Processors. Processors Total score in tests with multimedia files

Description of test systems and testing methods

In the light of the title of this material, the main heroes of the test were the Core i7-5820K and Core i7-4790K processors. However, to make the testing more meaningful, the background for comparing these processors was the results of past and current high-performance flagships - Core i7-4960X with Ivy Bridge-E design, as well as Core i7-5960X and Core i7-5930K with Haswell-E design.

As a result, the list of hardware components involved in testing looked like this:

Processors:
- Intel Core i7-5960X Extreme Edition (Haswell-E, 8 cores + HT, 3.0-3.5 GHz, 20 MB L3);
- Intel Core i7-5930K (Haswell-E, 6 cores + HT, 3.5-3.7 GHz, 15 MB L3);
- Intel Core i7-5820K (Haswell-E, 6 cores + HT, 3.3-3.6 GHz, 15 MB L3);
- Intel Core i7-4960X Extreme Edition (Ivy Bridge-E, 6 cores + HT, 3.6-4.0 GHz, 15 MB L3);
- Intel Core i7-4790K (Haswell Refresh, 4 cores + HT, 4.0-4.4 GHz, 8 MB L3).
CPU cooler: Noctua NH-D15.
Motherboards:
- ASUS X99-Deluxe (LGA2011-v3, Intel X99);
- ASUS Z97-Pro (LGA1150, Intel Z97);
- Gigabyte X79-UP4 (LGA2011, Intel X79).
Memory:
- 2x8 GB DDR3-2133 SDRAM, 9-11-11-31 (G.Skill F3-2133C9D-16GTX).
- 4x4 GB DDR3-2133 SDRAM, 9-11-11-31 (G.Skill F3-2133C9Q-16GTX);
- 4x4 GB DDR4-2666 SDRAM, 15-15-15-35 (G.Skill F4-2666C15Q-16GRR);
Video card: NVIDIA GeForce GTX 980 (4 GB/256-bit GDDR5, 1127-1216/7012 MHz).
Disk subsystem: Crucial M550 512 GB (CT512M550SSD1).
Power supply: Seasonic Platinum SS-760XP2 (80 Plus Platinum, 760 W).

Testing was performed in the operating system Microsoft Windows 8.1 Professional x64 with Update using the following driver package:

Intel Chipset Driver 10.0.17;
Intel Management Engine Driver 10.0.0.1204;
Intel Rapid Storage Technology 13.2.4.1000;
NVIDIA GeForce 344.75 Driver.

The Intel Core i7-4790K processor and the protagonist of today's test, the Core i7-5820K, were tested twice - not only in nominal mode, but also with their stable and long-term overclocking, achievable with the cooling we use:

Core i7-5820K @ 4.1GHz @ 1.225V;
Core i7-4790K overclocked to 4.5GHz at 1.2V.

Description of the tools used to measure performance:

Benchmarks:
- Futuremark PCMark 8 Professional Edition 2.3.293 - testing in Home (typical home PC use), Creative (PC use for entertainment and multimedia content) and Work (PC use for typical office work) scenarios.
- Futuremark 3DMark Professional Edition 1.4.828 - testing in Sky Driver, Cloud Gate and Fire Strike scenes.
Applications:
- Adobe Photoshop CC 2014 Processing Performance Test graphic images. Measured is the average execution time of a test script, which is a creatively redesigned Retouch Artists Photoshop Speed Test, which includes a typical processing of four 24-megapixel images taken by a digital camera.
- Adobe Photoshop Lightroom 5.7 - Performance Testing batch processing series of images in RAW format. The test scenario includes post-processing and export to JPEG at 1920x1080 resolution and maximum quality of two hundred 12-megapixel RAW images taken with a Nikon D300 digital camera.
- Adobe Premiere Pro CC 2014 - performance testing for non-linear video editing. Measures rendering time to H.264 Blu-ray for a project containing HDV 1080p25 footage with various effects applied.
- Autodesk 3ds max 2015 final rendering speed test. The time taken to render at 1920x1080 resolution using the mental ray renderer of one frame of the standard Space_Flyby scene from the SPEC test package is measured.
- WinRAR 5.1 - archiving speed testing. The time taken by the archiver to compress a directory with various files with a total volume of 1.7 GB. The maximum compression ratio is used.
- x264 r2491 - testing the speed of video transcoding to H.264/AVC format. To evaluate the performance, the original [email protected] AVC video file with a bit rate of about 30 Mbps.
- X265 1.4+142 8bpp - testing the speed of video transcoding to the promising H.265/HEVC format. For performance evaluation, the same video file is used as in the x264 encoder transcoding speed test.
Games:
- Civilization: Beyond Earth. Settings for 1280x800 resolution: DirectX11, Ultra Quality, Anti-aliasing = Off, Multithreaded rendering = On. Settings for 1920x1080 resolution: DirectX11, Ultra Quality, 8x MSAA, Multithreaded rendering = On.
- Company of Heroes 2. Settings for 1280x800 resolution: Maximum Image Quality, Anti-Aliasing = Off, Higher Texture Detail, High Snow Detail, Physics = Off. Settings for 1920x1080 resolution: Maximum Image Quality, High Anti-Aliasing, Higher Texture Detail, High Snow Detail, Physics = High.
- F1 2014. Settings for 1280x800 resolution: Ultra Quality, 0xAA, DirectX11. Settings for 1920x1080 resolution: Ultra Quality, 8xAA, DirectX11. The Texas track is used.
- Hitman: Absolution. Settings for 1280x800 resolution: Ultra Quality, MSAA = Off, High Texture Quality, 16x Texture Aniso, Ultra Shadows, High SSAO, Global Illumination = On, High Reflections, FXAA = On, Ultra Level of Detail, High Depth of Field, Tesselation = On, Normal Bloom. Settings for 1920x1080 resolution: Ultra Quality, 8x MSAA, High Texture Quality, 16x Texture Aniso, Ultra Shadows, High SSAO, Global Illumination = On, High Reflections, FXAA = On, Ultra Level of Detail, High Depth of Field, Tesselation = On , Normal Bloom.
- Metro: Last Light Redux. Settings for 1280x800 resolution: DirectX 11, High Quality, Texture Filtering = AF 16X, Motion Blur = Normal, SSAA = Off, Tessellation = High, Advanced PhysX = Off. Settings for 1920x1080 resolution: DirectX 11, Very High Quality, Texture Filtering = AF 16X, Motion Blur = Normal, SSAA = On, Tessellation = High, Advanced PhysX = Off. When testing, Scene 1 is used.
- Middle Earth: Shadow of Mordor. Settings for 1280x800 resolution: Lighting Quality = High, Mesh Quality = Ultra, Motion Blur = Camera and Objects, Shadow Quality = High, Texture Filtering = Ultra, Texture Quality = High, Ambient Occlusion = Medium, Vegetation Range = Ultra, Depth of Field = On, Order Independent Transparency = On, Tessellation = On. Settings for 1280x800 resolution: Lighting Quality = High, Mesh Quality = Ultra, Motion Blur = Camera and Objects, Shadow Quality = Ultra, Texture Filtering = Ultra, Texture Quality = Ultra, Ambient Occlusion = High, Vegetation Range = Ultra, Depth of Field = On, Order Independent Transparency = On, Tessellation = On.
- Thief. Settings for 1280x800 resolution: Texture Quality = Very High, Shadow Quality = Very High, Depth-of-field Quality = High, Texture Filtering Quality = 8x Anisotropic, SSAA = Off, Screenspace Reflections = On, Parallax Occlusion Mapping = On, FXAA = Off, Contact Hardening Shadows = On, Tessellation = On, Image-based Reflection = On. Settings for 1920x1080 resolution: Texture Quality = Very High, Shadow Quality = Very High, Depth-of-field Quality = High, Texture Filtering Quality = 8x Anisotropic, SSAA = High, Screenspace Reflections = On, Parallax Occlusion Mapping = On, FXAA = On, Contact Hardening Shadows = On, Tessellation = On, Image-based Reflection = On.

⇡ Performance in complex tests

Those users who purchase systems built on high-end processors are not very worried about the performance in complex tests. The fact is that such tests simulate the work of commonly used applications, which in fact can rarely load four or even more processor cores, especially if they work with Hyper-Threading technology. And this means that in this case, the PCMark 8 result is much more influenced by the frequency of the CPU, and not by its ability to multi-thread data processing.

Therefore, it is not surprising that the quad-core Core i7-4790K is noticeably ahead of the Core i7-5820K, because Devil's Canyon stands out from all modern Intel processors precisely with its high clock speed. Moreover, the advantage of the LGA1150 flagship is so high that even the overclocked Core i7-5820K does not always reach its results. However, do not take what you see in the above diagrams to heart. Further, in real resource-intensive applications, we will see a completely different picture.

⇡ Application performance

It is to perform computationally heavy tasks that it is worth using processors with a large number of cores. The six-core Core i7-5820K can provide higher performance than the Core i7-4790K in final rendering, video editing and transcoding, image array processing, data compression, and in many other cases. Moreover, in the most complex tasks, for example, in 3ds max 2015 or in modern x264 and x265 encoders, even the overclocked Devil's Canyon is inferior to the younger Haswell-E. This clearly indicates that if your activity is related to the creation of content, six-core processors are clearly preferable to quad-core ones. Moreover, such high performance has now noticeably fallen in price: the Core i7-5820K offers about the same level of performance as the Core i7-4960X a year ago, but two and a half times cheaper.

In passing, we note that, despite our complaints about overclocking, which we were able to achieve with the Core i7-5820K, the performance gain was not so small. Increasing the frequency of this processor to 4.1 GHz allowed for an average 15 percent increase in the speed of resource-intensive tasks. And this is clearly more than what you can count on when overclocking the Devil's Canyon, which is accelerated almost to the maximum by the manufacturer initially. It is curious that the overclocked Core i7-5820K in some cases boasts even higher performance than the current LGA2011-v3 flagship, the Core i7-5960X. However, do not forget that this is due to the low nominal frequency of the eight-core model, which, like the Core i7-5820K, is subject to overclocking to about the same levels. That is, if you do not look back at budget issues, then in order to obtain maximum performance in “heavy” tasks, it’s still worth relying on the older eight-core Haswell-E.

⇡ Game performance

Many owners of high-performance systems are not so much concerned with the speed of processors in resource-intensive applications, but with the gaming performance that they can provide. And this is where the confrontation “older quad-core versus younger six-core” can escalate with renewed vigor. As you know, many games do not need a large number of computing cores, and therefore the Core i7-4790K with a high clock speed at first glance looks more preferable. However, it is possible that the first impression is not entirely correct. After all, the Core i7-5820K cannot be called a brake, besides, it can offer more capacious cache memory and faster quad-channel DDR4, and such things also affect gaming performance.

Testing in real games is preceded by the results of the 3DMark synthetic benchmark, which gives a certain average metric of 3D gaming system performance.

Futuremark 3DMark is well optimized for the multi-core structure of modern processors, so it paints a kind of idealized picture of what would happen if game engine manufacturers bet on multithreading. Here, the six-core Core i7-5820K is slightly ahead of the older quad-core Core i7-4790K, and overclocking further increases this advantage.

Testing in real games rarely reveals fundamental differences between high-performance processors. With a modern gaming load, the bottleneck is not the computing resources of the platform, but its graphics subsystem. That is why in most cases it is completely indifferent which of the processors is used in a particular gaming platform. The number of FPS, most likely, will depend on this extremely slightly. However, this does not give reasons to refuse testing in games. Just for the sake of illustration, along with measuring gaming performance at a typical Full HD resolution of 1920x1080 with FSAA enabled, we also measure at 1280x800. The results in the first case show the level of FPS that can be obtained in real conditions right now, while the second test option allows us to evaluate the theoretical gaming performance of processors, which may be revealed in the future if we have faster options for the graphics subsystem.

Tests in Full HD resolution:

As you can see from the above diagrams, the gaming performance of the Core i7-5820K and Core i7-4790K is almost the same. When set to Full HD resolution, both processors can handle the full load of the flagship graphics card GeForce GTX 980, and we do not see any fundamental differences in the number of frames per second. If we approach the numbers on the diagram quite meticulously, then the Core i7-4790K is still a little faster, but this slight superiority is actually due not even to the higher clock speed of the quad-core, but to the features of the LGA2011-v3 platform, the PCI Express controller in which works with slightly higher latency due to its complexity.

Along the way, I would like to emphasize one more interesting fact: as a gaming processor, the Core i7-5820K turned out to be a slightly better choice than the flagship Core i7-5960X. The reasons for this phenomenon do not raise questions - the frequency of the eight-core Haswell-E is lower than that of the Core i7-5820K, and six cores are more than enough for any modern game. Therefore, it makes no sense to assemble gaming systems based on today's Extreme Edition processor. Other LGA2011-v3 six-cores will be at least as good.

Reduced resolution tests:

If, when measuring generalized gaming performance, we push the limits set by the limited performance of a graphics card, then the charts with performance results become much less uniform. And speaking of theoretical gaming CPU performance, the regular quad-core Haswell (Devil's Canyon) looks better than the Haswell-E and Core i7-5820K processors among them. CPU speed is more important for games than six cores, more cache and four-channel memory, at least in relation to Haswell microarchitecture media.

⇡ Energy consumption

All Haswell-E processors have the same TDP of 140W, and the Core i7-5820K is no different in this regard. This is quite natural, because the six-core and eight-core Haswell-E are based on the same semiconductor crystal, and the six-core clock speeds are higher than that of the older flagship Core i7-5960X processor. However, our study today compares the Core i7-5820K with the Core i7-4790K, which has a 40 percent lower TDP. Do the indicators of real energy consumption (and, consequently, heat release) differ so drastically?

The following graphs show the total consumption of the systems (without monitor) measured at the output of the socket into which the power supply of the test system is connected, which is the sum of the power consumption of all components involved in the system. The efficiency of the power supply itself is automatically included in the total indicator, however, given that the PSU model we use, Seasonic Platinum SS-760XP2, is 80 Plus Platinum certified, its impact should be minimal. To properly assess energy consumption, we have activated the turbo mode and all available energy-saving technologies.

When idle, the LGA2011-v3 platform consumes noticeably more than the LGA1150. We already know this, and there are several reasons for this phenomenon. Firstly, the Intel X99 system logic set is one and a half times more voracious than the Z97. Secondly, processors of the Haswell-E family are deprived of support for some energy-saving technologies, in particular, the C7 state. Thirdly, motherboards with processor socket LGA2011-v3 have a notoriously more complex design and are equipped with a large number controllers.

When solving a common multi-threaded video transcoding task with an x265 encoder, a system based on a six-core Core i7-5820K processor consumes 24 W more than a similar configuration with a Core i7-4790K. However, if we correlate this figure with the difference in their speed, it turns out that the junior representative of the line offers the best specific performance in terms of each watt of electricity consumed. True, during overclocking, the situation changes to the opposite - the consumption of the Core i7-5820K increases very sharply with an increase in the clock frequency above the nominal value.

The following diagram shows the maximum consumption under the load created by the 64-bit version of the LinX 0.6.5 utility with support for the AVX2 instruction set, which is based on the Linpack package, which has exorbitant energy appetites.

Do not pay attention to the low consumption of the Core i7-4960X: this processor does not support AVX2, and therefore it is not entirely correct to compare it with Haswell microarchitecture carriers in this test. As for the Core i7-4790K and Core i7-5820K, the difference in the appetites of systems based on them is only 20 W, and not more than 50 W, as one would expect based on official specifications. In fact, the Core i7-5820K reveals its gluttony only when overclocked. In the case of operating this processor in the nominal mode, it is quite possible to build a system on its basis that is distinguished by a rather moderate level of power consumption.

⇡ Conclusions

Before the appearance of processors of the Haswell-E generation on the market, low-end CPU models for the high-performance LGA2011 platform looked quite controversial. Their performance was no higher than that of older processors for the current common platform at that time, and in fact, all interest in the quad-core Sandy Bridge-E and Ivy Bridge-E was fueled by marketing alone: Intel positioned such products as special solutions for enthusiasts and the computer elite. communities. However, with the recent update of the high-performance platform and the introduction of the LGA2011-v3 connector, the situation has changed radically. Now the younger Haswell-E is a six-core processor, that is, a processor that is fundamentally different from the older CPUs for the LGA1150 platform, which have a maximum of four cores. Does this make the Core i7-5820K a really interesting and attractive choice for consumers looking to build a productive desktop? Yes and no.

On the one hand, a processor with six processing cores is an excellent tool for solving resource-intensive tasks. Despite the fact that the Core i7-5820K has frequencies at the level of energy-efficient quad-cores for the LGA1150 - and therefore at first glance it seems not fast enough, with a well-parallelized load, it gives a very decent level of performance, surpassing the older Devil's Canyon by 15-20 percent. Thus, in tasks related to the creation and processing of content, systems based on the Core i7-5820K may not only be in demand, but also preferable.

On the other hand, for gaming use in the Core i7-5820K, there is no point. Modern games absolutely do not need more than four cores, and the relatively low clock speeds can put the Core i7-5820K one step below the older LGA1150 processors. Of course, in most situations this does not happen yet, and the power of the junior six-core Core i7-5820K built on the progressive Haswell microarchitecture is enough to load the current flagship video cards. However, there is no guarantee that we will be able to say the same after the release of the next generation of video accelerators. Still, the Core i7-4790K offers several best resources for game load. Therefore, it makes no sense to use the Core i7-5820K at the heart of gaming systems that are not used in the process for systematic work with creative applications.

What was said in the two previous paragraphs could be a comprehensive guide to action when choosing best processor in the price range of 300-400 dollars, if not for one thing. Despite the fact that the cost of the Core i7-5820K and Core i7-4790K differs by only $50, the final price of systems with these CPUs will be noticeably more different. The fact is that the LGA2011-v3 platform sets a high cost of entry by itself: more expensive motherboards are offered for it, and the new DDR4 SDRAM costs more than the usual DDR3. Therefore, in reality, on an LGA2011-v3 configuration with a Core i7-5820K, motherboard mid-range and 16 GB of memory will have to spend $ 150-$ 200 more than a similar system with Core processor i7-4790K. And is it worth it - everyone must decide for himself, based on the purpose for which he is going to use his personal computer.

Review of Intel Core i7-5960X, i7-5930K and i7-5820K | Three New Enthusiast Processors

A little over a decade ago, Intel introduced the Pentium 4 Extreme Edition clocked at 3.4 GHz. This model used a single core with Hyper-Threading technology, 512 KB L2 cache, 2 MB L3 cache and an external bus frequency of 800 MHz. The last characteristic has not been seen for a long time in relevant solutions, Truth? This Pentium was designed on a 130nm process and had 178 million transistors. The processor was sold for $1000, installed in an ancient socket 478, and its thermal threshold reached more than 100 watts.

Who would have thought that in ten years, Intel's flagship chip would have a lower base clock speed, and it would accelerate to 3.5 GHz only in situations where thermal headroom allows. Here comes the new Core i7-5960X. Of course, it must be taken into account that today we are dealing with more complex technologies, and, as we already know, performance can be increased not only by increasing the clock frequency.

Core i7-5960X It carries eight physical cores that can process tasks in sixteen threads at once thanks to Hyper-Threading technology. Applications that are optimized for parallelization of tasks get a significant speedup in this way. Each core has 32 KB of L1 cache for instructions and data, as well as 256 KB of L2 cache. The volume of the L3 cache is as much as 20 MB, that is, 2.5 MB per core.

While the Extreme chip Edition version 2004 was optimal exclusively for computing tasks, many more features are implemented in the 2014 model. Core i7-5960X has its own built-in PCI Express controller, which provides processing of 40 lanes at a speed of 8 GT / s (official PCI Express 3.0 specification). The chip is also equipped with the world's first quad-channel DDR4 memory controller, officially designed for a data transfer rate of 2133 MT / s.

Digging a little deeper, it turns out that Core i7-5960X based on modern Intel Haswell architecture. But since this version is focused on servers and workstations, the architecture in it is called Haswell-E. You get additional PCIe lanes in the solution (there are only sixteen in Haswell desktop CPUs) and the aforementioned memory controller (existing Haswell processors are limited to supporting two DDR3 channels), but you lose the built-in HD Graphics graphics core that was so touted when the fourth generation Core processors appeared .

Additional Information:

If you want to learn more about the Intel Haswell architecture, which is the main processor core on Haswell-E, we suggest reading our review. Core i7-4770K .

Intel reasonably assumed that people buying powerful processor for a workstation or gaming system, discrete graphics cards will be used. Instead of installing an unnecessary GPU that eats up chip space, it was decided to use the freed up resources to create a more powerful central processor.

But even with no graphics core, Haswell-E's die area is over 355 mm², and there are 2.6 billion transistors on the die itself - almost fifteen times more than the Pentium 4 Extreme Edition. The chips are manufactured on the basis of a 22-nanometer process technology and are designed for a thermal package of 140 W. The expected price of this CPU is the same $1000.

Core i7-5930K and Core i7-5820K

Every time we test a thousand dollar Intel processor, we recognize its importance, but understand that enthusiasts would rather spend less money and use their technical knowledge to achieve maximum performance through overclocking. In the case of Haswell-E, the eight-core processor is only Core i7-5960X. In younger models, there are fewer cores and cache.

Fortunately, games generally won't lose performance when moving from eight to six cores, especially when it comes to efficient Intel architecture. However, the FPS increases noticeably with increasing clock speed. As a result, a more suitable candidate for an expensive gaming system based on high-end hardware would be Core i7-5930K. The processor is built on a similar Core i7-5960X crystal. Intel simply disables two cores and 5 MB of L3 cache. There are six cores, 15 MB of shared L3 cache, 40 PCI Express 3.0 lanes, and a quad-channel memory controller. The base frequency jumped to 3.5 GHz, and the peak frequency under the control of Turbo Boost technology - up to 3.7 GHz. Core i7-5930K priced at $583 and could potentially save you over $400.

If it's too expensive, then you should take a closer look at Core i7-5820K for $389. It's also a six-core chip with 15MB of shared L3 cache and a quad-channel DDR4 memory controller. However, Intel is cutting the number of PCI Express lanes from 40 to 28. To be honest, this loss is not as big as it might seem. There are enough lines for one, two and even three video cards, unless, of course, AMD and Nvidia allow arrays in the x8/x8/x8 configuration. According to Intel official data, the processor Core i7-5820K supports line splitting, but splitting must occur at the system board level.

Core i7-5820K slightly loses in clock frequency when compared with Core i7-5930K: Base is 3.3 GHz or up to 3.6 GHz with Turbo Boost.

Turbo Boost clock speeds on Core i7-5000 series processors

Model	Intel Core i7-5960X	Intel Core i7-5930X	Intel Core i7-5820X
Price (recommended), $	1000	583	389
Base frequency, GHz	3	3,5	3,3
One active core, GHz	3,5	3,7	3,6
Two active cores, GHz	3,5	3,7	3,6
Three active cores, GHz	3,3	3,6	3,4
Four active cores, GHz	3,3	3,6	3,4
Five active cores, GHz	3,3	3,6	3,4
Six active cores, GHz	3,3	3,6	3,4
Seven active cores, GHz	3,3	No	No
Eight active cores, GHz	3,3	No	No

Three Chips for Enthusiasts

All three models we tested are either Extreme Edition chips or have a K suffix, meaning they have unlocked multipliers that allow for more free overclocking compared to mainstream Intel Haswell processors.

It's nice that soft solder is used as a material between the Haswell-E die and the large heat spreader covering the Core i7-5000 series CPU. Compare with cheaper Haswell processors that use less efficient thermal paste. In our lab, chips with thermal paste warmed up quickly, thereby limiting voltage rise with both air and liquid cooling. The soldered thermal interface provides more efficient heat transfer, potentially raising the performance ceiling that can be achieved from Haswell-E.

Needless to say, companies selling top-end hardware have high hopes for Core i7-5960X and similar models. We have a large Noctua NH-D15 air cooler and an Intel BXRTS2011LC closed loop liquid cooling system in our lab. G.Skill sent DDR4-3000 memory modules with CAS 15 timings for testing. ASRock and MSI offered several impressive motherboards. Now the first review of motherboards on LGA 2011-3 from major manufacturers is being developed. It's not a typo, a new platform really comes into play.

Review of Intel Core i7-5960X, i7-5930K and i7-5820K | X99, LGA 2011-3 and DDR4: getting ready for a big update

Haswell-E will need a lot of new hardware.

Intel squeezed almost all of its potential out of the LGA 2011. The interface appeared with the Core i7-3960X (Sandy Bridge-E) processors almost three years ago. However, some variables break compatibility, such as the introduction of DDR4 memory.

Old and new processors have the same physical dimensions and mounting holes. However, Core i7-5000 CPUs are fixed differently than 4000 or 3000 series chips, so you won't be able to accidentally install a model for LGA 2011 into the LGA 2011-3 interface.

Special notches on the mount will not allow you to confuse Haswell-E with Ivy Bridge-E or Sandy Bridge-E, even though the number of legs is the same - 2011. For processors Core i7-5960X, Core i7-5930K or Core i7-5820K You need an X99 based motherboard.

But there is also good news. Maintaining dimensions from the previous generation ensures compatibility with older cooling systems. But before using, you need to make sure that your LGA 2011 heatsink or water block can handle the increased thermal limit of Haswell-E. Previous generation Intel flagship models were limited to 130W, the new Core i7s were limited to 140W, and overclocking can significantly increase power consumption figures.

X99 Express: Platform Controller Hub with familiar features

Intel chipsets are developing relatively slowly. The more functions are implemented in the processor itself, the less tasks fall on the PCH. And what remains rarely changes. If you were hoping to see any drastic changes in communication on the X99, then you are likely to be disappointed.

However, X79 is already outdated and X99 at least practically brings the new top-end chipset up to modern standards. It activates 14 USB ports, six of which support the USB 3.0 standard. There is an integrated Gigabit Ethernet network controller. Naturally, HD Audio. Don't forget about eight PCI Express 2.0 lanes for connecting expansion cards through expansion slots or through third-party controllers built into the board. Perhaps the most significant improvement is the introduction of support for 6Gb/s SATA devices.

It's frustrating that Intel still bundles PCH and CPU through four DMI 2.0 lines. Bidirectional throughput is 2 GB/s. It is not difficult to imagine that the combination of traffic peripherals, networks and storage systems will easily fill a fairly narrow channel.

The situation is saved by a large number of PCIe lanes in the two most expensive CPUs for connecting fast graphics adapters, SSD and GbE.

DDR4- new technology memory. What is it for?

Since modern multi-channel memory controllers are built into processors, we rarely hear about bandwidth limitations, which are most often associated with the graphics core integrated into the chip. Ivy Bridge-E supports up to four DDR3 channels at up to 1866 MT/s or more than 40 GB/s of bandwidth.

So why DDR4?

There is no need to switch to a new memory standard in the segment of computing products for enthusiasts. But if we consider Intel processors, which are starting to appear in the server and mobile devices, then the introduction of DDR4 will seem like a completely meaningful step.

For example, the lower supply voltage (1.2V) helps reduce power consumption compared to the 1.5V DDR3 modules that are ubiquitous today. This effect will be difficult to trace in today's article, since the DDR4 modules available in our laboratory are rated at 1.35 V, and in some cases they require even more voltage. In enterprise-class solutions, several Haswell-EP processors will use modules rated at 1.2 V, which provide noticeable power savings.

Factories producing DDR4 today use more advanced technologies to increase the density. This is important for server buyers who need high capacity. random access memory. The amount of memory required in servers is not comparable to the amount of memory typical of Haswell-E based PCs, for which 32 or 64 GB in eight slots will be more than enough.

DDR4 also enables higher data rates, starting at 2133 MT/s. However, this also increases delays. We noticed that Core i7-4960X, based on DDR3-1866 memory, is not far behind Core i7-5930K with DDR4-2133 in the SiSoftware memory test.

During our tests, we noticed that there are still some issues that need to be fixed. X99-based motherboards featured in our lab are constantly receiving new updates, most of which are related to DDR4 compatibility. But some don't work at all. Others are trying to set the speed above 2666 MT/s. On this moment we have to switch from BCLK 100 MHz to 125 MHz or higher. The 2800 and 3000 MT/s options are not very stable. Until issues with firmware, module compatibility, and pricing are resolved, it is DDR4 that may become an obstacle, due to which enthusiasts who are cautious in their choice will prefer to wait a bit.

When it comes to choosing new Intel Haswell-E generation processors, it's hard to talk about any special variety, because there are only three different models available to the buyer. The top i7-5960X, which has 8 cores (16 threads) and two younger models with 6 cores "on board". The top processor, of course, is selected for special specific tasks, if you intend to work with applications that require multithreading, then full 16 threads will help save you a lot of time. For the tasks of an ordinary user, the performance of younger models will be sufficient, especially since the price for them differs significantly downwards. And now a logical question arises - "which of the two junior processors to choose?" In fact, they are not much different from each other. The main advantage of the i7-5930K is the support for 40 PCI-Express lanes, which will be very useful if you use a large number of video cards (three or more). The i7-5820K processor only supports 28 lanes, which means that when using two video cards, the x16 - x8 configuration would be the best scenario, but it is worth noting that the price of the 5820K is lower. Evaluating both of these processors, we cannot unequivocally recommend one of them for you, here the choice will depend solely on the purpose of use, but in this review we will try to give you an idea which processor looks better in which tasks.

Specifications are shown in the table below. The prices are taken from the price list of the CSN store in Nizhny Novgorod for the OEM version of the processor.

As you can see, both processors have 15 MB of L3 cache, while the top 5960X already has 20 MB. The difference between the tested processors is the slightly increased frequency of the i7-5930K, although 200 MHz is not such a big difference, and in Turbo Boost mode the difference becomes 100 MHz.

It should also be noted that the 5820K and 5930K have a higher operating frequency than their older brother 5960X, with a difference of 300 and 500 MHz, respectively. This fact suggests that six-core processors can be slightly faster in ordinary applications and even games.

Extended specifications both processors:

Core: Haswell

Manufacturing process technology: 22 nm

Cache: L1 - 12 x 32 KB, L2 - 6 x 256 KB, L3 - 15 MB (shared)

Memory controller: Quad-channel DDR4, up to 2133 MHz

Socket: LGA2011-v3

Video core: none

Features: MMX, SSE, SSE2, SSE3, SSSE3, SSE4, SSE4.2, AES, AVX, M64T, F16C, Quick Sync Video, Hyper-Threading, Turbo Boost 2.0, VT-x

Testing

Test stand:

Motherboard: Asus x99 Delux

RAM: Corsair Vengeance LPX 2666 MHz, 16 GB

Video card: NVidia GeForce GTX 780

Storage: Crucial MX100, 512 GB

CPU Cooler: Corsair H75

Power supply: Corsair AX860i

Operating system: Windows 7 SP1, x64

PC Mark 8

So, let's start our testing with PCMark 8 in various modes.

The first test simulates the workload of processing 4K video files. First, the video quality is improved, then it is converted to a suitable format, the deshaking filter is applied and the image is displayed on a monitor with a resolution of 1920x1080. The lower the number, the better.

The next test is to process a certain set of photos using the ImageMagik editor. Processing consists of adjusting brightness, contrast, saturation, and color gamut. When one photo is processed, the rest are processed according to the same template automatically. The files used are in TIFF format, the file size is up to 67 MB. The lower the number, the better.

Image processing in Gimp

Gimp is a popular editing application various types images. Our test shows how well a computer can handle large photo libraries. The more points, the better.

Video encoding in Handbrake

This video encoder perfectly illustrates the capabilities of processors in multi-threaded mode. For the test, a video file with high definition and H.264 codec. The higher the scores, the better.

Multithreading in 7-Zip and MPlayer

Almost everyone is familiar with the concept of multithreading nowadays, multithreading allows us to work with several applications at the same time. To implement it, we need a powerful processor (ideally multi-core) and a large amount of RAM. Our multi-threading test performs bulk copying of encrypted files using 7-Zip while playing HD video with MPlayer, which is a pretty tough test for any computer. The more points, the better.

Total score in tests with multimedia files

We calculate the total score in working with multimedia based on the results of the three previous tests. A score of 1000 means that the test system is the same in performance as ours. reference system, which uses an Intel Core 2 Duo E6750 at stock frequency, 2 GB of Corsair DDR2 1066 MHz RAM, HDD Samsung SpinPoint P120S 250 GB motherboard ASUS board P5K Deluxe WiFi-AP. The calculation is linear, that is, if the system scores 1200 points, then it is 20% faster than the reference one. The more points, the better.

Cinebench R11.5 and R15 (64 bit)

Cinebench uses the rather advanced CINEMA 4D platform to realize very complex and spectacular scenes. This test is very difficult for any processor. We also note that CINEMA 4D is a real-life tool, with the help of which effects were implemented in films such as Spider-Man and Star Wars, so this test can be considered as a very real benchmark. The more points, the better.

Terragen 3

This application is an advanced scenery generator, which, again, is used to create effects in various films. The lower the number, the better.

Testing in games

Recall that the NVidia GeForce GTX 780 video card works in tandem with the tested processors, the monitor screen resolution is 1920x1080 pixels. For testing, the Fraps utility is used, for the reliability of the results, 3 measurements are taken and the average value is taken.

Battlefield 4

BioShock Infinite

Crysis 3

Power consumption

When testing processors, we completely disabled all power-saving technologies. We did this so that all test results were consistent, that is, all processors were in the same conditions. Also, we must not forget that any energy-saving programs negatively affect the maximum system performance. But for the power consumption test, on the contrary, we activated all these functions.

Testing consisted of two stages - measuring consumption in the idle state and in the state of intense load. Measurements were taken with an external device, so the graphs below show the total consumption of the system. For the rest test, we simply loaded Windows and left the desktop open, no other applications were running.

For the load test, we used the Prime95 stress test. Aero has also been launched. Then we waited for several minutes when the load stabilized and the power consumption became constant.

Performance analysis

Well, we have conducted a series of tests, compared the performance of new processors with processors released earlier, and we can draw some conclusions. As before, the higher the clock frequency, the more chances for the processor to lead, we see this from the graphs and this applies, first of all, to the Intel Core i7-4790K with its solid 4 GHz, and this despite the fact that it has only 4 cores. But the frequency of the processor is far from important in all cases, for example, in tests related to photo editing, the advantage of new processors with six cores is already quite obvious, since work with multi-threaded systems is more developed in various photo editors. In the popular Cinebench tests, the Haswell-E processors were also significantly faster, with the 5930K beating the 4970K by 26% and the 5820K being 22% faster. After overclocking, the advantage even approached 40%. Tests in games have shown that almost all the latest processors are capable of handling any modern game, so if you are building a top gaming system, you should not chase the best stone and buy an i7-5960X, as younger processors show similar results. The saved money is better spent on a more powerful video card, for games it will be much more useful.

According to the results of our tests, the undisputed leader, of course, is the eight-core monster i7-5960X, but the younger models 5930K and 5820K are not far behind it, only i7-4970K can compete with the new line of processors from older models. It is also worth noting the i7-4960X, if we consider a computer as a tool for working with video, then this processor may well serve as an alternative to the new Haswell-E.

Conclusion

Summing up all of the above, I would like to note that the new processors from Intel turned out to be quite productive, and in some tests involving multithreading, the results were even higher than we expected. As for the choice between the i7-5930K and i7-5820K, everything is quite simple, both of these processors demonstrate very high performance, and you will have to choose only based on how many video cards will be used in your system, if one, then those The 28 PCI-E lanes offered by the i7-5820K, and if two or three or more, the i7-5930K would be preferable as it offers 40 PCI-E lanes. It is also impossible not to mention the alternative to these processors. When choosing a stone for a high-performance system, pay attention to the i7-4790K, this processor has a relatively low cost, but it can compete with the tops in many applications and games.

Sadly, but the endless battles of fans of AMD and Intel processors are a thing of the past. Today, answering the question of which company's processor is more profitable to purchase for a productive desktop system has become very easy. AMD has practically given up competing with Intel in the upper price segments, and therefore, starting at about the $150 threshold, there really is no choice. Any reasonably priced latest-generation Core family processor will become the most the best option, and not because of some of its undeniable merits, but corny because of the lack of worthy alternatives. However, among all this simplicity, it is still possible to find a sufficient reason for discussion.

The fact is that Intel offers not one, but several platforms at once for desktop computers, which may be somewhat similar in their scope. And it's not at all about the fact that the market at the same time there are previous and following platforms that differ in their characteristics and performance is not so noticeable. Everything is simple here, because if you compare, for example, LGA1155 and LGA1150, then in any case, the choice should be made in favor of the second, newer version - more promising and, albeit a little, but still more productive. Intel itself is pushing us to this, shaping its pricing policy in such a way that new platforms, the release of which is synchronized with full cycle in the tick-tock strategy, they cost no more than their predecessors. In other words, paying attention to obsolete products when they have a more modern replacement does not make any rational sense.

The logical reason for the agony of choice arises from the fact that for the most productive desktops, Intel offers not the usual commonly used LGA1150 processors of the Haswell generation, but specialized elite CPUs with Haswell-E design in the LGA2011-v3 quasi-server form factor. Formally, the LGA1150 and LGA2011-v3 platforms do not intersect in terms of positioning, as Intel tried to separate them into different price segments. But in fact, the difference in the cost of the older processor for LGA1150 (Core i7-4790K) and the younger processor for LGA2011-v3 (Core i7-5820K) is not so fundamental - it does not exceed 15 percent. Of course, one should also take into account the fact that the accompanying ecosystem for the LGA2011-v3 processor itself will cost a little more, however, this price difference may not be decisive for many. Therefore, here and there you can hear a completely reasonable question about which version of the system is better to prefer for certain purposes - many users are ready to seriously consider the Core i7-5820K in LGA2011-v3 as an alternative to the Core i7-4790K. And if so, we decided to devote a separate study to this issue.

⇡ Learn more about Core i7-5820K

Intel's path to launching exactly the same as the Core i7-5820K, a junior processor for a high-performance platform, was not easy. Its first predecessor from the Sandy Bridge-E family that appeared in early 2012, the Core i7-3820, was a quad-core that did not even have an unlocked multiplier. In the next generation, Ivy Bridge-E, the junior Core i7-4820K processor received overclocking capabilities, but still, like the older Core i7 processors in the LGA1155 version, it was content with only four cores. The most significant step forward was made only with the introduction of the LGA2011-v3 platform and the Haswell-E processor design, as a result of which the Core i7-5820K became a full-fledged six-core CPU with a full set of overclocking features. Thus, today's junior processor for Intel's high-performance desktop platform can finally boast of fundamentally the best performance than the older processor for a common platform, if only because it has one and a half times more computing cores.

That Core i7-5820K is interesting. In relation to older processors for the LGA1150, it is a representative of a different weight category, but its recommended price is set at only $389, while the Core i7-4790K is estimated by the manufacturer only $50 cheaper - at $339. Quite a small markup for two extra cores, isn't it? But in addition, the Core i7-5820K boasts a more capacious third-level cache, the volume of which reaches 15 MB.

True, the presence of additional nuclei also has back side- lower clock speeds. The passport frequency of the younger Haswell-E is set at a rather modest level - 3.3 GHz, which is 700 MHz less than the core i7-4790K clock frequency. Of course, at the same time, Turbo Boost technology is implemented in the Core i7-5820K, but it can only increase the frequency up to 3.6 GHz, while the older Devil's Canyon accelerates to 4.4 GHz when the turbo mode is activated.

Since the Core i7-5820K belongs to the LGA2011-v3 platform, it is easy to guess about its other fundamental differences from the LGA1150 processors. Firstly, the junior representative of the Haswell-E family quite naturally has a four-channel memory controller that supports DDR4 SDRAM. However, to call it a clear advantage would still be wrong. On this stage DDR4 memory is rather poorly distributed, and it does not give a tangible performance boost. The extra charge that will have to be paid for the novelty of such a memory will be 50-60 percent at best.

Secondly, the Core i7-5820K has a more advanced PCI Express 3.0 controller than conventional Haswell, which supports 28 lanes instead of 16. This feature may be useful for gamers using multi-GPU configurations, or for enthusiasts who want to create a high-performance disk subsystem based on RAID controllers or solid state drives server class with PCI interface express. Moreover, only LGA2011-v3 processors (including the Core i7-5820K) can provide three PCI slots Express 3.0 at the same time, which allows you to build three-component SLI- or CrossfireX-configurations in the system based on them. However, it should be borne in mind that processors for LGA2011-v3, which are more expensive than the Core i7-5820K, provide the user with an even greater number of PCI Express lines - 40. But for the vast majority of cases, 28 lines available in the younger Haswell-E will be enough . For example, the difference in the speed of a two-component multi-GPU system with a full-fledged scheme of PCI Express 16x+16x slots and with the PCI Express 16x+8x variant provided by the Core i7-5820K is almost imperceptible.

Speaking about what other differences there are in Haswell-E and Haswell, one cannot fail to mention the absence of an integrated graphics core in high-performance processors for LGA2011-v3. Of course, this will not upset users of upper price range systems at all, but there is one subtlety. The lack of an integrated GPU also means that there is no Quick Sync feature, which could also be useful for the Core i7-5820K, as it allows for very fast and low-cost video transcoding.

Nevertheless, if you sum up all that has been said, then the Core i7-5820K looks like a very attractive offer. This processor is only slightly inferior in its characteristics to the $ 600 Core i7-5930K, which means that you can expect from it about the same performance as the recent flagship, the Core i7-4960X Extreme Edition of the Ivy Brige-E generation. But now that eight-core processors are being offered to enthusiasts for $1,000, the price of such high processing power has dropped significantly - six-core processors have become much closer to the people.

Let's brush up on the specs of the latest unlocked Intel processors in the Core i7 series:

	Core i7-5960X	Core i7-5930K	Core i7-5820K	Core i7-4790K
codename	Haswell-E	Haswell-E	Haswell-E	Devil's Canyon
Cores/Threads	8/16	6/12	6/12	4/8
Hyper Threading Technology	Eat	Eat	Eat	Eat
Clock frequency	3.0 GHz	3.5 GHz	3.3 GHz	4.0 GHz
Maximum frequency in turbo mode	3.5 GHz	3.7 GHz	3.6 GHz	4.4 GHz
Unlocked multiplier	Eat	Eat	Eat	Eat
TDP	140 W	140 W	140 W	88 W
PCI Express 3.0 lanes	40	40	28	16
HD Graphics	No	No	No	HD Graphics 4600
L3 cache	20 MB	15 MB	15 MB	8 MB
Memory support	4 channels DDR4-2133	4 channels DDR4-2133	4 channels DDR4-2133	2 channels DDR3-1600
Instruction set extensions	AVX2	AVX2	AVX2	AVX2
Package	LGA2011-v3	LGA2011-v3	LGA2011-v3	LGA1150
Price	$999	$583	$389	$339

As follows from the above table, the operating frequencies of the Core i7-5820K range from 3.3 to 3.6 GHz. But according to the CPU-Z diagnostic utility, the actual frequency of the Core i7-5820K under heavy load is almost always 3.4 GHz.

If the load falls on one or two computing cores, then this frequency can increase up to 3.6 GHz.

The operating voltages of the Core i7-5820K, like those of the older Haswell-E processors, are low: about 1.03-1.08 V. The uncore part of the processor, which includes, among other things, the L3 cache and the memory controller, in the nominal mode operates at a frequency of 3.0 GHz. This frequency is the same for all representatives of the Haswell-E series.

Curiously, the Core i7-5820K, like the other two Haswell-E models, is based on an eight-core semiconductor chip with an area of 356 mm 2 . However, a couple of cores, along with their corresponding part of the cache memory, is disabled at the production stage, and in fact we are talking about the fact that six-core processors use a rejection from the production of eight-core processors. This fact just explains the very similar behavior of all Haswell-E processors, which, for example, manifests itself when they are overclocked.

⇡ Overclocking

Despite the fact that the Core i7-5820K is the youngest representative in the Haswell-E line, it has all the same overclocking capabilities as its more expensive brothers. That is, firstly, its multiplier is unlocked, and secondly, the multipliers responsible for the formation of the frequency of the memory and the Uncore block are not fixed either. In addition, the processor allows you to choose between three options for the base frequency - 100, 125 or 166 MHz, for which the frequency dividers of the DMI and PCI Express buses are optimized. As for the supply voltage of the computing cores and all adjacent nodes, it is formed by a stabilizer built into the processor.

Like its older counterparts, the Core i7-5820K also has its case: a heat-dissipating cover is mounted on a semiconductor chip by soldering, which is considered the best option in terms of heat dissipation efficiency. And in this case, this can be said with complete confidence: Japanese enthusiasts tried to scalp a copy of such a processor and not only made sure that there was solder under the cover, but also captured what they saw on memorable photographs.

However, despite all this, we cannot say that the Core i7-5820K was able to please us with the results of its overclocking. The fact that it is based on exactly the same semiconductor crystal as in the older Haswell-E led to its similar behavior when increasing the frequency and voltage beyond their nominal values. Unfortunately, the excessive heat we experienced when testing the Core i7-5960X and Core i7-5930K again made it impossible to achieve any outstanding results. And one of the best air coolers we use for heat dissipation - the two-section Noctua NH-D15 tower - turned out to be powerless in front of six cores with Haswell microarchitecture even with only 25% overclocking.

As a result, stable operation of the Core i7-5820K turned out to be possible only at a frequency of 4.1 GHz and with an increase in the supply voltage to 1.225 V. Further overclocking with this supply voltage led to a loss of stability, and an increase in voltage was impossible due to the resulting overheating processor. During the passage of stability tests with a frequency of 4.1 GHz, the temperature of our Core i7-5820K instance reached 95 degrees, which can be considered quite acceptable for Haswell-E temperature regime, since throttling for these CPUs turns on when heated to 105 degrees.

Please note that we use the LinX 0.6.5 utility with support for AVX2 instructions to check overclocking - it is this program that is best suited for checking overclocker LGA2011-v3 systems. The fact is that AVX2 instructions are still not widely used in commonly used programs, but they cause unbridled heating of processor cores with Haswell microarchitecture. And this means that if the Core i7-5820K processor retains its stability in LinX 0.6.5, then problems in other cases will almost certainly not arise.

Unfortunately, the overclocking potential of the Core i7-5820K turned out to be no better than that of the Core i7-5960X and Core i7-5930K processors that visited our laboratory before. All this unequivocally indicates that all Haswell-Es are one field of berries. And the fact that six-core representatives of the family use crystals that for some reason did not pass the selection for full-fledged eight-core modifications well explains the low overclocking of the younger Core i7-5820K. In other words, the 25% increase in clock frequency that we obtained in the course of overclocking experiments is not an unfortunate exception to the rule, but a completely natural result that reflects the general trend. In terms of the maximum clock frequencies achievable under normal conditions, the younger six-core for the LGA2011-v3 platform is seriously inferior to the older Devil's Canyon, even though they are based on the same Haswell microarchitecture.

⇡ Description of test systems and testing methodology

As a result, the list of hardware components involved in testing looked like this:

Processors:
- Intel Core i7-5960X Extreme Edition (Haswell-E, 8 cores + HT, 3.0-3.5 GHz, 20 MB L3);
- Intel Core i7-5930K (Haswell-E, 6 cores + HT, 3.5-3.7 GHz, 15 MB L3);
- Intel Core i7-5820K (Haswell-E, 6 cores + HT, 3.3-3.6 GHz, 15 MB L3);
- Intel Core i7-4960X Extreme Edition (Ivy Bridge-E, 6 cores + HT, 3.6-4.0 GHz, 15 MB L3);
- Intel Core i7-4790K (Haswell Refresh, 4 cores + HT, 4.0-4.4 GHz, 8 MB L3).
CPU cooler: Noctua NH-D15.
Motherboards:
- ASUS X99-Deluxe (LGA2011-v3, Intel X99);
- ASUS Z97-Pro (LGA1150, Intel Z97);
- Gigabyte X79-UP4 (LGA2011, Intel X79).
Memory:
- 2x8 GB DDR3-2133 SDRAM, 9-11-11-31 (G.Skill F3-2133C9D-16GTX).
- 4x4 GB DDR3-2133 SDRAM, 9-11-11-31 (G.Skill F3-2133C9Q-16GTX);
- 4x4 GB DDR4-2666 SDRAM, 15-15-15-35 (G.Skill F4-2666C15Q-16GRR);
Video Card: NVIDIA GeForce GTX 980 (4 GB/256-bit GDDR5, 1127-1216/7012 MHz).
Disk subsystem: Crucial M550 512 GB (CT512M550SSD1).
Power supply: Seasonic Platinum SS-760XP2 (80 Plus Platinum, 760 W).

Testing was performed on the Microsoft Windows 8.1 Professional x64 with Update operating system using the following set of drivers:

Intel Chipset Driver 10.0.17;
Intel Management Engine Driver 10.0.0.1204;
Intel Rapid Storage Technology 13.2.4.1000;
NVIDIA GeForce 344.75 Driver.

Core i7-5820K @ 4.1GHz @ 1.225V;
Core i7-4790K overclocked to 4.5GHz at 1.2V.

Description of the tools used to measure performance:

Benchmarks:
- Futuremark PCMark 8 Professional Edition 2.3.293 - testing in Home (typical home PC use), Creative (PC use for entertainment and multimedia content) and Work (PC use for typical office work) scenarios.
- Futuremark 3DMark Professional Edition 1.4.828 - testing in Sky Driver, Cloud Gate and Fire Strike scenes.
Applications:
- Adobe Photoshop CC 2014 - performance testing for graphics processing. Measured is the average execution time of a test script, which is a creatively redesigned Retouch Artists Photoshop Speed Test, which includes a typical processing of four 24-megapixel images taken by a digital camera.
- Adobe Photoshop Lightroom 5.7 - performance testing for batch processing of a series of images in RAW format. The test scenario includes post-processing and export to JPEG at 1920x1080 resolution and maximum quality of two hundred 12-megapixel RAW images taken with a Nikon D300 digital camera.
- Adobe Premiere Pro CC 2014 - performance testing for non-linear video editing. Measures rendering time to H.264 Blu-ray for a project containing HDV 1080p25 footage with various effects applied.
- Autodesk 3ds max 2015 final rendering speed test. The time taken to render at 1920x1080 resolution using the mental ray renderer of one frame of the standard Space_Flyby scene from the SPEC test package is measured.
- WinRAR 5.1 - archiving speed testing. The time taken by the archiver to compress a directory with various files with a total volume of 1.7 GB is measured. The maximum compression ratio is used.
- x264 r2491 - testing the speed of video transcoding to H.264/AVC format. To evaluate the performance, the original [email protected] AVC video file with a bit rate of about 30 Mbps.
- X265 1.4+142 8bpp - testing the speed of video transcoding to the promising H.265/HEVC format. For performance evaluation, the same video file is used as in the x264 encoder transcoding speed test.
Games:
- Civilization: Beyond Earth. Settings for 1280x800 resolution: DirectX11, Ultra Quality, Anti-aliasing = Off, Multithreaded rendering = On. Settings for 1920x1080 resolution: DirectX11, Ultra Quality, 8x MSAA, Multithreaded rendering = On.
- Company of Heroes 2. Settings for 1280x800 resolution: Maximum Image Quality, Anti-Aliasing = Off, Higher Texture Detail, High Snow Detail, Physics = Off. Settings for 1920x1080 resolution: Maximum Image Quality, High Anti-Aliasing, Higher Texture Detail, High Snow Detail, Physics = High.
- F1 2014. Settings for 1280x800 resolution: Ultra Quality, 0xAA, DirectX11. Settings for 1920x1080 resolution: Ultra Quality, 8xAA, DirectX11. The Texas track is used.
- Hitman: Absolution. Settings for 1280x800 resolution: Ultra Quality, MSAA = Off, High Texture Quality, 16x Texture Aniso, Ultra Shadows, High SSAO, Global Illumination = On, High Reflections, FXAA = On, Ultra Level of Detail, High Depth of Field, Tesselation = On, Normal Bloom. Settings for 1920x1080 resolution: Ultra Quality, 8x MSAA, High Texture Quality, 16x Texture Aniso, Ultra Shadows, High SSAO, Global Illumination = On, High Reflections, FXAA = On, Ultra Level of Detail, High Depth of Field, Tesselation = On , Normal Bloom.
- Metro: Last Light Redux. Settings for 1280x800 resolution: DirectX 11, High Quality, Texture Filtering = AF 16X, Motion Blur = Normal, SSAA = Off, Tessellation = High, Advanced PhysX = Off. Settings for 1920x1080 resolution: DirectX 11, Very High Quality, Texture Filtering = AF 16X, Motion Blur = Normal, SSAA = On, Tessellation = High, Advanced PhysX = Off. When testing, Scene 1 is used.
- Middle Earth: Shadow of Mordor. Settings for 1280x800 resolution: Lighting Quality = High, Mesh Quality = Ultra, Motion Blur = Camera and Objects, Shadow Quality = High, Texture Filtering = Ultra, Texture Quality = High, Ambient Occlusion = Medium, Vegetation Range = Ultra, Depth of Field = On, Order Independent Transparency = On, Tessellation = On. Settings for 1280x800 resolution: Lighting Quality = High, Mesh Quality = Ultra, Motion Blur = Camera and Objects, Shadow Quality = Ultra, Texture Filtering = Ultra, Texture Quality = Ultra, Ambient Occlusion = High, Vegetation Range = Ultra, Depth of Field = On, Order Independent Transparency = On, Tessellation = On.
- Thief. Settings for 1280x800 resolution: Texture Quality = Very High, Shadow Quality = Very High, Depth-of-field Quality = High, Texture Filtering Quality = 8x Anisotropic, SSAA = Off, Screenspace Reflections = On, Parallax Occlusion Mapping = On, FXAA = Off, Contact Hardening Shadows = On, Tessellation = On, Image-based Reflection = On. Settings for 1920x1080 resolution: Texture Quality = Very High, Shadow Quality = Very High, Depth-of-field Quality = High, Texture Filtering Quality = 8x Anisotropic, SSAA = High, Screenspace Reflections = On, Parallax Occlusion Mapping = On, FXAA = On, Contact Hardening Shadows = On, Tessellation = On, Image-based Reflection = On.

⇡ Performance in complex tests

⇡ Application performance

In passing, we note that, despite our complaints about overclocking, which we were able to achieve with the Core i7-5820K, the performance gain was not so small. Increasing the frequency of this processor to 4.1 GHz allowed for an average 15 percent increase in the speed of resource-intensive tasks. And this is clearly more than what you can count on when overclocking the Devil's Canyon, which is accelerated almost to the maximum by the manufacturer initially. It is curious that the overclocked Core i7-5820K in some cases boasts even higher performance than the current LGA2011-v3 flagship, the Core i7-5960X. However, do not forget that this is due to the low nominal frequency of the eight-core model, which, like the Core i7-5820K, is subject to overclocking to about the same levels. That is, if you do not look back at budget issues, then to get maximum performance in “heavy” tasks, you should still rely on the older eight-core Haswell-E.

⇡ Game performance

Testing in real games is preceded by the results of the 3DMark synthetic benchmark, which gives a certain average metric of 3D gaming system performance.

Tests in Full HD resolution:

As you can see from the above diagrams, the gaming performance of the Core i7-5820K and Core i7-4790K is almost the same. When set to Full HD resolution, both processors cope with the full load of the flagship GeForce GTX 980 graphics card, and we do not see any fundamental differences in the number of frames per second. If we approach the numbers on the diagram quite meticulously, then the Core i7-4790K is still a little faster, but this slight superiority is actually due not even to the higher clock speed of the quad-core, but to the features of the LGA2011-v3 platform, the PCI Express controller in which works with slightly higher latency due to its complexity.

⇡ Energy consumption

When idle, the LGA2011-v3 platform consumes noticeably more than the LGA1150. We already know this, and there are several reasons for this phenomenon. Firstly, the Intel X99 system logic set is one and a half times more voracious than the Z97. Secondly, processors of the Haswell-E family are deprived of support for some energy-saving technologies, in particular, the C7 state. Thirdly, motherboards with an LGA2011-v3 processor socket have a deliberately more complex design and are equipped with a large number of controllers.

⇡ Conclusions

On the other hand, for gaming use in the Core i7-5820K, there is no point. Modern games absolutely do not need more than four cores, and the relatively low clock speeds can put the Core i7-5820K one step below the older LGA1150 processors. Of course, in most situations this does not happen yet, and the power of the junior six-core Core i7-5820K built on the progressive Haswell microarchitecture is enough to load the current flagship video cards. However, there is no guarantee that we will be able to say the same after the release of the next generation of video accelerators. Still, the Core i7-4790K offers slightly better resources for a gaming load. Therefore, it makes no sense to use the Core i7-5820K at the heart of gaming systems that are not used in the process for systematic work with creative applications.

What has been said in the two previous paragraphs could be an exhaustive guide to action when choosing the best processor in the $300-400 price range, if not for one thing. Despite the fact that the cost of the Core i7-5820K and Core i7-4790K differs by only $50, the final price of systems with these CPUs will be noticeably more different. The fact is that the LGA2011-v3 platform sets a high cost of entry by itself: more expensive motherboards are offered for it, and the new DDR4 SDRAM costs more than the usual DDR3. Therefore, in reality, an LGA2011-v3 configuration with a Core i7-5820K, a mid-range motherboard and 16 GB of memory will cost $150-$200 more than a similar system with a Core i7-4790K processor. And is it worth it - everyone must decide for himself, based on the purpose for which he is going to use his personal computer.

The end of summer - the beginning of autumn has always been a hot time in the IT industry, since it is during this period that a large proportion of various announcements and events fall. AMD has already managed to please gamers with an addition to the family of AMD Volcanic Islands video cards - the AMD Radeon R9 285 solution based on the graphics AMD core Tonga Pro. Its eternal rival NVIDIA is preparing a response in the form of launching a new series of NVIDIA GeForce GTX 9xx video accelerators. Although the processor market cannot boast of the same activity, however, the beginning of autumn here was also marked by a high-profile event - the release of the platform SocketLGA2011- v3 and a new generation of high performance processors Intel Haswell-E.

Despite the fact that the top solutions from the Intel Sandy Bridge-E and Intel Ivy-Bridge-E families still have no equal in terms of performance, they are already obsolete. The same can be said about the entire Socket LGA2011 platform. The Intel X79 Express chipset has long fallen short of the modern requirements of a high-performance system. To some extent, motherboard manufacturers tried to correct the situation by soldering third-party controllers for SATA 6 Gb / s and USB 3.0 ports. However, it could not go on for so long, because a new platform and a new microarchitecture were required. Thus, the appearance of a new family of high-performance processors became almost a fait accompli, the only question left was: “When?”

August of this year was the answer. Perhaps, it will seem to someone that Intel has delayed too much with the announcement of the Intel Haswell-E line. However, there are quite objective reasons for this: it was necessary to "run in" the new Intel Haswell microarchitecture in practice, as well as to enable memory manufacturers to saturate the market with DDR4 standard solutions.

The main innovation in the Intel Haswell-E family was the introduction of an 8-core device in the model range, while previously top-end Intel desktop processors included a maximum of 6 cores. The second improvement concerned the memory controller: there was a transition from the DDR3 standard to DDR4 with a much greater throughput. According to the company, all this allows for multi-threaded applications to achieve a gain of 79% compared to 4-core solutions. It's a pity that today software, which takes full advantage of the principles of multi-core and multi-threading, not so much. Therefore, in real conditions, the difference in performance may not be so impressive. But this is more a question for application developers, because, for their part, processor manufacturers have already done everything possible to increase the speed of the PC. But we will consider the real performance of new products a little later during testing, but now let's focus in more detail on the theoretical part of the issue.

Microarchitecture Features

The processors from the Intel Haswell-E family are based on an 8-core semiconductor crystal. Apparently, in less powerful models (Intel Core i7-5930K and Intel Core i7-5820K), which include 6 cores, the remaining 2 will simply be blocked. Following its "Tick-Tock" strategy, Intel used a well-established 22nm process to manufacture Intel Haswell-E solutions, but at the same time changed the microarchitecture from Intel Ivy Bridge to Intel Haswell.

Intel Haswell- E

Processor crystal from the lineIntel Ivy Bridge- E

As a result, all the structural level improvements implemented in Intel Haswell processors (as well as their successors: Intel Haswell Refresh and Intel Devil's Canyon) migrated here: improved sampling and branch prediction mechanisms, support for AVX2 and FMA3 instructions, increased cache performance the first and second levels, the presence of a voltage converter on the side of the CPU.

As we noted earlier, the maximum number of cores has increased: 8 for the Intel Core i7-5960X Extreme Edition versus 6 for the previous flagship. Taking into account Intel technologies Hyper-Threading, at the output we get 16 independent threads instead of 12. The increase in the number of cores also affected the amount of cache memory in the third level (L3). Now its volume can reach as much as 20 MB.

As for the memory controller, in addition to the transition to the DDR4 standard, the frequency of supported modules has also increased. The manufacturer guarantees stable operation of memory sticks at speeds up to 2133 MHz. For representatives of the Intel Ivy Bridge-E generation, this figure was limited to 1866 MHz. Although, it is obvious that by overclocking it will be possible to achieve even better results. At least, some memory manufacturers are hinting at this, already offering sets of modules capable of operating at frequencies of 3000 and 3333 MHz. The configuration of the controller itself remains the same: 4 channels with a maximum memory capacity of up to 64 GB.

Naturally, such indicators also affected the complexity of the processor itself. So the 8-core modification of Intel Haswell-E includes 2.6 billion transistors, while the flagship of the Intel Ivy Bridge-E family has only 1.86 billion. The crystal area also increased: from 257 mm 2 to 356 mm 2 . In order not to further increase its dimensions, the manufacturer abandoned the integrated graphics core. Indeed, why is it in a processor costing $400 - 1000?

The only node that has not changed is the PCI Express line distribution block. As before, their number reaches 40 pieces (x16+x16+x8 or x8+x8+x8+x8+x8). All of them comply with the PCI Express 3.0 standard. Thus, Intel Haswell-E solutions are the best suited for creating powerful gaming configurations with multiple graphics cards.

Platform Socket LGA 2011-v3

The introduction of a new microarchitecture and expansion of capabilities led to a change in platform. The processor is still the main link and includes the main controllers. The chipset is responsible for ensuring the correct operation of peripherals: sound and disk subsystems, network controllers, USB connectors, and PCI Express 2.0 expansion slots. Data exchange between them is carried out via the DMI 2.0 bus.

More detailed functionality of the new Intel chipset X99 and motherboards based on it, we will explore in a separate review. Here we just note that the capabilities of the Socket LGA2011-v3 platform have increased in almost all parameters compared to the Socket LGA2011.

As for the connector itself, despite the general similarity in design, fasteners, and even the same number of pins, it is not compatible with processors. Intel Sandy Bridge-E and Intel Ivy-Bridge-E.

modelrow

The Intel Haswell-E processor line is represented by three models: Intel Core i7-5960X Extreme Edition, Intel Core i7-5930K and Intel Core i7-5820K. On the market, they will replace the corresponding solutions from the Intel Ivy-Bridge-E family: Intel Core i7-4960X Extreme Edition, Intel Core i7-4930K and Intel Core i7-4820K.

As you can see from the table, only the flagship of the series is equipped with 8 cores. The rest of the processors have 6 cores, including the younger one. intel model Core i7-5820K. Apparently, Intel took into account user reviews regarding the 4-core Intel Core i7-4820K, which practically does not differ in performance from the Intel Core i7-3770K. Now, the difference between the top "civilian" processor Intel Core i7-4790K and the junior representative of high-performance solutions (Intel Core i7-5820K) will be more noticeable.

The increase in the number of transistors also led to an increase in heat dissipation: from 130 W to 140 W. Moreover, in order not to go beyond this indicator, the manufacturer in some cases had to slightly lower the operating frequency. This is especially noticeable on the example of the flagship of the Intel Core i7-5960X Extreme Edition series, which operates at a nominal speed of 3000 MHz, and accelerates to 3500 MHz in turbo mode. Recall that its predecessor, Intel Core i7-4960X Extreme Edition, the frequency was 500 - 600 MHz higher, depending on the operating mode. However, all new products have an unlocked multiplier, so if necessary, additional speed can always be obtained by overclocking. And the presence of a high-quality thermal interface between the crystal and the processor cover will also have a positive effect on their overclocking potential. True, you will have to take care of the cooling system yourself, since high-performance solutions from Intel are not equipped with a standard cooler. Although, this is even for the better, since he would have been immediately replaced anyway.

To save the user from searching for a cooling system with sufficient efficiency, the manufacturer offers to purchase a proprietary maintenance-free type CBO called "Intel TS13X" for a separate amount. However, looking at its characteristics, it is still better to think about buying a good air cooler, like the Noctua NH-D15.

And finally, the most important thing is the cost of new products. Intel has remained true to itself here too, setting the price tag for the flagship of the series at $1000. The 6-core Intel Core i7-5930K and Intel Core i7-5820K are much cheaper - $583 and $389 respectively. It is these models that are likely to make up the lion's share of sales, and the Intel Core i7-5960X Extreme Edition will be the lot of overclockers, game fans and just lovers of very powerful configurations. No matter what ill-wishers say there, there are also a sufficient number of such buyers. It remains only to be glad for the users who are lucky enough to become the owners of such an 8-core "monster".

CPUIntelCorei7-5930 K

Unfortunately, we did not manage to get the flagship of the series for testing, so we will start our acquaintance with the Intel Haswell-E line with a 6-core model, namely with IntelCorei7-5930 K. For greater convenience, we decided to submit the full specification of the novelty in the form of a comparative table with its predecessor Intel Core i7-4930K.

		IntelCorei7-5930 K	Intel Core i7-4930K
Marking (engineering sample)		SR20R(QFRB)
Processor socket		SocketLGA2011-v3
Clock frequency, MHz	nominal
	in turbo mode
Factor	nominal
	in turbo mode
Base frequency, MHz
L1 cache size, KB		6 x 32 (instruction memory) 6 x 32 (data memory)	6 x 32 (instruction memory) 6 x 32 (data memory)
L2 cache size, KB		6 x256
L3 cache size, MB
microarchitecture		Intel Haswell	Intel Ivy Bridge
codename		Intel Haswell-E	Intel Ivy Bridge-E
Number of processor cores / threads
Instruction Support		MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, EM64T, VT-x, AES, AVX, AVX2, FMA3	MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, EM64T, VT-x, AES, AVX
Number of PCI Express 3.0 lanes
Supply voltage, V
Maximum design power (TDP), W
Critical temperature, °C
Maximum operating temperature T case , °C
Process technology, nm
Technology support		Enhanced Halt State (C1E) Enhanced Intel Speedstep Hyper Threading Execute Disable Bit Intel Virtualization Intel Turbo Boost 2.0	Enhanced Halt State (C1E) Enhanced Intel Speedstep Execute Disable Bit Intel Virtualization Intel Turbo Boost 2.0
Built-in memory controller
Maximum memory, GB
Memory types
Maximum frequency, MHz
Memory subsystem bandwidth, GB/s
Number of memory channels
Maximum number of modules per channel
Built-in graphics core

CPUIntel Core i7-5930 K

CPUIntel Core i7-4930K

Compared to Intel's Ivy Bridge-E series solutions, the physical dimensions of the new processors remain the same. However, this changed the configuration of the heat-distributing cover and the arrangement of contacts on the back side. In addition, improvements have also occurred at the electrical level, which are primarily associated with the appearance of an internal power converter. This simplifies the design of the VRM module on the motherboard, since only one base voltage is now supplied to Intel Haswell-E processors. This technology It proved itself very well on the Socket LGA1150 platform, so its appearance in the older series is fully justified.

As we noted earlier, as a result of all the modifications carried out, the Socket LGA2011-v3 connector has lost forward and backward compatibility with the Socket LGA2011. However, the mounting system for the coolers has remained the same. In other words, when changing platforms, you will not need to buy new system cooling. The only thing worth remembering is that the TDP of new products has increased by 10 W compared to solutions from the Intel Ivy Bridge-E series.

Analysis of technical characteristics

In normal operation, the speed of the Intel Core i7-5930K is 3500 MHz with a reference frequency of 100 MHz and a multiplier of "x35". At the time of taking the readings, the core voltage was 1.037 V. Thus, the hero of the review is 100 MHz faster than its predecessor (Intel Core i7-4930K) and at the same time requires a slightly lower supply voltage for its correct operation: 1.037 V versus 1.080 V.

In dynamic boost mode using proprietary technology Turbo Boost 2.0 multiplier increases by 2 points to the value of "x37". At the same time, the processor clock frequency increases to 3700 MHz, and the voltage to 1.071 V. However, such a speed was observed very rarely and only with a small load.

Most of the time, the processor spends at a frequency of 3600 MHz, and the supply voltage varies between 1.056 - 1.071 V. As you can see, the mechanism of the Intel Turbo Boost technology has also undergone some changes. Recall that the Intel Core i7-4930K in turbo mode was 200 MHz faster, but its supply voltage was 1.232 V.

When the processor is idle, its multiplier is reduced to the value "x12", thereby reducing the frequency to 1200 MHz. The Intel Core i7-4930K behaved in the same way in the absence of load. However, here the speed of 1200 MHz is achieved at a voltage of 0.684 V, while in the case of the Intel Core i7-4930K, 0.824 V was required.

The cache memory of the Intel Core i7-5930K is distributed as follows:

cache memory of the first level L1: 32 KB for data with 8 associativity channels and 32 KB for instructions also with 8 associativity channels are allocated for each of the 6 cores;
cache memory of the second level L2: 256 KB for each core with 8 channels of associativity;
L3 Cache: 15 MB for all cores with 20 associativity channels.

Recall that its predecessor L3 cache was 3 MB less.

The DDR4 RAM controller operates in quad-channel mode and is guaranteed to support modules with a frequency of up to 2133 MHz. The maximum memory capacity is 64 GB.

Testing

To test the Intel Core i7-5930K processor, we were kindly provided with the motherboard GIGABYTE board GA-X99-Gaming G1 WiFi and a set of Corsair Vengeance LPX DDR4-2800 series RAM with a total of 16 GB (4 x 4 GB). The rest of the test bench configuration remained unchanged.

When testing, we used the Stand for testing Processors No. 2

Motherboards (AMD)	ASUS F1A75-V PRO (AMD A75, Socket FM1, DDR3, ATX), GIGABYTE GA-F2A75-D3H (AMD A75, Socket FM2, DDR3, ATX), ASUS SABERTOOTH 990FX (AMD 990FX, Socket AM3+, DDR3, ATX)
Motherboards (AMD)	ASUS SABERTOOTH 990FX R2.0 (AMD 990FX, Socket AM3+, DDR3, ATX), ASRock Fatal1ty FM2A88X+ Killer (AMD A88X, Socket FM2+, DDR3, ATX)
Motherboards (Intel)	ASUS P8Z77-V PRO/THUNDERBOLT (Intel Z77, Socket LGA1155, DDR3, ATX), ASUS P9X79 PRO (Intel X79, Socket LGA2011, DDR3, ATX), ASRock Z87M OC Formula (Intel Z87, Socket LGA1150, DDR3, mATX)
Motherboards (Intel)	ASUS MAXIMUS VIII RANGER (Intel Z170, Socket LGA1151, DDR4, ATX) / ASRock Fatal1ty Z97X Killer (Intel Z97, Socket LGA1150, DDR3, mATX), ASUS RAMPAGE V EXTREME (Intel X99, Socket LGA2011-v3, DDR4, E-ATX )
Coolers	Scythe Mugen 3 (Socket LGA1150/1155/1366, AMD Socket AM3+/FM1/ FM2/FM2+), ZALMAN CNPS12X (Socket LGA2011), Noctua NH-U14S (LGA2011-3)
RAM	2 x 4 GB DDR3-2400 TwinMOS TwiSTER 9DHCGN4B-HAWP, 4 x 4 GB DDR4-3000 Kingston HyperX Predator HX430C15PBK4/16 (Socket LGA2011-v3)
video card	AMD Radeon HD 7970 3 GB GDDR5, ASUS GeForce GTX 980 STRIX OC 4 GB GDDR5 (GPU-1178 MHz / RAM-1279 MHz)
HDD	Western Digital Caviar Blue WD10EALX (1TB, SATA 6Gb/s, NCQ), Seagate Enterprise Capacity 3.5 HDD v4 (ST6000NM0024, 6TB, SATA 6Gb/s)
power unit	Seasonic X-660, 660 W, Active PFC, 80 PLUS Gold, 120 mm fan
operating system	Microsoft Windows 8.1 64-bit

Choose what you want to compare Intel Core i7-5930k Turbo Boost ON to

Before proceeding directly to the analysis of the graphs presented above, we want to say a few words about the effect of Intel Turbo Boost 2.0 technology on the speed of a computer. Disabling it leads to a drop in system performance by an average of 1-2%, depending on the application. Such results are quite natural, since the difference in frequency between nominal and turbo mode is only 200 MHz. However, taking into account the fact that the activation of Intel Turbo Boost 2.0 technology practically does not lead to an increase in the supply voltage, this function we recommend leaving it enabled.

Now let's move on to comparing the novelty with its competitors. Of course, they are high-performance solutions from the Intel Ivy Bridge-E family: Intel Core i7-4960X Extreme Edition and Intel Core i7-4930K. We also included top versions of "civilian" processors (Intel Core i7-4770K and AMD FX-9370) in the test to analyze whether it makes sense to choose the Socket LGA2011-v3 platform when building a powerful PC.

The hero of the review, the Intel Core i7-5930K, showed very high performance in test benchmarks, however, compared to its predecessor, the performance increase was not so big - about 5%. The new flagship of the Intel Ivy Bridge-E series is on average 2% ahead. In other words, owners of top processors for the Socket LGA2011 platform will not receive a noticeable increase in performance after purchasing the Intel Core i7-5930K. And given that when updating the system, you will need to change more motherboard, as well as to purchase expensive DDR4 memory, the expediency of such an undertaking looks doubtful. It would be much more reasonable to invest these funds in installing a more powerful video card or some other type of components.

However, if a computer is purchased "from scratch" or a system assembled on the basis of less productive platforms (Socket LGA1155/1150, Socket AM3+, Socket FM2/FM2+) is upgraded, then the situation changes dramatically. Despite the seemingly increased frequencies of the Intel Core i7-4770K and AMD FX-9370 solutions, they can't even compete with the hero of the review in terms of performance. In this case, the average difference in performance is 23% and 62%, respectively (of course, in favor of the Intel Core i7-5930K). Therefore, the additional costs caused by the purchase of a more expensive motherboard and memory look more justified, since it is with the help of such solutions that you can maximize the capabilities of a bundle of several video cards in games or achieve maximum productivity with complex calculations.

Although the TDP level of the novelty is nominally 10 W higher than that of its predecessor, the Intel Core i7-4930K, in practice both systems consumed approximately the same amount of energy. Yes, and to call the Intel Core i7-5930K an "oven" against the background of the AMD FX-9370 somehow also does not turn the tongue.

Overclocking

Since the representatives of the Intel Haswell-E family belong to high-performance solutions, they all have an unlocked multiplier, which greatly simplifies the overclocking process. Although, if necessary, overclocking can be done by changing the reference frequency.

In our experiment, we used the first method. By raising the multiplier to the value "x46", the speed of the novelty increased to 4600 MHz, while the base frequency was fixed at around 100 MHz, and the voltage had to be increased to 1.365 V. In this mode, the Intel Core i7-5930K passed the stress test in the LinX program without errors 0.6.4. During the experiment, the maximum recorded temperature was 93°C for the hottest core (using the Scythe Mugen 3 bench cooler). As a result, the speed increase was 31.4% relative to the nominal frequency (3500 MHz), which can be considered a pretty good result for air cooling.

The optimization of the Intel Core i7-5930K parameters affected the performance of the computer as follows:

			In nominal mode	During acceleration

		Computing Suite



SiSoft Sandra 2012	Arithmetic	Overall performance, GOPS
		Whole dhrystone, gypsum
		Whetstone double float, GFLOPS
	Multimedia	Overall multimedia performance, MPixels/s
		Multimedia integers, MPixels/s
		Multimedia FP32/FP64 floating point MPixels/s


		CPU (Single Core), pts

Fritz Chess Benchmark 4.2, knodes/s


Batman Arkham City		DirectX 11 (fps)
Resident Evil 5 Benchmark		DirectX 10, Anti-Aliasing x8 (fps)


Average value

The average increase in productivity was 15.54%. Such an increase in speed in most cases will be noticed by the user during operation, so it makes sense to optimize the parameters of this processor. However, it should be borne in mind that quite a lot will fall on the cooling system. huge pressure, which not every air cooler can handle.

With this result, we were able to take third place in the overall overclocking rating of the Intel Core i7-5930K model, which is published on the popular overclocking resource HWBot.org.

The best result at the time of writing was 4854 MHz. It is noteworthy that in this case, overclocking was carried out by increasing the reference frequency, and not by the multiplier. A liquid-type system was used for cooling.

Processor family Intel Haswell-E is a worthy continuation of a series of high-performance solutions produced by Intel. As in the cases with the Intel Sandy Bridge-E and Intel Ivy Bridge-E lines, here the manufacturer also tried to combine modern developments in the field of processor building and the IT industry.

So, representatives of the Intel Haswell-E family are based on the advanced 22-nm process technology and the Intel Haswell microarchitecture. They received support for DDR4 memory and new AVX2 and FMA3 instructions, and the number of processing cores in the maximum configuration was increased from 6 to 8. In a word, Intel Haswell-E is not a rebranding of old solutions, but a truly new generation of processors with many improvements at the structural and electrical levels.

We explored their capabilities using the example of a 6-core model IntelCorei7-5930 K. With a recommended cost of about $ 583, it demonstrates phenomenal performance and easily leaves the flagships of the Intel Haswell and Intel Devil's Canyon series. The average difference in performance is 20-23%, but in multi-threaded applications it can reach all 40-50%. The same results are observed in games.With 40 lanes of PCI Express 3.0, the Intel Core i7-5930K is an excellent choice for building a powerful gaming configuration with multiple graphics cards.In addition, if necessary, you can always increase the performance of your computer by overclocking the processor. by changing the multiplier or by raising the reference frequency.In our case, we easily increased its speed to 4600 MHz, which resulted in a 15% increase compared to the speed at the nominal frequency (3500 MHz). get a good cooling system, perhaps even a liquid type. The use of a high-quality thermal interface between the crystal and the heat-distributing cover, TDP of 140 W still makes itself felt.

Against the background of representatives of the previous Intel Ivy family Bridge-E model The Intel Core i7-5930K doesn't look that impressive. In terms of performance, it can be compared with the Intel Core i7-4960X Extreme Edition and only 5% outperforms the Intel Core i7-4930K. Thus, the transition from the old generation to the new one does not look like a very expedient undertaking, because in this case you will have to change not only the processor itself, but also the motherboard, and also spend money on expensive DDR4 memory.

Although on the other hand, AMD still does not have a decent answer. Therefore, if you want to assemble a modern powerful gaming computer or productive desktop workstation, then here, as they say, without options - only processors from the Intel Haswell-E family.

We are grateful to the company