Kaby Lake processor: review, testing, reviews

The end of Intel's production tick-to-tick pattern means that Kaby Lake has already become the third architecture based on a 14-nm processor. Starting with Broadwell (5th generation, "tick"), the manufacturer introduced the new microarchitecture Skylake (6th generation, "so"), which was optimized in the 7th generation. Improved energy efficiency and increased frequency was achieved due to less intense layout of transistors. Intel has released a large number of new processors from Kaby Lake, ranging from mobile KBL-U to 15 and 28 W and KBL-H by 45 W to KBL-S workstation models with a design consumption of 35-91 W. There are 3 overclocked options, including ³3.

Lake Kabi

The first official launch of Kaby Lake took place in September 2016 and included 6 mobile processors designed for installation in premium laptops and mini-PCs. They showed good results, and in early 2017, Intel introduced more than 25 new models. The main feature of Kaby Lake processors is the support for Optane Memory and chipsets of the 200 series. In addition, the Gen9 graphics were updated by Main10 and other low power video playback systems, and the circuit underwent a correction to improve the frequency curve of the voltage.

Overview of Kaby Lake processors

Intel defines its production lines for the segments Y, U, H and S. Recent changes in the naming scheme made it difficult to determine which segment the chip belongs to, if you do not know the TDP or the kernel schematic.

The Y series, using the nomenclature of Kaby Lake Pentium, Core m3, Core і5 / і7 and Core і5 / і7 vPro, is a 2- and 4-core processors with hyperpotential and rated thermal power of 4.5 W that are targeted at small and light mobile PC. Such low energy consumption is achieved thanks to an ultra low base frequency. This allows you to install batteries of lower capacity, providing a low weight and long battery life.

The U series consumes 28 and 15 watts, it has 2 cores with hyper-threading, but with a much higher clock speed. Includes processors Kaby Lake Pentium, Celeron, Core і3 / і7. They are often cheaper than the Y series, because they are not limited to stringent voltage and frequency requirements and are used in premium gaming laptops. Some processors are equipped with an additional 64 MB or 128 MB eDRAM chip, which serves as DRAM buffer with main memory and affects the graphics speed.

The H series chips have an estimated power consumption of 45 watts and provide the maximum performance of mobile devices. Intel is promoting them under the VR Ready brand, which indicates their use in virtual reality systems. Produced in various combinations of components and speed.

The S-series is designed for desktop PCs. Nothing noteworthy. It was released 3 modifications Core і7 with 4 cores of Kaby Lake processor and hyper-threading, one of which allows overclocking, and the other is characterized by low power. Also, several 4-core і5 are produced in similar modifications and 2-core chips і3.

In the new model series KBL-S, we can distinguish the possibility of overclocking Core і3-7350K, a 2-core processor with hyperpotency, an estimated thermal power of 60 W, a base frequency of 4.2 GHz (without turbo mode) and a configurable multiplier. This was a response to the requests of enthusiasts, who thus achieve the performance of the CPU, corresponding to devices of a higher class.

Speed Shift v2

One of the new features of Skylake was the Speed Shift function. If there is a correct driver, the system can refuse to control the turbo mode of the processor in favor of the processor itself. Using an internal metric collection in combination with access to system sensors, the CPU can adjust the frequency with greater accuracy and faster than the OS. The purpose of Speed Shift is to allow the system to respond more quickly to performance requests (for example, interact with a touch screen or browse the web), reduce delays and improve the user experience. Therefore, when the operating system is limited to predefined P-state parameters, a processor with Speed Shift support with the correct driver can almost continuously change the CPU frequency multipliers in a wide range of values.

The first iteration of the Speed Shift reduced the peak set time from 100 to 30 ms. The only limitation was the driver, which is now included with Windows 10 and is supplied by default.

With the advent of the new architecture, the hardware control of Speed Shift has improved. Intel did not change the name of the technology, but the improvements were significant. The driver has not changed, so it works with all the modifications of Speed Shift, but the processor can now reach the maximum frequency for 10-15 ms, and not for 30.

Optical memory

One of the goals of the memory industry is to create something with the speed of DRAM, but more robust, so that the data is preserved even in the absence of power. DRAM uses energy to update data, but it is the main source of data transfer software. Most of the acceleration of software depends on the speed of access to memory or the ability to have data closer to the core when necessary, so the presence of large, close, non-volatile memory can increase performance and reduce power consumption. Most of the decade was spent on its creation. Intel (and Micron) officially announced their decision, 3D XPoint, a year ago, but before that, it was not officially announced about its implementation.

Media capabilities

Although from the point of view of the functionality of Intel Kaby Lake is not very different from Skylake, there are obvious improvements in the graph. As in the case of CPU cores, the 14nm + process allowed to increase the frequency and improve the performance of the GPU, but perhaps more impressive changes are the updated multimedia capabilities. The main architecture of the Gen9 graphics processor has not changed, but Intel has revised video processing units, adding functionality and improving efficiency.

Hardware acceleration 4K

The main difference in the Kaby Lake-U / Y mediadvizhke is the availability of full hardware acceleration for encoding and decoding of 4K-video format HEVC Main10. This contrasts with Skylake, which supports 4k p30, but does this using a hybrid process that distributes the load between the CPU, the media processors and the shader cores of the GPU. As a result, Kaby Lake not only processes more HEVC profiles, but it only spends a fraction of its power with much higher bandwidth. Also in the new architecture was implemented 8-bit coding and 8/10-bit decoding of the codec VP9 from Google. Skylake offered hybrid decoding of the codec, which did not provide sufficient energy efficiency. The new hardware acceleration scheme HEVC Main10 and VP9 is part of the MFX unit. The video quality engine received support for HDR and Wide Color Gamut.

According to Intel, Kaby Lake U / Y is capable of processing up to 6 4K30 codecs AVC and HEVC simultaneously. Support for HEVC decoding is designed for 4Kr60 to 120Mbps, which is necessary for the playback of premium content and UHD Blu-ray. Thanks to the process improvements, even 4.5-watt Y chips are able to process HEVC 4Kr30 in real time. Thus, in the series U and Y, one of the main complaints about Skylake was resolved: the lack of hardware accelerated decoding of the 4Kp60 HEVC Main10. There are other improvements that provide a more satisfying multimedia experience for consumers.

Connectivity options

The flow of graphics processor Kaby Lake U / Y is the same as that of Skylake. This means that iGPU serves up to 3 displays simultaneously.

One of the disappointing aspects of Skylake, which was not eliminated in Kaby Lake-U / Y is the lack of its own HDMI 2.0 port with support for HDCP 2.2. Intel stands for the addition of LSPCon in DP 1.2. This approach has been used on several motherboards and even mini-PCs, such as Skull Canyon NUC (NUC6i7KYK) and ASRock Beebox-S.

Chipsets

The new PCH controller concentrators are interfaced with LGA1151 sockets and, thus, support both Skylake and Kaby Lake. The 100 series chips, such as the Z170, are also compatible with the new processors after the BIOS update.

Today, Intel's chipsets are fairly predictable. Z-series is focused on multigraphic chips and overclocking, H is distinguished by the absence of the latter, Q is designed for platforms with support for vPro and B is oriented to cheaper solutions.

There are also 3 mobile chipsets with similar differences, including a set for Xeon in CM238, which allows using new E3-1500 v6 processors.

Compatible Boards

Motherboard for Kaby Lake processors - ASUS Maximus ІХ Code, GIGABYTE Z270X, Supermicro С7Z270-CG, ASRock Z270, MSI Z270, ECS Z270H4-І. They have new controllers, including USB 3.1 10 Gb / s ASMedia ASM2142, which uses two PCIe 3.0 bands to support up to 2 ports. Previously, only 1 PCI 3.0 slot was used for this.

The Realtek ALC1220 audio controller has also been updated: there is an output of 120 dBA and an input of 113 dBA. This should provide the best measurable quality. The network connection is still made by the Intel I219-V Gigabit Ethernet controller. A big change here should be the introduction of the multi-gigabit Aquantia 5G / 2.5G AQC107. The new became the 10-Gbit / s USB 3.1 interface on the front panel in the MSI Z270 Gaming M7. Currently, it is activated via ASM2142, using two PCIe strips to provide one USB 3.1.

Technically, all motherboards equipped with Kaby Lake should be able to support Optane Memory. LED-backlighting also plays a big role in the motherboards of the 200 series: it is deprived only a few models in each price category.

Performance

As expected, there is no speed advantage. According to users' reviews, the 3-GHz Kaby Lake i7-7700K processor works similarly to the Core i7-6700K with a clock speed of 3 GHz (with hyperactivity turned off). The only difference is the support of memory. If Skylake is compatible with DDR4-2133, then Kaby Lake - with DDR4-2400, but this has little effect on almost all benchmarks.

Power consumption

One of the main advantages of the Kaby Lake processor is the same frequency at lower power or greater with the same power as Skylake. The i7-7700K supports turbo mode at 4.5 GHz with a thermal power of 91 watts. All tested processors Kaby Lake, even with manual overclocking, the consumption is close to the estimated, although usually the supplier of the CPU significantly overestimates the voltage required for stable operation of the chip.

Overclocking

According to the feedback from users, the perception of the increase in the clock frequency in Kaby Lake has changed due to the new AVX Offset function, which is found in the BIOS of each Z270 motherboard. It is known that AVX instructions cause damage to overclocking, reducing stability and making it difficult to advance the code without AVX. Now the user can apply an offset (for example, -10x), which will reduce the multiplier when the AVX command is encountered. This means that when the Kaby Lake processor is overclocked to 4.8 GHz with an AVX offset of 8x, the AVX command will be executed at 4.0 GHz, releasing less heat and keeping the system stable.

According to users, the 4.8 GHz AVX frequency is easily achievable even at a reasonable voltage. І7-7700K reaches 4.9 GHz with AVX -10 offset, and і5-7600K - 5.0 GHz, even with AVX enabled.

By and large, overclocking і7-7700K from 4.2 to 4.8 GHz does not give a practical advantage. The difference of 600 MHz corresponds to a 13-14% increase in productivity, which is not so much. However, given the voltage profile of the chips, the frequency of 4.5 GHz provides good temperatures and voltages, still exceeding ¿7-4790K or і7-6600K.

Test results

According to user feedback, the comparison of Kaby Lake processors confirms that Core і7-7700K wins in almost every test (except for a few, where і7-5775C is still better due to 128 MB of eDRAM).

Core і5-7600K works much the same, except for scenarios with a small number of threads (for example, when tracing rays), but in performing everyday tasks the processor certainly does not yield to anything. Core і5-7600K due to lack of growth ІРС is essentially the base і5-6600K, except for a few additional megahertz. The processor overclocked well - its temperature is much better than І7-7700K, but it's nothing unusual.

The elephant in the china shop, however, is Core і3-7350K. At $ 159, it's only $ 11 from Core і5-7400, which costs $ 170, but has 2 full two cores, albeit at a lower frequency (3 GHz vs. 4.2 GHz).

Has the new architecture of Intel become a new milestone?

For the most part, Kaby Lake does not offer any big changes. Optane memory support is a plus, but otherwise it's just a shift in the power and efficiency curve. The energy consumed at 3.0 GHz last year now yields 3.3 GHz, which means saving on time spent doing work or saving electricity. Speed Shift v2 - a really nice feature, but it is limited to users of Windows 10. Of greater interest is a set of new controllers (ALC1220, E2500, Aquantia). The optimization architecture does not cause admirable surprise, but it provides a 10% increase in efficiency.