrid processors and at the same time also introduced another, this time the final, version of the Bulldozer architecture, which, however, found its place in processors exclusively for the mobile market. Let's take a closer look at how this all happened.

So, at the beginning of 2015, Intel introduced the Broadwell generation processors, focused on thin and light laptops. These processors were the next tick in the tick-tock development cycle and, in fact, their main advantage was the transition to a modern 14nm process and numerous optimizations aimed at improving efficiency. At the same time, Intel wanted to conduct some experiment with the Broadwell processor design: it was initially focused on mobile and ultra-mobile solutions, and full-fledged desktop processors based on it were not planned at all. However, under pressure from customers, these plans were subsequently amended, and based on Broadwell, CPU models were nevertheless created for desktop systems. However, this led to a delay in the process of bringing them to the market, which was also seriously affected by production problems with 14-nm technology, which for a long time failed to obtain an acceptable yield of suitable crystals. As a result, desktop Broadwells were released only in June, that is, almost a year after the announcement of the first mobile processors based on this microarchitecture, which were included in the Core M family.

As a result, Broadwell entered the desktop processor market at a time when the next generation of the Skylake CPU was only a couple of months away. Intel has attempted to separate these novelties by positioning, setting Broadwell to more stringent thermal packages and providing them with some additional features: high-performance Iris Pro graphics and an additional eDRAM buffer that acts as an L4 cache. However, Broadwell's desktop measures could not be made popular. On the contrary, because of this, they have acquired the clear features of niche and highly specialized products that are in demand only in very rare cases. As Intel representatives themselves later stated, the whole saga with the release of Broadwell for desktop PCs is a big miscalculation. The resulting products not only came out clearly at the wrong time, but they also turned out to be too expensive and rather unbalanced in terms of characteristics, as a result of which their sales were, in fact, disrupted.

However, one cannot deny the fact that the Core i7-5775C and i5-5675C were able to demonstrate that AMD can no longer be considered the developer of the highest performing integrated graphics cores. Both Broadwell desktop processors released by Intel were able to offer noticeably better graphics performance than the competitor's older APUs, the AMD A10-7850K and A10-7870K. Thus, during their lives, the Core i7-5775C and i5-5675C still managed to fully play at least one important role - representative.

But already in August the place of the most modern Intel processors occupied models of the Skylake generation. And it was already a completely complete update. Indeed, along with the use of 14-nm technology, Intel also implemented a new improved microarchitecture in them. Of course, it cannot be said that the Skylake microarchitecture has become some kind of revelation, but the specific performance of the new processors has increased by another 10-15 percent compared to Haswell, and in addition, they received a new ninth-generation graphics core. Another important innovation that happened with the advent of Skylake processors was the transition to a new LGA1151 ecosystem with support for DDR4 memory.

To date, Skylake processors have successfully penetrated all market segments, from thin and light laptops to high-performance desktops. There are also traditional overclocker models Core i7-6700K and Core i5-6600K in the Skylake line, which by now have been able to break many records due to the transfer of the power stabilizer from the processor to the motherboard and, as a result, increased frequency potential. By the way, an almost forgotten overclocking technique unexpectedly returned with Skylake - by increasing the BCLK frequency. Moreover, at the very end of the past year, it became known that this approach is also applicable to conventional processors without the K index in the name, so from an overclocking point of view, Skylake has become a very important and expected update indeed.

However, we still cannot say that the market life of Skylake is good and cloudless. There are at least two problems with these processors. The first is production. Despite all the winning reports, the 14nm manufacturing process is still not working the way Intel would like. The low yield of suitable crystals capable of operating at high frequencies leads to a shortage of older Skylake models. As a result, the Core i7-6700K and Core i5-6600K processors are sold at slightly inflated prices, and in some regions are not available at all. The second problem concerns the mechanical design of the processors, whose textolite thickness has become noticeably thinner. Because of this, when using coolers with strong pressure, there is a risk of damage to the CPU. Of course, the probability of such an outcome is not too high, but, nevertheless, some cooler manufacturers were even forced to change the design of the mount they use.

While Intel was able to introduce fundamentally new processors twice during the past year, AMD was clearly not living in the present, but in the future. All her hopes are connected with the promising Zen microarchitecture, which will get into real products only at the end of this year or even at the beginning of next year. And therefore, in 2015, there were practically no announcements of new AMD products at all. During this time, the company only managed to clearly confirm its intentions to no longer develop the long-suffering Bulldozer microarchitecture, the latest reincarnation of which was the fresh Excavator design. However, this design will not get into the processors of the FX series at all, and its halo of habitat is limited exclusively to hybrid APUs, and on this moment only for mobile purposes. Actually, the first processors using Excavator cores were presented - under the code name Carrizo - in the middle of last year.

In theory, Carrizo was supposed to be an alternative to Intel's U-class processors, which are used in thin and light laptops. And they made some really big improvements compared to the previous generation of Kaveri mobile processors - performance increased, and TDP decreased by about 20 percent. However, even in this single niche, AMD has not been able to catch up with Intel. Its mobile processors turned out to be worse in terms of power consumption, weaker in performance and were noticeably inferior to mobile Skylake even in terms of the power of the graphics core.

As for the desktop market segment, which is more interesting for us, AMD did not carry out any innovations in it at all. The company only slightly raised the frequencies of Kaveri hybrid processors, giving them a new code name Godavari along the way.

This time we will talk about how to choose a 2015 processor. The selection criteria can be very different - this is the number of cores, and the manufacturer, what processor frequency to choose, and we will analyze much more now.

Also, I will try to show you the new processors of 2015, different prices and features.

Processor manufacturers

At the moment, there are two large companies involved in the development of processors - Intel And AMD. You may think that there is nothing to choose here, but this is not so. Both companies have their pros and cons. For example, Intel has quite a breakaway lead and there are reasons for this, namely higher performance and modern manufacturing technologies. AMD is the leader in terms of budget, the processors of this company, of course, are cheaper. There are constant disputes on the Internet about which processor from which company is better, and these disputes continue to this day. But you should not be guided by what people write or tell you, it is better to check everything yourself. The most optimal variant will be shown only by various tests.

Processor Reliability

We can say that processors are almost perfect, since they rarely fail by themselves. Out of 1000 processors, a maximum of 1 or 2 can fail in the first six months. Therefore, in order to avoid any problems, you need to have the right hands when assembling computers, and also, if you don’t know what you are doing, don’t get into internal components.

Processor overclocking and upgrading

Often, users practically do not change the processor to a newer and more powerful one. Also, it is worth noting that manufacturers of various components, including processors, often push us to buy a new device rather than upgrade the old one. You can do the following, purchase a processor with a margin, since the life of a computer is more than 5 years.

There is such a thing as overclocking. At first glance, the idea seems very tempting, but in fact, the performance gain from overclocking will be only 10-30% . And there is also a risk of severe overheating of the processor, and then its failure, so you need to be very careful when overclocking. It is also worth forking out for good system cooling so that the components do not overheat.

Moreover, in 2015 there are already inexpensive processors that have good performance, so overclocking almost makes no sense. It is best to purchase a good video card and then the performance will increase very well.

Features of choosing a processor

According to the results of tests at the end of 2015, the price range of inexpensive but good processors is about 5600 rubles. It is worth paying attention to the fact that processors with integrated video cards are approximately equal in price for both companies, and AMD processors without integrated video cards are much more profitable than Intel. You can also consider powerful processors in the price range from 250 y.e, which are owned by Intel.

On the Internet, there is often criticism about the boxed cooling system for the processor and it is recommended to use it from third-party manufacturers. Such cooling systems are very high quality, have a low noise level. It is desirable for powerful processors to have a good cooling system, this also applies to overclocking.

All the processors that I will list below are divided into several groups, and also sorted by performance. There are also models that have been on sale since October. Perhaps the price will be slightly different, so you can look in detail on the Internet, just by driving in a specific model.

Budget processors

AMD A4-4000- the processor is relatively inexpensive, suitable for a budget computer. Allows you to work on the Internet, play games, watch videos.

Intel Celeron G1820- a processor that performs the same functions as the previous one.

Medium processors

AMD A8-7600- this processor is good, it is 4-core with a built-in video card. You can play modern games at minimum, and even at medium settings.

Intel Core i5-4590- powerful 4-core processor, for powerful computers. Games run almost at maximum settings. At the moment there is also Intel Core i5-6500 which is suitable for DDR4 memory.

AMD Athlon X4 860K- also a 4-core processor, suitable for medium computers. Games at medium settings, sometimes at high settings if the corresponding video card is present.

Powerful processors

Intel Core i7-4790K- a powerful and fast processor that draws everything that is possible. 4 core. At the moment, there is a newer model, which, according to some characteristics, is even better, this is - Intel Core i7-6700K. There will be an interesting video about these two processors below, which I recommend watching.

The most powerful processor for the period of October 2015

Intel Core i7-5960X- the most powerful, 8-core processor, but the ratio of price and performance is extremely low. It makes sense to purchase such a processor in a purely sporting interest. If you have a lot of material resources, then please.

Price/Performance Table

And yet, perhaps, it is not worth it on my part to exaggerate so zealously. The development of modern technologies, albeit thoroughly slowed down, but does not stop for a minute. In the end, the one who even crawls moves towards the goal. The introduction of 14nm CPUs is an important and long-awaited event in the industry. Another bar has been overcome, new heights are on the way.

Unfortunately, as such, there was no competitive confrontation this year. AMD did not please us with new architectural solutions, limiting itself to cosmetic "finishing" of already known architectures. As they say, thanks for that, but we expect a real breakthrough from the Reds. I hope it happens in 2016.

Iron shop No. 28. Results of 2015: processors and computer platforms

AMD

If we are talking about central processors, then in 2015 AMD turned out to be the undisputed winner in the nomination “do not move bags”. As I said, the processor giant did not introduce any new products. It turns out that the latest development of the "red" - Kaveri hybrid processors - will be almost two years old in early January. In fact, for 24 months AMD has not shown anything fundamentally new. Their Bulldozer processor architecture with all the branches is completely outdated. The situation with current platforms is in a deplorable state. It is not surprising that it was in 2015 that the company passed in all positions. Thus, in the third quarter, the Reds earned $1.03 billion. This is 26% less income mined a year earlier. The company is unprofitable at 197 million. If we talk only about processors (CPU and GPU), then AMD sold chips in the amount of 424 million dollars in the third quarter, which is 46% less than a year ago. The increased sales of the new lines of Radeon R7 / R9 300 and Radeon R9 Fury video cards are partly pulled out by the “reds”. As one famous meme says: “everything is very bad.”

AMD FX CPUs turn four years old in September

What remains to be done by a company that has fallen into such an unenviable situation? That's right, roll out. In 2015, AMD made the decision to enter the market by selling intellectual property. The corporation wants to make money on patents. At the moment, the Reds are choosing between several business models. First, a company can license the patents themselves. How, for example, NVIDIA does it in relation to Intel graphics solutions. Secondly, AMD can license developments for devices that, in principle, it does not plan to release. The “reds” have interesting modifications to Radeon graphics designed for smartphones and tablets. Finally, thirdly, if necessary, AMD can sell patents from the portfolio. It is important to understand that the company does not plan to become an analogue of ARM. She won't be able to. Technology licensing is just another way to get out of the same place. In addition, it is a very profitable business. You don't have to look far for an example: Qualcomm Corporation in the first three quarters of the current financial year earned $6.162 billion from licenses alone. NVIDIA receives 66 million greenbacks from Intel every quarter.

And once again: this year AMD did not introduce a single fundamentally new central processor. There has been some buzz around Carrizo notebook chips based on the latest incarnation of Bulldozer's modular architecture, the Excavator. Talk about these “stones” has been going on since 2014, but how many laptops have been on sale since then? The question, however, is rhetorical.

We did not wait for new architectural solutions from AMD this year. All hopes are tied to 2016.

In 2015, AMD partners, contract manufacturers TSMC and GlobalFoundries, did a good job. Bringing to perfection 28- and 32-nm technological norms allowed the "red", firstly, to release a line of hybrid processors Godavari (Kaveri Refresh). There is nothing new in these APUs. The frequencies of both the x86 computing part and the Radeon graphics module were expectedly increased. Secondly, eight-core FX chips appeared, the heat pack of which was kept within 95 watts. That's all "achievements".

On the integrated graphics AMD A10-7870K you can play GTA V even in Full HD resolution

Obviously, 2015 was a transshipment year for AMD. The company has undergone a major restructuring, a change in leadership and major layoffs. Next year, the "Reds" promise to speak on all fronts. The main event for the corporation (perhaps the entire industry) will be the release of processors built on the Zen architecture.

Quite officially, the head of the company, Lisa Su, announced that computing cores, based on the latest development, will execute 40% more instructions per clock than solutions based on the Excavator used in the mythical Carrizo chips. The given figure, to put it mildly, dumbfounded. Especially when you consider that it is the cores that operate at the same frequency that are compared. The novelty will be produced at the GlobalFoundries factories using the 14nm 14LPP process technology. Already this year, the developers received prototypes for their tests. AMD itself is satisfied with both the performance of new chips and the percentage of yield of chips suitable for commercial implementation.

Attention, fasten your seat belts - we're going up!

Developed back in 2010, the modular Bulldozer architecture proved to be a failure. Time has put everything in its place. Zen is the exact opposite of Bulldozer. It is known that each core will include six integer operations execution units, two 256-bit floating point operations execution units, its own instruction decoder, instruction fetch unit, branch prediction unit, and its own cache. Such a core can be fully called full-fledged. Plus, the Zen architecture will support Simultaneous Multithreading technology, the principle of which is similar to Hyper-Threading. All this is reassuring.

For Zen chips, the AM4 platform will be introduced, supporting both APUs and x86 systems without integrated graphics. AMD will switch to using DDR4 memory. It is known that ASMedia will develop the logic for the Reds. AMD itself will abandon the production of chipsets.

In words, the architecture of Zen is great. But how will it be in practice?

There is only one point that is alarming: in 2010, a large number of beautiful speeches about the benefits of modular architecture were also heard from the lips of the management team of the corporation. In fact, everything turned out differently. Therefore, I do not want to make any predictions about Zen chips. They will come out - we will test and form a final opinion. I hope that AMD will succeed this time. The market needs competition.

A new product line of AMD chips. For one platform - AM4 - both hybrid processors with integrated graphics and powerful FX chips will be produced

Intel

The market needs competition. Intel has been resting on its laurels for a year now and is competing almost with itself. For the sake of justice, I note that the processor giant cannot be blamed for this. The chipmaker has a strategic plan that must be carried out at any cost. However, in 2014, Intel faced the first serious problems. The transition to more advanced technological standards requires more and more costs. At the same time, the complexity of such solutions increases exponentially. The tick-tock strategy failed for the first time in a decade.

As you know, Moore's law is 50 years old this year. For half a century, microelectronics has evolved based on this elementary rule of thumb, according to which the number of transistors in processors should double every 24 months. The industry has faced big amount obstacles in its path, but always emerged victorious. At one time, the turning point was the appearance of three-dimensional transistors. But now Intel is facing a bigger problem. With each new process used by the corporation to create commercially successful solutions, the time is approaching when it will simply not be possible to use silicon due to the physical limitations of this semiconductor material. Already now, releasing 14-nm processors Broadwell and Skylake, Intel is faced with a rather low yield of suitable crystals.

A decrease in the technical process leads to an inevitable increase in the production cycle of new chips.

Next stop - 10 nanometers, 50 silicon atoms! The first decisions will appear no earlier than the second half of 2017. Intel's ambitious plans include development and design integrated circuits based on 7- and even 5-nm process technology, but for this it is necessary that the industrial use of EUV scanners (Extreme Ultraviolet - radiation with a wavelength of 13.5 nm) becomes economically viable. According to some reports, this event will not happen until 2020.

Moore's law in numbers

But back to today. Like I said, the tick-tock concept was starting to fizzle out. Let me remind you that "tick" means the release of processors with the old architecture, but with a new (read - more subtle) technical process. “So” is the release of chips on a completely new architecture, but based on proven technological standards. For example, these are the Haswell solutions released in the first half of 2013. It took Intel exactly two years to rein in 14nm manufacturing and launch Broadwell desktop chips in June.

The line of chips turned out to be somewhat crumpled. Only five models are presented, but only two have LGA packaging (that is, they are independently installed in a socket on the motherboard). So those who believe that this time Intel hasn't really released "tick" processors are right.

In short, Broadwell is the same Haswell, but transferred to a 14-nanometer process technology. And yet a couple of architectural improvements have been implemented. So, the volumes of buffer zones and the scheduler window were increased. Exactly one and a half times the volume of the table of associative translation of addresses of the second level (L2 TLB) has grown - up to 1500 entries. Plus, the entire translation scheme has acquired a second miss handler. These changes allowed Broadwell processors to better predict complex code branches. The speed of execution of multiplication operations has increased from five cycles to three cycles. Division operations sped up the pace by using a 10-bit divider. Finally, vector gather instructions from the AVX2 set have been optimized. As a result, at the same frequency, the Broadwell architecture is faster than Haswell by an average of 5%.

Intel Broadwell Crystal

The whole point of desktop Broadwell is to use 128 MB of eDRAM memory, which plays the role of a fourth-level cache and is called Crystalwell, and the Iris Pro 6200 graphics module. It is rightfully considered the most productive integrated graphics to date. Even solutions from AMD are inferior to it. The Reds managed to lose to Intel in this field as well.

Iris Pro 6200 Performance

Just two months later, the Skylake line of “so” processors was released. A full-fledged series, numbering under two dozen models. All the same 14 nanometers, but on a new architecture. The flagships were the Core i5-6600K and Core i7-6700K models, respectively. The main platform feature is the presence of a dual memory controller that allows you to work with both DDR3 and DDR4 RAM. Intel itself quite simply dotted the i's between the new platform and the old architectural solutions. So, according to the manufacturer, the Core i7-6700K should be 10% faster than the Core i7-4790K, 20% faster than the Core i7-4770K and 30% faster than the Core i7-3770K. As testing has shown, the given figures are a very optimistic indicator, achievable by no means in all tasks. In some cases, Skylake's top "stones" outperform Haswell by only 3-5%. Sometimes they don't get ahead at all. The performance of the cache of the second and third levels has only noticeably increased. On average, throughput increased by a decent 60-70%! At the same time, cache delays remained at the same level. The integrated Skylake graphics (HD Graphics 530) are on average 20-40% faster than the HD Graphics 4600 used in Haswell.

Performance Comparison of Haswell, Broadwell and Skylake Architectures

There is no fundamental difference in performance between Haswell and Skylake. It is a fact. Therefore, there is no point in chasing a 5 percent advantage. In addition, today, while store warehouses are filled to capacity with old products that need to be sold and at the same time turn out to be in the black, there is a noticeable difference in cost. For example, in Moscow retail prices for motherboards and the processors themselves are overpriced by an average of 2000-3000 rubles. Still, Skylake chips consume less power. As a result, they heat up less and accelerate better. Intel, apparently, realized their mistake and abandoned the use of an integrated power converter (FIVR). And the platform itself turned out to be more functional.

Intel Core i7-6600K

About platforms

This is not the end of the Intel theme. The chipmaker is constantly criticized for the fact that he changes platforms very often, forcing him to take a motherboard in addition to the newly released processor. Skylake is one of those. But "often" is how much? After all, Intel has long been practicing the introduction of a new platform and new chipsets for the “so” generation. The cycle averages two years, but what will happen with 14-nanometer "stones" is unknown. Perhaps the problems with the transition to the 10-nanometer process technology will make the LGA1151 platform the longest-lived. This practice allows us to offer the user not only modern chips, but also functional motherboards for every taste and color. Logic of the 100th series (codename - Sunrise Point), designed to work with Skylake, for example, has native support for the NVMe protocol and 20 lines PCI Express 3.0, which is enough to integrate very fast SSDs into the system. The most sophisticated was the Z170 Express chipset. It allows you to solder up to six SATA 3.0 ports and up to 10 USB 3.0 connectors on the board without any auxiliary controllers. It is also the only chipset that supports overclocking features. All chipsets can work with both DDR3 and DDR4 RAM.

Z170 Express logic set

Without a doubt, the Z170 Express logic is of the most interest to enthusiasts. Firstly, only with it is it possible to overclock the Core i5-6600K and Core i7-6700K processors by the multiplier. There are no more overclocker "stones" for the LGA1151 platform. Secondly, motherboards built on this chipset allow you to overclock not only K-chips, but also all others by increasing the frequency of the clock generator. The web is full of results of the Core i3-6100 model, overclocked with liquid nitrogen to 6 GHz and higher. In practice, the user can save some money and take, for example, a Core i5-6400 processor, and then independently increase its speed from 2.7 GHz to 4-4.3 GHz. Intel simply allowed motherboard manufacturers to remove the binding of the frequency of the DMI and PCI Express buses to the frequency of the clock generator.

The LGA1151 platform went back to the roots of overclocking by changing the clock frequency.

When overclocking processors Intel Skylake it is worth paying close attention to a few things. Firstly, the transfer of the built-in power converter back to the motherboard again makes such a parameter as the quality of the power subsystem relevant. It becomes paramount during overclocking of the central processor. Secondly, when overclocking chips without an unlocked multiplier, it is worth considering that the integrated graphics core is automatically disabled for processors, as well as the functions of dynamic frequency change, C-states and turbo boost. Plus, the performance of AVX instructions is noticeably reduced.

ASRock Z170 OC Formula

Apart from LGA1151, there were no interesting platforms on the market. AMD does not make any sense to change anything for the existing APUs and FX series chips. You just have to wait for Zen and AM4. The release of Intel Broadwell-E processors has been delayed until Computex 2016. In any case, these chips will be compatible with the LGA2011-v3 platform. Therefore, it is not surprising that in 2015 Skylake processors and the LGA1151 platform, as well as the release operating system Windows 10 did not contribute to the growth of the market motherboards. It decreased by 10% in quantitative terms. So, ASUS and Gigabyte shipped approximately 17-17.5 million devices each. Other manufacturers have seen a more noticeable regression in sales. According to the information portal DigiTimes, it is planned that in 2016 the number of motherboards shipped will decrease by another 10%. It is possible that next year such manufacturers as Onda, ECS and Biostar will leave the market, although a requiem for them has been sung many times over.

ECS Z170-CLAYMORE

About memory

Still, the advent of the public platform LGA1151 affected the growth in demand for DDR4 memory. In just two months, it fell in price by 17% and came close to DDR3 in this indicator. To date, the average cost of a single 4Gb DDR4-2133 chip (according to DRAMeXchange) is $2.25. In September, such a chip cost $2.72, and in June - $3.62. As you can see, the price fell by 17% and 32%, respectively.

Talk about the high cost of DDR4 memory is speculation.

With all this, the cost of DDR3 also slightly decreases. A 4 Gb chip currently costs an average of $1.92, although back in September it was estimated at $2.17 and in June at $2.66. Declining demand for personal computers and increasing competition in the market are affecting. As a result, the difference in cost between DDR4 and DDR3 is 16.7%. Leading manufacturers everywhere are switching to 20nm process technology (SK Hynix - to 21nm process technology). AMD also plans to use DDR4 in the near future. The new Zen processors can shake up sales of personal computers. Intel will not stand aside either. All of this will keep the price down. random access memory.

The evolution of DDR memory

To date, two South Korean giants - Samsung and SK Hynix - rule the ball in the DRAM memory market. They control almost ¾ of the entire industry. According to the calculations of these companies, Samsung in the third quarter of 2015 took 45.9% of all sales, and SK Hynix - 27.6%. Last year, South Koreans owned 68.3% of the market.

In third place is Micron, but the company's business is getting worse every year. Americans occupy only 19.8% of the DRAM market. Even buying Elpida Memory in 2013 did not help Micron. The remainder of the market is shared between Nanya Technology and Winbond Electronics. They account for 2.8% and 1.3% of global DRAM sales, respectively.

The price of DDR4 drops noticeably, while the performance grows. And at a breakneck pace. The built-in memory controller of Skylake processors supports DDR4-2133 modules, however, the chip has dividers that allow you to run kits with an effective frequency up to 4133 MHz. A few months later, such a set appeared on sale. Tried overclocking giant - the company G.Skill. The 8 GB Trident Z PC4-33000 F4-4133C19D-8GTZ kit (two 4 GB modules) operates at an effective frequency of 4133 MHz at CL19 delays (25-25-45-2N). Such a set costs $ 300, that is, the same as the Core i7-6700K processor. Expensive, but what matters in this news is the very fact of how quickly DDR4 overcomes frequency barriers that are unattainable for the same DDR3.

G.Skill Trident Z PC4-33000 F4-4133C19D-8GTZ

About our

In fact, the appearance of Intel Skylake CPUs, as well as the LGA1151 serving platform and DDR4 memory, were the main events of 2015. But personally, I was pleased with the progress in the domestic microelectronics market. Now in Russia there are at least two self-sufficient developments.

The first is x86-compatible Elbrus processors developed by MCST. More precisely, in December, Izhevsk Radiozavod announced the delivery of the first batch of Elbrus-401 computers, which is based on the Russian quad-core Elbrus-4C chip. 24 GB DDR3 RAM is integrated into the system, expandable up to 96 GB, HDD 1 TB and solid state drive 128 GB mSATA form factor. Discrete is responsible for displaying the image. AMD graphics card Radeon 6000. Acceptable filling for an office computer. Probably even redundant for most tasks.

Yes, yes, yes, no one needs this “calculator” for 400 thousand rubles.

The volume of the first batch was 80 computers. The cost of one such computer is estimated at 400,000 rubles. Anticipating panic cries about who needs this "vacuum cleaner" for such money, I will answer that it is needed - the defense industry. The high cost is due to small-scale production of chips. In order to reduce the price by several dozen times, it is necessary to order a very large batch from a contract manufacturer. At least a few hundred thousand copies.

"Elbrus-401"

"Elbrus-4C" was presented back in 2014, but only now the development has found
(semi)commercial use. The processor is manufactured on a rather old 65nm process technology at the TSMC factory. Similar rules were considered relevant in 2007. Now Intel, as we have seen, is using 14 nanometers with might and main. Switches to this process technology in 2016 and AMD. Elbrus-4C has almost a billion transistors and, with a TDP level of about 60 W, operates at a frequency of 800 MHz. The peak performance of the domestic chip is 25 GFLOPs in 64-bit mode and 50 GFLOPs in 32-bit mode. To date, the speed of "Elbrus-4C" is approximately at the level Intel Atom D510 (1.66GHz, DDR2-800). This 45-nanometer chip was released in 2010 and was actively used in low-cost netbooks. It is no match for modern 4-core "stones".

Motherboard with "Elbrus-4C"

The 8-core Elbrus-8C processor, which is scheduled for release in 2016, looks much more promising. Firstly, the developers have transferred the architecture to a completely modern 28-nanometer process technology. Today, the current AMD and NVIDIA chips are produced according to exactly the same standards. The contract manufacturer is the same - TSMC. A thinner process technology allowed not only to double the number of cores, but to increase it to 1.3 GHz clock frequency. Each Elbrus-8C core is equipped with 512 KB of L2 cache. In total, the processor has 16 MB of shared L3 cache. The memory controller is 4-channel, DDR3-1600 modules are supported, including ECC. This results in peak throughput in single precision (FP32) operations of approximately 250 GFLOPs and 125 GFLOPs in double precision (FP64) operations. It turns out that it is five times faster than Elbrus-4C. For comparison, the double precision performance of the current 8-core AMD FX-8350 CPU is only 74 GFLOPs, while the Core i7-5960X is 384 GFLOPs. So Elbrus-8C is no longer so far behind modern AMD and Intel central processors. Let's see how MCST will deal with sales of this device.