Intel core i7 generation comparison. ⇡ Changes in microarchitecture that do not exist. ⇡ Kaby Lake line for desktop computers

Manufactured on Nehalem, Bloomfield and Gulftown microarchitectures. In this case, the internal clock frequency fluctuates around 3000 MHz. Integrated graphics are not supported on all models. The data bus frequency, as a rule, does not exceed 5 GHz per second.

Some configurations come with unlocked multipliers. To learn more about processors, you should consider Intel Processors Core i7 on specific microarchitectures.

Nehalem microarchitecture CPU

Processor Core clock frequency has around 2.8 GHz. In this case, there are four cores. The bus frequency of the CPU reaches 2400 MHz. Maximum voltage the system can withstand 1.4 V. The 2600K model is released with four cores. It has a clock frequency of 2.53 GHz. The CPU multiplier is of the unlocked type. The main bus frequency fluctuates around 2400 MHz. Model Core The i7 2700K has a clock speed of 2.93 GHz. The specified modification for four cores has an LGA connector. The bus frequency itself does not exceed 2400 MHz.

Bloomfield range

The Core i7 4720 processor has four cores. In this case, the chip area is 263 mm 2. The clock frequency itself is 2.6 GHz. The Core i7 4730 configuration has four cores. A total of 731 million transistors are used in it. The CPU clock frequency is 2.8 GHz. The Intel modification is designed for 3.07 GHz. In this case, the chip area is 263 mm 2. The bus itself is available at 213 MHz.

Gulftown microarchitecture CPU

The Core i7 970 model was released by the manufacturer with six cores. Its clock frequency does not exceed 3.2 GHz. The model has a bus at 2660 MHz. The Core i7 980 has a clock frequency of exactly 3.3 GHz. The chip area in this situation is 239 mm 2 . The bus itself is provided at 2660 MHz. The Core i7 processor has 990 transistors, 1170 million units. The clock frequency of the model does not exceed 3.4 GHz. The LGA connector is supported in this case.

Main functions

The area of ​​high-speed memory for processors based on the Gulftown microarchitecture is very extensive, therefore Intel Core i7 deserves good reviews from owners. The cache memory is directly related to the architecture. The model's cores are used dynamically. Thus, the system provides high performance. If we consider the Intel Core i7 4790, then the IM bus in this case is provided at 5 MHz. It plays an important role when exchanging information.

The system bus in the processor based on the Gulftown microarchitecture is used by SV. It is perfect for transmitting data to the controller unit. The interface is provided by the manufacturer with MI support. Direct connection is made through system board. All major operational teams are supported by it.

Performance

An Intel Core i7 laptop can support a maximum of four threads. In this case, the base frequency parameter is quite high. An IP program is provided to organize instructions. Processing data directly does not take much time. It is also important to note that the clock frequency parameter directly depends on the speed of computational cycles.

The calculated power in Intel processors is specified through a dot. The maximum frequency parameter is 38 GHz. The direct power of the CPU on the Gulftown microarchitecture is at the level of 83 Watts. When working at the base frequency, all cores in the processor are used.

Memory Module Specifications

The Intel Core i7 CPU on the Gulftown microarchitecture boasts a large amount of memory. In this case, it is supported in various formats. The number of channels directly influences system performance. In this modification there are two of them. Additionally, it is important to mention that the Intel CPU supports flex memory.

The throughput is at a very high level. In this case, reading data does not take much time. This was largely achieved through support for dual-channel memory. The high speed of data storage is another advantage of this system. ECC memory is supported by processors. A standard chipset for this is installed.

Graphics Specifications

For Intel Core i7 processors on the Gulftown microarchitecture, the parameter graphics frequency is at 350 MHz. In this case, it is also important to take into account the rendering indicator. It affects the base frequency quite strongly. The graphics subsystem itself can significantly improve rendering.

Support for the NS format is provided for Intel models. If we consider the Intel Core i7 2600K, then the maximum system volume is 1.7 GB. This indicator is very important for interface support. It also affects memory availability. To increase the interaction of a personal computer with a processor, the PPC system is used. Its resolution is 4096 x 2304 pixels.

Direct support

When describing the Intel Core i7 processor, it is important to mention Direct support. In this case, specific collections of application programs are taken into account. "Direct" series 11.1 is excellent for processing system files. If we talk about the graphical component, it is important to mention the Open Chart system. It affects the calculation of problems quite strongly. In this case, a lot depends on the support of multimedia files.

The Libera system was created to display two-dimensional graphics. If we talk about the Quick Video technology, then in this case you need to take into account the conversion speed. According to experts, the system interacts normally with portable media players. The Quick Video technology also affects the speed of video editing. Additionally, it provides placement on the Internet important information on work safety. Creating videos using this technology is very simple.

Expansion options

An Intel Core i7 computer uses the Express edition to transfer data. Today there are many versions of it, which, in fact, are not very different. However, in general, the Express editorial team is very important when it comes to connecting to personal computer various devices.

If we talk about version 1.16, then it can significantly increase the data transfer speed. This system can only work with PC type devices. It allows you to reproduce up to 16 channels directly. In this case, the basic modulator of the central processor is not involved in data processing.

Data Protection Technology

This technology allows you to work with the AE system, which is a set of commands. Thanks to it, you can quickly encrypt data. In this case, the process occurs safely. The AE system is also used to decrypt data. The program's many tools allow you to solve a wide range of problems. In particular, the AE system is capable of working with cryptographic data. It solves problems with applications quite quickly.

The Data Project technology itself was created to decipher random numbers. Authentication is carried out through them. Additionally, it should be noted that the Data Project technology includes the Key system. It is designed to generate random numbers. It helps a lot in creating unique combinations. The Key system is also involved in decoding algorithms. It is well suited for enhancing data encryption.

Platform Protection Technology

"Platform Protection" technology for Intel CPUs is provided in the 10.1 series. Speaking about it, first of all it is important to mention the Guard system. It was created for safe work with various applications. In this case, various operations can be performed with them.

The Gard system is also used to connect microcircuits. The Trusted program is used directly to protect platforms. It allows you to work with a digital office. The measured trigger function is supported by Platform Protection technology.

There is also an option for secure command execution. In particular, the system is able to isolate some flows. Wherein running applications they are not affected. To cancel hardware programs, the Anti-Tef system is used. In this case, the vulnerability of the CPU is significantly reduced. The Anti-Tef system is also designed to combat malicious software.

Anton Suchkov,

It's been about a year since Intel company introduced processors Ivy Bridge. In past developments, much attention was paid to power consumption, as well as increasing the power of the video core integrated into the CPU. What awaits us with the arrival of the latest new product - the processor Intel Haswell, and what direction is Intel currently heading?

Introduction

The next round of evolution gives rise to the LGA 1150 platform, which offers new model ranges system logic, and also, of course, processors. Last time, certain user groups practically scolded Intel for making such little progress compared to the previous generation. In other words, "old" Sandy Bridge looked good against the background of the new Ivy Bridge.

Time passed, and the platform with seventh series chipsets, as well as processors of the 4000th line, completely entered the market, tightly occupying all segments. The question “what to choose when buying from scratch” was no longer discussed. Prices for new models have dropped so much that they are not just equal, but even lower than the previous generation. Those who like to do a planned upgrade, however, disagreed, and many decided to wait for Haswell, expecting a decent jump in efficiency after replacing devices. The time has come; So is it time to upgrade?

Platform and architecture

Intel still uses the proven "tick-tock" development strategy, when two generations of processors are released using the same production standards, but the microarchitecture changes. New products with the code name Haswell are “so”, that is, they are produced using a 22 nm technological process, just like Ivy Bridge, while the work with 22 nanometers itself should already be debugged.

To set yourself new processor, you will have to purchase a motherboard with a suitable socket: LGA 1150. Yes, this time you will have to change the motherboard, and you won’t be able to get by with “little losses”, as was the case when switching from Sandy Bridge to Ivy Bridge. The lineup motherboards will use eighth series chipsets with Intel Z87, H87, Q87, B85 and H81 indices. As an example, let's look at one of the most common logic models for high-performance systems - the Z87 Express.

It is difficult to come up with something fundamentally new, so Intel probably follows the policy of “the best is the enemy of the good.” Thus, we do not see any fundamental differences in updated version older chipset for desktop systems; most changes have undergone quantitative improvements. To be more precise, the chip supports up to six native USB 3.0 ports versus four on the Z77, and now all six SATA ports operate at a transfer speed of 6 Gbps (previously, we recall, there were only two of these). Although AMD supporters know that the company has been equipping motherboards with chipsets that support all six SATA 6 Gb/s ports for quite some time, so the increase in the number of ports cannot be called a “revolution in the industry.” In a word, there is nothing outstanding here, take at least the PCI Express 3.0 lanes: they are still divided and work according to the x8 + x8 formula. That is, 3-Way SLI systems, which are so loved by 3DMark record breakers, will pass by most Z87-based motherboards. Memory support, limited to the DDR3 1600 MHz standard, also looks a little strange, especially now, when many models of higher frequency memory have appeared, and, in principle, there are no problems with its operation on systems.

By the way, about the positioning of the new platform: who is it for? The answer is simple - for everyone at once. In the near future, the Haswell line of processors should fill all segments, starting from the CPU entry level installed in typewriters to high-performance workstations. The “enthusiast sector” still stands apart, in which the platform with the LGA 2011 socket and multi-core processors, including their Extreme versions, rules the roost. By the way, it will also have an update; Ivy Bridge-E processors should appear in the foreseeable future, but that’s a completely different story.

As for the new architecture, then everything is not so simple. On the one hand, there are a number of changes, on the other hand, most of them went towards creating a more powerful video core. The main characteristics of the crystal, such as the number of physical cores and cache memory level, remain the same. The older model in the Intel Core i7-4770K line has four cores and a frequency of about 3.5 GHz, rising to 3.9 GHz in Turbo Boost mode. Powered by Intel Technology Hyper Threading the number of active threads increases to eight, the volume of the third level cache is 8 MB. Let's put it simply, we have already seen exactly the same parameters on the Intel Core i7-3770K processor, not just similar, but exactly the same. As for the TDP, it has only increased, from 77 W to 84 W.

Therefore, if we want to get more performance, we can only hope for an improved microarchitecture, which, as we know, does not add much efficiency, all other things being equal. The stagnation in the development of CPU power is especially noticeable if you look at the announcements of new processor cores over the past couple of years.

On the other hand, Haswell is equipped with support for new instructions such as AVX2 and FMA3. Additional kits commands can significantly help the processor in solving certain problems - however, if the software product does not support them, then there is no point in having them (yet). As practice shows, software developers need some time to adapt their creations; Sometimes it takes a year or two before manufacturers begin implementing instructions into code en masse. In any case, the Haswell user receives an unspoken advantage, which will be used on a “as soon as possible” basis.

So, we have already figured out where the transistor budget went (which, by the way, is only 0.2 billion more than that of Ivy Bridge) - into the new video core. During previous announcements, users received a huge number of dissatisfied reviews regarding GPU, built into the CPU. Most of the negativity was directed at the fact that powerful processor You don’t need graphics, since they don’t really allow you to play games, and they say they take up space. But let’s not forget that the main competitor in the processor market, AMD, has long been “eating the dog” in the implementation of GPUs. Moreover, let's be honest with ourselves, Intel's computing power is fine. Of course, you always want more, but still, there are no problems in this area, which could not be said about integrated graphics a few years ago.

Now HD Graphics not only allows you to easily “play” videos on YouTube without completely involving the central processor, but also watch videos in 1080p quality, which is especially in demand in the mobile segment. At one of Intel's recent presentations, interesting information was announced that Haswell is actually the first architecture that was developed by the company specifically for use in laptops, or, more accurately, actively promoted Intel ultrabooks. This state of affairs is primarily due to the fact that the PC market is not developing as actively as the mobile sector, and therefore the benefits of strengthening its position in it are not obvious.

Graphics options comparison table

The new HD 4600 core is better in everything. Almost all video encoding and decoding blocks have been updated (including the “4K” video class). Added support for DirectX 11.1, in addition, OpenGL 4.0 and OpenCL 1.2. Of course, the performance of the relatively old HD 4000 video card has increased. The main advantage of the new product in comparison with its predecessor was the increase in the number of stream processors from 16 pcs. up to 20 pcs. But this is not the strongest, but only an average graphics core among Haswell; Accordingly, some junior processors will receive an HD 4200 core, and models with the letter “T” can be equipped with an HD 5200 - the pride of Intel.

It is also worth noting that the “embedded” unit supports connecting up to three monitors, which can be combined into a single work area with 4K resolution.

One of the older processors in the line, the Intel Core i7-4770, arrived for testing. This CPU will delight its user with four cores with support for eight threads, a frequency range of 3.4 GHz - 3.8 GHz, as well as 8 MB of cache memory.

Compared to the Ivy Bridge generation, the processor has “lost” several contacts and has different “keys” for installation in the MP socket, which, coupled with Intel’s policy, makes it impossible for it to work on LGA1155 motherboards.

As you can see, the processor does not have the letter “K” at the end of its name, and therefore is not designed for overclocking and does not carry a multiplier unlocked for increase. Although we remember that at one time there was information about the return of “bus” overclocking, and indeed, in the BIOS options of motherboards for Haswell, a new item appeared for working with bus multipliers. So, now you can set the bus to 166 MHz, for example, instead of the 100 MHz standard, and get much higher clock speeds. However, as it turns out, this option is only available for “K” series processors; for all others (including the sample being tested), the option is hidden and it is impossible to overclock the processor. For example, the only thing that the motherboard participating in the stand (MSI Z87-G65 Gaming) was able to do even in auto-overclocking mode was to set the Turbo multiplier to the maximum position and fix it there. Therefore, unfortunately, the issue of Haswell overclocking remains open, which will not prevent us from making a comparison with the Intel Core i7-3770K in nominal operating mode.

Characteristics table

As we can judge from the table, most of the parameters of both processors are similar; the more interesting it will be to compare the new product with the new architecture.

Test stand

• Processor – Intel Core i7 3770K, Intel Core i7 4770
• Cooling System – Deep Cool Ice Warrior
• Thermal interface - Arctic Silver 5
• RAM – Corsair XMS3 1600 MHz, 9-9-9-24, 2x4GB
• Motherboard – ASUS Maximus V Formula Thunder FX, MSI Z87-G65 Gaming
• Hard drive – Intel SSD 320 Series, 160 GB
• Power supply – Seasonic Platinum 1000W
• Monitor – Dell U2711b, 2560 x 1440
• Video card - integrated
• Operating system - Win7 x64

Test results

Of course, the most important characteristic for a processor is its performance, so let's move on to testing.

The Intel Core i7-4770 processor turned out to be a little slower in some operations when working with memory, but overall it performed well, slightly beating its opponent in the WinRar and Fritz tests. Improvements are also noticeable when rendering in 3DSmax and, of course, Cinebench. Video encoding was especially good, where a significant advantage of 7 fps was achieved. As for the temperature regime, not everything is so smooth here. The increased TDP immediately hit the cores, showing performance about ten degrees Celsius higher than that of the previous generation CPU.

If the advantage of the new product in computing performance is not so great, then the graphical component of the tests has been worked out beyond all praise. In fact, this is the difference between “impossible to play because of the brakes” and a completely playable mode for undemanding users. I especially want to draw attention to the game GRID 2, when launched, the Intel HD Graphics logo is displayed. If you really want to, you can play this arcade simulator even at a resolution of 1920 x 1080 and low quality settings (not the lowest, but about there); The frame rate will be exactly 30 FPS.

The game can also enable additional options that are only available on integrated graphics. Intel cores, and the latest generation. As practice has shown, these options pleasantly change the graphics, adding several effects, while reducing the average frame rate by only 2-3 FPS.

Conclusion

Most likely, there will be people who are dissatisfied with the final result shown Intel processor Haswell. But in general, the new product has proven itself well, remaining slightly faster and more efficient than the previous generation in the computing department and having several undisclosed “trump cards” for the future, in the form of additional instructions. Very soon the market will be flooded with motherboards with 8-series onboard logic and Haswell processors, and prices will drop to Ivy Bridge levels, so the question of “what to buy” will be resolved by itself.

The question of upgrading older generations to Haswell remains open, because the increase in operating speed is not so noticeable to the unaccustomed eye, and the power requirements of the integrated video core are quite specific for users of home systems. In other words, unless you consider yourself an enthusiast and are not willing to shell out a decent amount of money for an upgrade, then there is no point in immediately switching from Ivy Bridge models to the new product. Although we have been observing this state of affairs for a couple of years now and perhaps it’s time to get used to it.

2017, which began a few days ago, is a year of big processor announcements. So, this year AMD should present processors at new architecture Zen, and Intel is about to introduce a new LGA2066 enthusiast platform. But all this will come later. In the first days of the new year, other processors will come to the fore - Intel Kaby Lake, which are aimed at mass systems, where the LGA1151 platform is currently used, followers of Skylake.

And to be honest, this is the most uninteresting announcement of the entire set of new products that is expected in the near future. A lot has been known about Kaby Lake for a long time, and all this information does not give much optimism. It is well known that the new processor is a slightly tweaked Skylake, which means it does not bring any special surprises. The fact is that Kaby Lake, in fact, is a forced patch on the canvas of Intel processor plans, and it was made in a relatively simple way and in a hurry.

A similar insignificant processor announcement has already happened in Intel history- in 2014, the company missed the deadline for the release of Broadwell and was forced to update its product range at the expense of and. Today's situation is largely similar: problems with the implementation of the next 10nm process technology are forcing Intel to come up with additional intermediate steps in the processor update relay.

However, Kaby Lake is still not such a pass-through model. In it, the microprocessor giant was able to introduce some improvements in the graphics core, but most importantly, the production of Kaby Lake now uses the second generation 14nm process technology. What all this can give to ordinary users and enthusiasts, we will analyze in this article.

⇡ New old technical process, or What is “14 nm+”

The key principle for Intel in the development of new processors, well known by the code name “tick-tock”, when the introduction of new microarchitectures alternated with the transition to more advanced ones technological processes, stalled. Initially, each stage in this pipeline took 12-15 months, but the commissioning of new production technologies with reduced standards gradually began to require more and more time. And in the end, the 14nm process finally broke the entire measured rhythm of progress. With the release of processors Broadwell generation such critical delays arose that it became clear: regular and methodical “tick-tock” no longer works.

Thus, mobile representatives of the Broadwell family hit the market almost a year later than originally planned. Older desktop processors appeared with a delay of almost a year and a half. And mid-level solutions based on this design have not reached the stage of mass products at all. Moreover, the introduction of Broadwell microarchitecture into complex multi-core processors it happened so slowly that when it finally reached senior server products in the middle of last year, the mobile segment had gone almost two generations ahead - and this is also clearly an abnormal situation. Even for companies the size of Intel, maintaining current state simultaneous implementation of several processor designs and several manufacturing technologies poses quite a serious challenge.

The upcoming transition to the next production technology promises no less problems, so the first processors released using the 10nm process technology can be expected no earlier than the second half of 2017. But if we remember that Intel began to use 14 nm technology in the third quarter of 2014, and Skylake processors appeared in mid-2015, it turns out that between Skylake and their 10 nm successors there is a too long, two-year pause that can have a negative impact both on the company's image and on sales. Therefore, in the end, Intel, in order to get rid of the constant lag behind the original plans and, if possible, unify its products, decided to radically change the development cycle and add an additional clock cycle to it. As a result, instead of the “tick-tock” principle, a new three-stage principle “process - architecture - optimization” will now be used, which implies longer operation of technical processes and the release of not two, but at least three processor designs according to the same standards.

This means that, in accordance with the new concept, Broadwell and Skylake should now be followed not by a transition to 10 nm standards, but by the release of another processor design using the old 14 nm standards. It was this additional design, developed as part of additional “optimization”, that received the code name Kaby Lake. We are already familiar with its first media, aimed at use in ultra-mobile devices - they came out at the end of last summer. Now the company is expanding Kaby Lake to other markets, including traditional personal computers.

Due to the fact that Kaby Lake is a kind of impromptu, which was forcedly designed by the microprocessor giant amid problems with the transition to a 10-nm process technology, the optimizations embedded in this processor relate not to microarchitecture, but primarily to production technology. The manufacturer even says that Kaby Lake is produced using the second generation of 14-nm process technology - 14-nm+ or 14FF+. In short, this means that quite significant changes have been made to the semiconductor structure of processor chips, but the resolution of the lithographic process still remains the same. More specifically, Intel's proprietary three-dimensional transistors (3D Tri-gate) in Kaby Lake received , On the one side, higher silicon channel ribs, and on the other hand, increased gaps between the gates of transistors, which actually means a lower density of semiconductor devices on the chip.

Unfortunately, Intel refuses to provide any specific information about how much its 14nm process technology has changed with the release of Kaby Lake. And most likely, this is due to the fact that these changes can be considered some kind of step back. When the company launched its 14nm manufacturing technology and announced its Broadwell generation of processors, it eagerly shared details and claimed that its FinFET process was superior to similar technologies used by other semiconductor manufacturers: TSMC, Samsung and GlobalFoundries. Now that the dimensions and profile of transistors have changed again as part of the 14nm+ process, their characteristics apparently no longer look as advantageous as before.

However, the absolute dimensions of transistors are only interesting for theoretical discussions about which semiconductor manufacturer owns the most advanced technology. A qualitative description of the changes is enough for us. Increasing the height of the edges of three-dimensional transistors, which are their channel, opens up the possibility of reducing signal voltages and, accordingly, minimizing leakage currents. Expanding the gaps between gates, on the contrary, requires increasing voltages, but it reduces the density of the semiconductor crystal and simplifies the production process.

These two changes, carried out simultaneously, partly compensate for each other - and therefore Kaby Lake crystals operate at the same voltages as Skylake. But Intel wins on another front: the improved technical process gives a better yield of usable crystals. Moreover, the resulting rarefaction in the arrangement of transistors makes it possible to reduce their mutual thermal and electromagnetic influence, and this entails an increase in the frequency potential. As a result, Intel managed to do without degrading the energy efficiency characteristics of the new design, but at the same time get a higher frequency or even overclocking reincarnation of Skylake.

Of course, this raises certain questions that relate to the cost of semiconductor crystals grown using the 14-nm+ process. Intel says the average transistor density in Kaby Lake has not changed compared to Skylake, but this is likely due to a redesign and better use of previously unused areas of the chip. However, Intel apparently still needed to change some equipment at the factories where Kaby Lake was launched. This, in particular, is indirectly indicated by the extended time of the announcement of Kaby Lake. Obviously, the company was unable to launch mass production of both ultramobile dual-core and powerful quad-core crystals precisely because of the need to reconfigure or re-equip production lines.

But the main thing is that the new technical process, which can be called Intel’s third 3D tri-gate process, really allowed the company to start producing chips with a higher clock frequency. For example, the base frequency of the older desktop Kaby Lake reached 4.2 GHz, while the flagship Skylake had 200 MHz more low frequency. Of course, in the absence of improvements in the microarchitecture, all this evokes some associations with Devil’s Canyon, but Kaby Lake is not just an overclocked Skylake. It turned out thanks to deep tuning, which affected semiconductor base processor.

⇡ Changes in microarchitecture that do not exist

Despite significant changes in manufacturing technology, no improvements have been made at the microarchitectural level in Kaby Lake, and this processor has exactly the same IPC (instructions executed per clock) characteristic as its predecessor, Skylake. In other words, the entire advantage of the new product lies in the ability to work at increased clock speeds and in certain changes in the built-in media engine regarding support hardware encoding and 4K video decoding.

However, for mobile processors Even seemingly minor innovations can have a noticeable effect. Ultimately, process improvements translate into improved energy efficiency, meaning the next generation of ultramobile devices will be able to offer longer battery life. In processors for desktop computers, we can get an additional increase of 200-400 MHz in clock frequencies, achieved within the previously installed thermal packages, but nothing more.

At the same time, at the same clock speeds, Skylake and Kaby Lake will produce completely identical performance. The microarchitecture in both cases is the same, so even the usual performance increase of 3-5 percent simply cannot come from anywhere. This is easy to confirm with practical data.

Usually, to illustrate the advantages of new microarchitectures, we use simple synthetic tests that are sensitive to changes in certain processor units. This time we used the benchmarks included in the AIDA64 5.80 test utility. The following graphs show the performance of older quad-core processors from the Haswell, Broadwell, Skylake and Kaby Lake generations running at the same constant frequency of 4.0 GHz.

All three groups of tests: integer, FPU and ray tracing rendering agree that at the same frequency Skylake and Kaby Lake produce completely identical performance. This confirms the absence of any microarchitectural differences. Therefore, it is right to treat Kaby Lake as Skylake Refresh: new processors bring an increase in performance only due to increased frequencies.

But Kaby Lake’s clock speeds don’t make much of an impression either. For example, when Intel released Devil's Canyon, the increase in nominal frequency reached 13 percent. Today, the increase in frequency of the older Kaby Lake model compared to the older Skylake is only about 7 percent.

And if we take into account that in the 14-nm Broadwell and Skylake processors the maximum frequencies were rolled back compared to their 22-nm predecessors, it turns out that the older Kaby Lake is only 100 MHz higher in frequency than Devil’s Canyon.

⇡ Kaby Lake line for desktop computers

Intel introduced the first processors of the Kaby Lake generation back in the summer. However, at that time these were only representatives of the energy-efficient Y and U series, aimed at tablets and ultra-mobile computers. All of them had only two cores and a GT2 class graphics core, that is, they were relatively simple chips. The bulk of Kaby Lake, including quad-core ones, are being released only now. Moreover, we are talking about updating the range of all classes of processors at once, including the 4.5-watt Core Y-series; 15- and 28-watt Core U-series with HD Graphics and Iris Plus; 45-watt mobile Core, including their versions with a free multiplier; 45-watt mobile Xeon; as well as a set of S-series processors for desktop computers with thermal packages of 35, 65 and 95 W.

Today's announcement affects a total of 36 various models processors, of which only 16 are desktop. But we will talk about them in detail today.

Previously when updating model range For desktop processors, Intel preferred to stagger the release of quad-core and dual-core chips. But this time the plan is slightly different. The company still did not immediately dump the entire range of updated LGA1151 processors onto the market, but the first batch of Kaby Lake desktop processors turned out to be more widespread than usual: it includes not only quad-core Core i7 and Core i5, but also dual-core Core i3. That is, during the second stage of the update, which will approximately take place in the spring, only processors from the budget Pentium and Celeron families will be presented.

The seventh generation Core i7 desktop processor family (which includes the Kaby Lake design) includes three models:

The Core i7 family still includes quad-core processors with support for Hyper-Threading technology and 8 MB of L3 cache. But compared to Skylake frequencies new Core i7 increased by 200-300 MHz, and in addition, the processors now have official support for DDR4-2400. Otherwise, the new items are similar to their predecessors. Recommended prices also remained at the usual level: Kaby Lake will replace representatives of the Skylake family in the old price categories.

Approximately the same picture emerges with Kaby Lake processors belonging to the Core i5 class. Except that the range here is much wider.

The Core i5 line of quad-core processors lacks Hyper-Treading technology, has a 6 MB L3 cache and offers slightly lower clock speeds compared to the Core i7. But, as in the case of the Core i7, Kaby Lake generation Core i5 processors are 200-300 MHz faster than their predecessors. Otherwise, they inherited the characteristics from Skylake without any significant changes.

But important changes have occurred in the Core i3 series. When the Kaby Lake design was introduced into this family, an overclocking processor with an unlocked multiplier was added to it, which, according to established tradition, received the letter K in the model number.

The Core i3 series combines dual-core processors with support for Hyper-Threading technology, equipped with 3 or 4 MB of L3 cache. The characteristics of the new Kaby Lake generation products again repeat the specifications of the corresponding Skylake with the only difference being the clock frequencies, which have become 200 MHz higher.

However, in addition to updated versions of the usual dual-core processors, a fundamentally new model has now appeared in the Core i3 series - Core processor i3-7350K, characterized by its overclocking capabilities. Previously among dual core processors Intel has never had such proposals (an experiment in the form does not count), but now the company seems to have decided to officially lower the barrier to entry into the world of overclocking. And the Core i3-7350K seems to be a very interesting option for budget-conscious enthusiasts, because its price is as much as 30 percent lower than the cost of the overclocker Core i5. Moreover, it is very likely that due to the reduced core area with low heat dissipation, this processor will also be able to please with high overclocking potential, which we will try to test in practice at the first opportunity.

A few words should be said about the graphics core of the new products. All desktop processors of the Kaby Lake generation received the same integrated GT2-level graphics, which includes 24 actuators - exactly the same number as the GT2 core of Skylake processors. And since the core GPU architecture hasn't changed in the new processor design, Kaby Lake's 3D performance remains the same. The appearance of a higher numerical index 630 in the HD Graphics name is entirely due to the new capabilities of the hardware media engine, to which tools for fast encoding/decoding of video in VP9 and H.265 formats were added, as well as full support for materials in 4K resolution.

⇡ New features of Intel QuickSync

From the point of view of traditional processor capabilities, Kaby Lake does not look like a serious step forward compared to Skylake. This feeling is created due to the fact that the new processor does not have any microarchitectural improvements. Nevertheless, Intel called the new processor its own code name - Kaby Lake, which is trying to convey the idea that this is not just Skylake with increased operating frequencies. And this is partly true. Some fundamental improvements that may be noticeable to end users are in the graphics core of the new CPUs. Despite the fact that the GPU architecture of Kaby Lake processors belongs to the ninth generation (like Skylake), its multimedia capabilities have expanded significantly. In other words, the basic design of the graphics core (including the number of execution units) in Kaby Lake remains the same, but the units responsible for encoding and decoding video content have undergone significant improvements in both functionality and performance.

Most importantly, the Kaby Lake media engine can now fully hardware accelerate encoding and decoding of 4K HEVC video with the Main10 profile. In Skylake, we recall, HEVC Main10 decoding was also announced, but there it was implemented using a hybrid scheme, and the load was distributed between the media engine, shaders of the built-in GPU and the computing resources of the processor itself. Because of this, high-quality playback was achieved only in the case of 4Kp30 video; more complex formats could not be played efficiently and without frame drops even on older CPU models. With Kaby Lake, such problems should not arise: the new processors decode HEVC video relying only on the media engine, and this allows them to digest complex profiles and high resolutions without loading the processing cores: with high efficiency, without frame drops and with low power consumption . Intel promises that specialized blocks of the Kaby Lake media engine can be strong enough not only to play 4K video at 60 and even 120 frames per second, but also to simultaneously decode up to eight standard 4Kp30 AVC or HVEC streams.

In addition, the Kaby Lake media engine received hardware support for the VP9 codec developed by Google. Hardware video decoding is possible with 8- and 10-bit color depth, and encoding with 8-bit. In Skylake, work with VP9 video, just as in the case of HEVC, was carried out using a hybrid hardware-software scheme. As a result, Kaby Lake may be very useful for those who like to watch 4K videos on YouTube, since the VP9 codec is being actively implemented in this service.

In general, the situation with hardware support for various video formats in Kaby Lake is as follows:

The block diagram of the Kaby Lake graphics part is shown in the illustration below. There are almost no structural differences from Skylake, but they are present at a lower level. Thus, hardware support for HEVC Main10 and VP9 has been introduced into the MFX (Multi-Format Codec) block. As a result, this particular unit was able to independently decode video in VP9 and HEVC formats with 10-bit color depth, as well as encode HEVC with 10-bit color and VP9 with 8-bit color.

In addition to MFX, the VQE (Video Quality Engine) block, which is responsible for the operation of the hardware encoder, has also been updated. Innovations are aimed at improving quality and performance when working with AVC video. Thus, Intel wants to gradually introduce the ability to work with HDR content and is systematically expanding the supported color at different stages of the pipeline. However, it must be kept in mind that this moment All encoding functions are oriented towards 4:2:0 chroma subsampling only. This is not a problem for amateur video work, but professional applications require more accurate 4:2:2 or 4:4:4 encoding, which is not yet available within Intel QuickSync.

It must be said that usually users of Intel desktop processors do not pay too much attention to the capabilities of media engines. After all, they are part of the graphics core, which in conventional high-performance systems is disabled in favor of a discrete video card. However, in fact, in modern Intel platforms, the media engine can be used even if you have a discrete video card. To do this, you just need not disable the built-in graphics, but activate it through Motherboard BIOS board as a secondary video adapter. In this case, two graphics adapters will be detected in the operating system at once, and after installing the Intel HD Graphics driver, the Intel QuickSync processor media engine will become available for use.

Let's give a few simple examples practical benefits of such a configuration.

Here, for example, is the situation with playing complex media content on the Core i7-7700K - 4Kp60 HEVC Main10 video with a bitrate of about 52 Mbit/s. Decoding is performed using Intel Quick Sync.

There are no frame drops, processor load is at minimum values. The integrated graphics of the Core i7-6700K, and even more so of processors with earlier designs, could not play this video without dropping frames. Therefore, to play such videos, previously it was necessary to rely on software decoding, which only works on high-performance platforms, and even then not always.

Another example is video transcoding. As part of our introduction to Kaby Lake, we looked at the performance of transcoding a source 1080p video using various software and hardware encoders. For testing purposes, we used the popular HandBrake 1.0.1 utility, which allows you to perform transcoding both through Intel QuickSync and programmatically using x264 and x265 encoders.

The tests used the standard Fast 1080p30 quality profile.

The performance benefits that can be achieved when transcoding using the hardware capabilities of a media engine are more than significant. Despite the fact that in both cases the result was approximately the same in quality with a bitrate of about 3.7 Mbit/s, the Intel QuickSync engine can offer many times more high speed transcoding, which also occurs with minimal load on the computing processor cores. True, the speed of hardware transcoding in Kaby Lake has hardly increased compared to Skylake.

Another example is streaming. Since Intel QuickSync allows you to encode video without loading the processing cores of the processor, streamers for their broadcasts can easily get by with one system with a Kaby Lake processor. For example, popular program for online broadcasts, OBS Studio supports H.264 encoding using the Intel media engine and is in this case capable of working in parallel with gaming applications running on a discrete video card, without reducing their performance.

In other words, even in a productive system equipped with external graphics card, there are a lot of applications for Intel QuickSync. And its increased functionality in Kaby Lake comes in handy. The hardware multimedia capabilities of this unit, which has become almost omnivorous, truly expand the scope of use of a typical personal computer.

Speaking about the graphics core built into Kaby Lake, we cannot fail to mention that, like in Skylake, it can support up to three 4K monitors simultaneously. However, despite expectations, native support for the HDMI 2.0 interface has not appeared in new generation desktop processors. This means that monitors connected via HDMI on most motherboards will only be able to provide a maximum resolution of 4096 × 2160 @ 24 Hz. Full 4K resolution, as before, will be available only when using a DisplayPort 1.2 connection. However, there is an alternative solution that allows system manufacturers to equip HDMI 2.0 outputs; it consists of using additional LSPCon (Level Shifter - Protocol Converter) converters installed in the DP path. However, this approach naturally requires additional costs.

However, Intel promises that systems based on Kaby Lake processors will be able to play premium 4K content protected by DRM (for example, from a premium Netflix account) without any special compatibility issues. If there is no HDMI 2.0 port, a system with DisplayPort connected to a 4K TV or monitor that supports HDCP2.2 will also work for this.

As a result, the Kaby Lake media engine provides an answer to the main complaint against Skylake - the lack of hardware acceleration of 4Kp60 HEVC Main10. Plus, some other useful features and improvements have been added, as a result of which the integrated Kaby Lake graphics are truly better suited to work with the increasingly popular 4K video and content streaming services. However, you need to keep in mind that hardware improvements alone are not enough to introduce new functions, and there is a lot of work ahead to update and adapt the software.

⇡ Chipsets for Kaby Lake: Intel Z270 and others

By tradition, along with new processors, Intel also brings to the market new sets of system logic. That is, despite the fact that the “tick-tock” principle was replaced by the “process - architecture - optimization” principle, everything with chipsets remains the same: they are updated at every turn of progress. However, this time the minor improvements in Kaby Lake compared to Skylake allowed us to maintain full compatibility with the old platform. Kaby Lake is not only installed in the already familiar LGA1151 processor socket, but also works perfectly in motherboards with older 100th series chipsets.

The optimizations that occurred in the production technology of new processors did not require changes in the power supply. As in the case of Skylake, Kaby Lake should have it on the board, and not in the processor. At the same time, the requirements for voltages and currents remained the same as before. This means that there are no circuitry obstacles to installing Kaby Lake in old LGA1151 boards. The only thing that is required to support new CPUs with old boards is the presence of the appropriate microcode in the motherboard BIOS. And most boards based on Z170 and other chipsets of the previous generation received the necessary update in a timely manner.

New logic sets with model numbers from the 200 series were designed by Intel more out of habit and simply so that motherboard manufacturers have some reason to update platforms. Therefore, it is not surprising that in terms of capabilities, the differences from previous chipsets turned out to be minimal and, one might say, even cosmetic. None really useful additions in the form of support USB interfaces 3.1 or Thunderbolt did not appear in the Intel Z270 and other chips in the series, and the main improvement that Intel is pushing for is support for promising drives Intel Optane.

This is how they relate to each other purely specifications older chipsets in the hundredth and two hundredth series:

Moreover, with regard to the main marketing argument in favor of the 200 series chipsets - Optane support, Intel is largely disingenuous. In fact, Optane drives will not require any special interfaces or connectors. To operate, they will need a regular M.2 slot with a PCI Express 3.0 x4 bus installed in it, and many older LGA1151 boards have such slots. In the case of new logic sets, we are simply talking about the fact that the number of PCI Express lanes in them is slightly increased, and this allows board manufacturers to easily add more than one M.2 slot to their platforms. The fact is that, as expected, the first versions of Intel Optane will not replace conventional SSDs. They will receive extremely small volumes and will be positioned as additional caching drives, so it is planned to allocate a separate independent slot for them, which is easier to implement in chipsets of the 200 series. In addition, a special Rapid Storage Technology driver will be made for the new chipsets, which will contain some operating algorithms optimized for Optane, essentially similar to new version Intel Smart Response Technology.

Thus, the significant difference between the Z270 and the Z170 should be considered not the far-fetched Optane support, but the increased by four (to 24) maximum number of PCI Express 3.0 lanes supported by the chipset. Moreover, this change was reflected in the change in the I/O Port Flexibility scheme, within which it is now allowed simultaneous implementation 30 high-speed interfaces at once. The number of SATA and USB ports has remained at the old level, but in the Z270, the USB 3.0 standard can accommodate not 8, but 10 ports.

Many new chipsets of the 200 series consist of more than just one Intel Z270. We decided to focus on it because it is the most equipped and the only one that supports overclocking the processor (both through changing the multipliers and the frequency of the base clock generator). However, besides this, the line of new chipsets includes a couple of simpler consumer chipsets - H270 and B250, as well as a couple of chipsets for the corporate environment - Q270 and Q250, which are distinguished by the presence of a set of Intel Standard Manageability functions for remote control and administration.

The most interesting for ordinary users The H270 and B250 differ from the Z270 not only in the lack of overclocking capabilities. They have reduced the number of PCI Express 3.0 lanes and USB 3.0 ports, and also reduced the number of M.2 interfaces that can be connected to the Intel RST driver. In addition, low-end chipsets do not allow dividing processor bus PCI Express on several slots.

A complete picture of the correspondence of the characteristics of the 200 series logic sets can be obtained from the following table.

⇡ Test processor: Core i7-7700K

For testing, we were provided with the senior representative of the Kaby Lake desktop line, Core i7-7700K.

This quad-core processor with support for Hyper-Threading technology and an 8-MB L3 cache has a nominal clock speed of 4.2 GHz. However, the test showed that in practical conditions Core frequency The i7-7700K is 4.4 GHz with an all-core load and 4.5 GHz with a low-threaded load. Thus, in terms of frequencies, the older Kaby Lake managed to overtake not only the old one, but also the old one, which until recently remained the highest-frequency Intel processor for desktop systems.

The operating voltage of our sample was 1.2 V: there are no significant differences from processors of previous generations.

When idle, the Kaby Lake frequency drops to 800 MHz, and, in addition to the usual Enhanced Intel SpeedStep technology, the processor supports more new technology Intel Speed ​​Shift. It transfers frequency control from operating system the processor itself. Due to this, a significant improvement in response time to a changing load is achieved: the processor comes out of energy-saving states faster and, if necessary, turns on turbo mode faster. But there is a limitation: Speed ​​Shift technology only works in Windows 10.

Left - Core i7-7700K (Kaby Lake), right - Core i7-6700K (Skylake)

Certain changes have also occurred with appearance CPU. True, they are more of a cosmetic nature. For example, Intel did not abandon the use of thin PCB, which appeared in Skylake, in Kaby Lake. But the shape of the heat distribution cover has changed. It has additional tides that increase the contact area with the cooler sole. However, this will most likely have little effect on the efficiency of heat removal. After all the main problem in the path of heat from the processor chip - the polymer thermal interface is not best quality, which is located under the processor cover. And in this regard, everything is as before: highly efficient solder remains the prerogative of flagship processors in LGA2011-v3 performance.

There are also changes on the processor side. However, Kaby Lake remains compatible with the LGA1151 socket, so there are very few differences compared to Skylake. The stabilizing circuit remained the same, so the set of hanging elements was preserved. A slight difference can only be noticed in their relative position.