Intel Core i7-4790K processor review | Devil's Canyon series for enthusiasts
Just think - it's been a year since we wrote about Intel as a company aimed at developing desktop PCs. And even in that case we discussed the processor Intel Core i7-4960X as a slightly updated product for a platform that appeared about three years ago (we're talking about the X79 Express, in case you forgot). And that's a pretty good indicator of how long Intel has ignored us enthusiasts.
Although it is perhaps not entirely fair to say so; Maybe Intel was just starting to run out of steam, while AMD products were better suited to meet the needs of power users who wanted high-end desktop performance. As a result of such sad stagnation, we stopped writing reviews like "Core i7-4770K on Haswell architecture: preliminary review and test". We were quite strict about Intel Core i7-4770K, the same attitude on our part was towards. None of Intel's flagships gave us compelling reasons to choose them, and they failed to emphasize the importance of the transition from the Z77 to the Z87 to the Z97 Express.
But it's good that Intel is listening to these murmurs.
We received a processor for testing Intel Core i7-4790K. Based on the nomenclature, it is still built on the Haswell architecture for the LGA 1150 interface. It will even be suitable for existing motherboards with the Z87 chipset, if their manufacturers are responsible for updating the firmware. I wonder what Intel managed to do with the new processor Intel Core i7-4790K to attract the attention of enthusiasts?
Getting to know the Intel Core i7-4790K
To begin with, one of the complaints expressed on behalf of power users and system builders who had to separate the integrated heat distribution plate from the processor (a process called “delidding”) was taken into account. We don't know much about the Next-Generation Polymer Thermal Interface Material that the company uses in its Devil's Canyon processors, but we know that Intel has rushed (uncharacteristically for a manufacturer) to introduce it into commercial solutions. The speed with which the company came to the decision to use this particular material is one of the reasons why it will no longer be possible to go back to the principle that made it possible to successfully overclock Sandy Bridge.
From below Intel Core i7-4790K equipped with additional capacitors, which, according to Intel, should provide stable uninterruptible power supply processor crystal.
These modifications are implemented in one product, which is ready to replace the previous model Intel Core i7-4770K, also compatible with LGA 1150. The most basic specifications are very similar: four physical cores work with eight threads thanks to Hyper Threading, there is also 8 MB of shared L3 cache, the already familiar integrated HD Graphics 4600 video card and a 16-lane PCI Express 3.0 controller, with This supports DDR3 memory with a maximum speed of 1600 MT/s.
Intel Core i7-4790K has a base frequency of 4 GHz and reaches 4.4 GHz thanks to technology Turbo Boost, which, in turn, requires maximum power 88 W (from 84 W).
| Product | Intel Core i7-4790K | Intel Core i7-4770K |
| Price | $339,99 | $324,50 |
| Number of Cores | 4 | 4 |
| PCIe lanes | 8 | 8 |
| Base frequency, GHz | 4 | 3,5 |
| Maximum Turbo Boost frequency, GHz | 4,4 | 3,9 |
| Technical process, nm | 22 | 22 |
| TDP, W | 88 | 84 |
| Memory support | 32 GB, DDR3-1600 | 32 GB, DDR3-1600 |
| Video card | HD Graphics 4600 | HD Graphics 4600 |
Devil's Canyon processors are ready to run on 9-series motherboards, although Intel says 8-series platform makers should simply update the firmware.
Intel Core i7-4790K processor review | Overclocking and thermal pad
We were very big fans of aggressive overclocking at first, reaching peak values at settings that weren't designed for long-term system performance. Everything is for the benefit of big numbers, right?
I often heard from many readers that processors and video cards functioned well for six months, but when initial settings no longer provide job stability. Now it seems that it is more important to achieve high but stable clock speeds that can comfortably handle any load. When we talk to system builders (those who need overclocking while also having to comply with warranty obligations on their configurations), the discussions become more candid and we find out which parameters will receive attention over time.
So, today we have to compare Intel's previous flagship with its new product, so we launched both processors in stock and overclocked states using new set our benchmarks.
| Test equipment | |
| Processors | Intel Core i7-4790K (Haswell) 4.0 GHz (40 * 100 MHz), LGA 1150, 8 MB shared L3 cache, Hyper-Threading on, Turbo Boost on, power saving on Intel Core i7-4770K (Haswell) 3.5 GHz (35 * 100 MHz), LGA 1150, 8 MB shared L3 cache, Hyper-Threading on, Turbo Boost on, power saving on. |
| Motherboard | MSI Z97 Gaming 7 (LGA 1150) Intel Z97 Express, BIOS 1.3 |
| Memory | G.Skill 16 GB (4 x 4 GB) DDR3-1600, F3-12800CL9Q2-32GBZL @ DDR3-1600 @ 1.5 V |
| Storage device | Samsung 840 Pro 256 GB, SATA 6 Gb/s |
| Video card | Nvidia GeForce GTX Titan 6 GB |
| power unit | Corsair AX860i, 80 PLUS Platinum, 860 W |
| Radiator | Noctua NH-U12S, 100% fan cycle |
| Software and drivers | |
| operating system | Windows 8 Professional x64 |
| DirectX | DirectX 11 |
| Graphics driver | Nvidia GeForce Release 337.88 |
IntroductionIt seems that just recently the ideology of processor overclocking involved purchasing cheap CPU models and overclocking them to the level of older models or even stronger. That is, thanks to their skills and experience, enthusiasts actually saved money without losing the performance level of their own systems. But with the growing popularity of overclocking, this situation no longer suited computer equipment manufacturers, who, in the end, were able to turn the whole overclocking upside down. For example, Intel, which is the main developer and supplier central processing units, recently it has become possible to overclock only special models that have a higher price than regular CPUs. It is not at all surprising that all this has had a strong impact on the overclocking movement. It has become less widespread, turning from a common practice into a kind of sport. However, it cannot be said that the ranks of enthusiasts have thinned catastrophically. Processors with unlocked frequency increase capabilities continue to be in good demand, which leads to the emergence of more and more CPU modifications for overclockers.
However, by releasing specialized overclocking CPUs, Intel has also taken on certain obligations: the processors that it offers at an increased price with an eye to overclocking use must have good frequency potential, otherwise what’s the point in them? And this is where the problem arises. Currently, when developing new processor architectures, the main priority for Intel is not increasing the frequency potential, but improving the efficiency of processor cores. And this doesn't fit well with overclocking. Since the emergence of Sandy Bridge family processors on the market, the release of each new generation of microarchitecture entails a decrease in frequency potential. And if processors of the Sandy Bridge generation were often able to boost up to frequencies of 4.8-5.0 GHz for everyday use, then modern Haswells fall far short of this milestone, stopping at frequencies of 4.3-4.5 GHz, which, in fact, , do not exceed the nominal ones too much.
All this began to threaten Intel with both monetary and reputational losses, because many owners of classic desktop systems refuse to switch to new generations of CPUs precisely because of their poor overclocking. The current situation began to strain Intel so much that in December last year, six months after Haswell processors went on sale, management decided on the need to take urgent measures that would improve Intel's image among enthusiasts. This is how the idea of Devil’s Canyon was born - a new version of Haswell for overclockers, which was eventually developed and put into mass production in record time.
Two new Devil’s Canyon series processors, along with another interesting new product - the anniversary overclocker Pentium G3258 processor - were announced in June and went on sale in July-August. Two new products were presented: a “simple” quad-core Core i5-4690K and a flagship quad-core processor for the LGA 1150 platform with support for Hyper-Threading technology, Core i7-4790K. These two processors seem to complement the LGA 1150 Haswell Refresh update released in May. The new Devil's Canyon acts as a Haswell Refresh for enthusiasts, replacing its Core i5-4670K and Core i7-4770K predecessors with models with slightly higher clock speeds. However, a simple increase in frequency would hardly arouse even a drop of interest on the part of overclockers, so the matter is not limited to this. In addition, the microprocessor giant promises a noticeable improvement in overclocking potential, and this is what makes the new processors remarkable. Let's get to know them in more detail.
Features of Devil's Canyon
Considering that Devil’s Canyon was developed in emergency mode, there are no improvements at the microarchitecture level. The advantage of these processors over their predecessors is determined only by such engineering solutions, which could be cranked without going deep into the “guts” of the processor. However, while Haswell Refresh differs from Haswell of the “first wave” only by slightly increased clock frequencies, Devil’s Canyon can be counted on three fundamental differences from previous overclocker Haswells.The first change will be of interest to those users who use overclocking processors in nominal mode. Considering that with the release of Haswell Refresh, the frequency increased for all processors for the LGA 1150 platform, the old modifications Core i5-4670K and Core i7-4770K are no longer the fastest CPUs in their class. Therefore, the new Core i5-4690K and Core i7-4790K frequencies had to be increased. Moreover, in this case we are not talking about a 100-MHz increase, as in Haswell Refresh, but about a more noticeable acceleration. For example, the older model Core i7-4790K, in its nameplate frequency, was even able to reach the psychologically important mark of 4.0 GHz, which until now none of Intel’s processors for desktop systems could cross.
The second change is aimed directly at improving the frequency potential during overclocking. In Devil's Canyon, the composition of the material located between the semiconductor crystal and the processor cover has changed. This step is a direct response to criticism from overclockers who expressed dissatisfaction with the poor thermal conductivity of the internal thermal interface of old Haswell and Ivy Bridge, due to which the processor cores overheated even in systems with very efficient cooling systems. However, the change that has occurred inspires only moderate optimism. The fact is that most effective method mating the crystal and the cover - through flux-free soldering with high thermal conductivity solder - never returned to Devil’s Canyon. In fact, Intel simply replaced one type of thermal paste with another. The new material, called NGPTIM (Next-Generation Polymer Thermal Interface Material) in the documentation, is distinguished by its chemical composition and has better thermal conductivity than before.
If you remove the processor cover from Devil’s Canyon, the following picture opens before your eyes.
The appearance of the thermal interface has not changed compared to the old Haswell ones. Naturally, there is no solder in Devil’s Canyon; the gray substance used inside is approximately the same as before. However, the plasticity of this material has improved, it does not look dry, and you can remove it like regular high-quality thermal paste - with a soft cloth. As for thermal conductivity, experiments show that for NGPTIM it has not increased too noticeably. This heat-conducting material loses in direct comparison even to such an inexpensive and widely available thermal paste like Arctic MX-2. Therefore, one should not expect serious changes in the temperature regime of new overclocking processors. Practical experiments show that at the same frequency and voltage, Devil’s Canyon heats up about 10 degrees less than their predecessors, which actually doesn’t change anything fundamentally. Radical enthusiasts will have to continue scalping Devil’s Canyon processors to achieve maximum overclocking results.
By the way, the third innovation in Devil’s Canyon is aimed precisely at such advanced overclocking fans who do not disdain extreme approaches to overclocking. In order to improve the stability of supply voltages, Intel has developed a new electrical wiring diagram, which is reflected in the appearance of additional external capacitors on the “belly” of the processor and under its cover. The new power supply improves voltage consistency under high load, essentially acting as an analogue of the anti-voltage drop that is typically found on enthusiast motherboards. Regular “home” overclocking of Haswell without using extreme cooling methods actually does not require anything like this, but when experimenting with cooling the processor with dry ice or liquid nitrogen, additional stabilization may come in handy.
There are no other changes in Devil's Canyon compared to Haswell. Contrary to wild expectations, the semiconductor crystal in Devil’s Canyon remained unchanged. Its stepping is C0: the same core stepping can be observed both in Haswell Refresh and in regular Haswells from a year ago. Therefore, all the promised improvements in frequency potential are solely the result of changing the internal thermal interface and improving the power supply circuit.
Specifications
The most important news that can be seen in the specifications is a significant increase in the clock speed of the Core i7-4790K processor. While the nominal frequency of its predecessor, the Core i7-4770K, was set at 3.5 GHz with the ability to auto-overclock in turbo mode to 3.9 GHz, the new product aims to conquer the 4-GHz milestone. It is at this level that the standard frequency of the older Devil’s Canyon, Core i7-4790K, is set, that is, this processor always operates at a frequency of at least 4.0 GHz. Thanks to Turbo Boost technology, the maximum nominal frequency of this new product under low load can reach up to 4.4 GHz. Thus, the Core i7-4790K received a 14 percent speedup compared to its predecessor, which looks like a very generous gift from Intel, especially considering the fact that the recommended price is identical to the Core i7-4770K. The only side effect of the resulting increase in clock frequency is an increase in the calculated heat dissipation, which increased from 84 to 88 W.
The younger processor from the Devil’s Canyon duo, Core i5-4690K, cannot boast of the same significant increase in frequency. From a specifications point of view, it could well be included in the Haswell Refresh ranks, because its frequency has risen by only one 100 MHz step compared to the Core i5-4670K. The old overclocker Core i5 operated at frequencies of 3.4-3.8 GHz, and its successor Devil’s Canyon received frequencies of 3.5-3.9 GHz. At the same time, for the Core i5-4690K the manufacturer also pushed the thermal envelope to 88 W, but this was done purely formally. Having the same workers Core specifications The i5-4690 (non-overclocker Haswell Refresh) has a TDP of 84 W.
Note that the new K-series processors for the LGA 1150 platform, unlike their predecessors, received support for vPro and TXT security technologies, VT-d virtualization and, formally, TSX-NI instructions. This partly makes Devil’s Canyon interesting, including in the corporate sector. However, it should be recalled that recently an error was found in the TSX-NI set, due to which its implementation has become impossible for now. With the release of new BIOS versions, this set will simply be disabled in processors at the microcode level. But most likely, ordinary users will not be affected by this: TSX-NI instructions are needed to implement the transactional memory model in multi-threaded environments, and their use can only be justified in specialized applications, typical only for multiprocessor servers.
Let's compare the characteristics of Devil's Canyon processors with the specifications of their predecessors.
The new processors belonging to the Devil’s Canyon series continue to be designed for the LGA 1150 ecosystem. They are compatible with motherboards based on 9-series chipsets, and with the vast majority of motherboards with 8-series chipsets (in the latter case, a firmware update is required). At the same time, among the boards based on the Z87 chipset, there are also models that are not capable of working with Devil’s Canyon, so if you plan to build a system on new processors and motherboards from a year ago, be sure to check their compatibility on the motherboard manufacturer’s website.
I would also like to emphasize the fact that with the release of the Core i7-4790K and Core i5-4690K, Intel has widened the performance gap between the older representatives of the i5 and i7 series. If previously the main difference between such processors for enthusiasts was only a more capacious L3 cache and Hyper-Threading technology, now the half-GHz difference in clock frequency should also be taken into account. In other words, the 40% difference in the cost of the older i5 and i7 has become more justified, which is likely to persuade many to switch to the Core i7-4790K rather than to its younger brother.
As for practical aspects, the copy of the Core i7-4790K processor we received for testing functioned in full accordance with the stated specifications - at a frequency of over 4.0 GHz. With Turbo Boost 2.0 technology activated, with a multi-threaded load, its frequency was 4.2 GHz, but in the case of single-threaded activity it increased to the promised 4.4 GHz. The operating voltages were in the range of 1.12-1.20 V, which, it should be noted, is slightly higher than the voltages that are characteristic of both conventional Haswell and their newer Haswell Refresh modifications.
Considering that the Core i5-4690K, like its older brother, has a TDP of 88 W, one would expect its operating voltages to also be relatively high. But no, the sample we took worked nominally at 1.06 V, and when Turbo Boost 2.0 technology was activated, this voltage increased to only 1.09 V. At the same time, the typical real frequency of the Core i5-4690K with active turbo mode and high computing load is 3 .7 GHz.
Overclocking
Devil's Canyon is aimed at the enthusiast audience by Intel. Therefore, both processors included in this series have unlocked multipliers, allowing you to independently overclock the processor, bus, memory and graphics core. However, there is nothing new in this list, and in terms of overclocking functions, the Core i7-4790K along with the Core i5-4690K are similar to the previous Haswell K-series.Previously, Haswell with an unlocked multiplier was somewhat disdainful among overclockers. This is due to high heating, due to which the maximum frequency suitable for everyday use of the processor did not exceed 4.3-4.5 GHz. But Devil’s Canyon is presented by Intel as “working on mistakes” primarily in terms of heat removal from the processor chip. That is, the manufacturer directly promised enthusiasts that new CPUs should heat up less and, due to this, are able to conquer higher frequencies without any surgical intervention (scalping) or replacing the internal thermal interface.
However, in practice, immediately after its release, Devil’s Canyon did not demonstrate anything good. This even forced us to refuse to write a review of the engineering samples sent by Intel. In overclocking, they turned out to be no better than their predecessors, which suggested that something was wrong with them. Now we are waiting for production copies of processors to appear in our laboratory. And guess what? Nothing changed!
The first to take part in overclocking experiments was the junior quad-core Core i5-4690K. The maximum frequency achievable for it during overclocking turned out to be a modest but familiar 4.4 GHz. It should be understood that in this case we are not talking about an overclocker record, but about the frequency at which this processor can operate in a regular desktop system 24/7. Therefore, for cooling we used the Noctua NH-D15 air cooler with a dual-tower design, which, nevertheless, is one of the most effective solutions in this class.
To ensure stability during testing, the voltage had to be raised to 1.15 V. In this case, no problems arose, but when running the test in LinX 0.6.5, the temperature of the hottest core increased to 98 degrees, that is, almost to the 100-degree limit . Thus, the temperature factor, as before, continues to hinder the overclocking of the Core i5-4690K. The inefficiency of the internal thermal interface leads to overheating of the processor cores, and it is almost impossible to cope with this by increasing the performance of the cooling system.
As a result, increasing the processor supply voltage, which is necessary for stable operation during overclocking, is possible only within very limited limits. And in the end we get just such a sad overclocking. The maximum 4.4 GHz for our Core i5-4690K is not only no better than the typical overclocking of the Core i5-4670K released more than a year ago, but it is also the rated frequency of the Core i7-4790K in turbo mode.
Maybe the flagship processor overclocks better? Well, if only a little bit. The maximum frequency we achieved during overclocking experiments was 4.5 GHz. To ensure stability, the supply voltage had to be increased to 1.2 V - at a lower voltage the system crashed. blue screen" The temperature in this case, during testing in LinX 0.6.5 on the hottest core, reached 96 degrees.
In other words, even with the use of highly efficient air cooling, there is no noticeable overclocking that is noticeably higher than before. And the fault for this again lies with the internal thermal interface, which simply does not allow heat to be removed from the processor chip and constrains users from increasing the supply voltage. Sandy Bridge processors, produced using 32 nm technology, could operate freely at a voltage of 1.4-1.5 V, and at the same time, all the generated heat could be dissipated by the air cooling system with good efficiency. In the production of Haswell and Devil's Canyon, a more advanced 22 nm process technology is used, but the voltage can only be increased to 1.2 V, and then overheating follows. This is a very significant manifestation Low quality existing thermal interface.
Unfortunately, all the Haswell problems that appear during overclocking have completely moved to Devil’s Canyon. When a load occurs, the temperature of the processor cores increases sharply and significantly, and keeping it within acceptable limits requires considerable effort. It turns out that the new LGA 1150 processors, with normal non-extreme overclocking, are almost no better than their overclocking predecessors of the Haswell family. To turn Haswell into a new Sandy Bridge, simply changing the chemical composition of the polymer thermal interface is not enough. So we can only hope that flux-free soldering of the semiconductor chip to the processor cover with low-melting solder based on indium metal will return in future LGA1150 products for Broadwell generation enthusiasts. At the moment, the only current processors that use soldering are the just released Haswell-E, but they, unlike Devil’s Canyon, belong to the premium price category.
How we tested
We compared the performance of the new Devil’s Canyon processors with the speed of their predecessors, senior representatives of the Haswell line from last year, Core i5-4670K and Core i7-4770K. However, we did not stop there. Starting this week, AMD is introducing new prices for its FX processors that make them (relatively) competitive again. For example, the older FX-9590 receives an official price tag of $230, which allows it to be compared with Intel's Core i5. That is why you will also find the AMD FX-9590 among the testing participants.In order to show the scalability of Devil’s Canyon performance when overclocked, we tested the Core i5-4690K and Core i7-4790K both when operating in nominal mode and at the maximum achieved frequency, which, however, does not differ too much from the nominal one.
As a result, the list of hardware components involved in testing looked like this:
Processors:
AMD FX-9590 (Vishera, 8 cores, 4.7-5.0 GHz, 4x2 MB L2, 8 MB L3);
Intel Core i7-4790K (Haswell Refresh, 4 cores + HT, 4.0-4.4 GHz, 4x256 KB L2, 8 MB L3);
Intel Core i7-4770K (Haswell, 4 cores + HT, 3.5-3.9 GHz, 4x256 KB L2, 8 MB L3);
Intel Core i5-4690K (Haswell, 4 cores, 3.5-3.9 GHz, 4x256 KB L2, 6 MB L3);
Intel Core i5-4670K (Haswell, 4 cores, 3.4-3.8 GHz, 4x256 KB L2, 6 MB L3).
CPU cooler: Noctua NH-D15.
Motherboards:
ASUS M5A99FX Pro R2.0 (Socket AM3+, AMD 990FX + SB950);
ASUS Z97-Pro (LGA1150, Intel Z97).
Memory: 2x8 GB DDR3-2133 SDRAM, 9-11-11-31 (G.Skill F3-2133C9D-16GTX).
Video card: NVIDIA GeForce GTX 780 Ti (3 GB/384-bit GDDR5, 876-928/7000 MHz).
Disk subsystem: Intel SSD 520 240 GB (SSDSC2CW240A3K5).
Power supply: Seasonic Platinum SS-760XP2 (80 Plus Platinum, 760 W).
Testing was performed on the operating system Microsoft Windows 8.1 Professional x64 with Update using the following set of drivers:
AMD Chipset Drivers 14.4;
Intel Chipset Driver 10.0.14;
Intel Management Engine Driver 10.0.0.1204;
Intel Rapid Storage Technology 13.0.3.1001;
NVIDIA GeForce 340.52 Driver.
Performance
Overall PerformanceTo evaluate processor performance in common tasks, we traditionally use the Bapco SYSmark test package, which simulates user work in real common modern office programs and applications for creating and processing digital content. The idea of the test is very simple: it produces a single metric characterizing the weighted average speed of the computer during everyday use. Recently, this benchmark was updated once again, and now we use the most latest version– SYSmark 2014.
The increase in clock speed of the Core i5-4690K compared to the Core i5-4670K was 3 percent. The older Core i7-4790K received a 14 percent higher frequency than its predecessor, the Core i7-4770K. If we keep this in mind, the results obtained become well understandable, because these are the proportions in which the performance in SYSmark 2014 increased. As a result, the Core i7-4790K outperforms the Core i5-4690K by as much as 27 percent, which is much more noticeable than the previously observed difference in performance older Core i5 and Core i7. However, if you take into account overclocking, which allows you to increase the frequency of the Core i5-4690K by 25 percent, and the Core i7-4790K by only 13 percent, then it is not at all surprising that the discrepancy in results drops to 17 percent.
A deeper understanding of SYSmark 2014 results can be provided by familiarizing yourself with the performance estimates obtained in various system usage scenarios. The Office Productivity scenario simulates typical office work: writing texts, processing spreadsheets, working with email, and surfing the Internet. The script uses the following set of applications: Adobe Acrobat XI Pro, Google Chrome 32, Microsoft Excel 2013, Microsoft OneNote 2013, Microsoft Outlook 2013, Microsoft PowerPoint 2013, Microsoft Word 2013, WinZip Pro 17.5 Pro.
The Media Creation scenario simulates the creation of a commercial using pre-shot digital images and videos. For this purpose, the popular packages Adobe Photoshop CS6 Extended, Adobe Premiere Pro CS6 and Trimble SketchUp Pro 2013.
The Data/Financial Analysis scenario is devoted to statistical analysis and investment forecasting based on a certain financial model. The scenario uses large amounts of numerical data and two applications: Microsoft Excel 2013 and WinZip Pro 17.5 Pro.
In applications different types Devil's Canyon's relative advantage over their Haswell predecessors remains constant. This is precisely the benefit of increasing the clock frequency - it always gives a predictable positive result. The difference in performance between Core i5 and Core i7 varies depending on the load. We have already talked more than once about what tasks Core i7 processors are good for. They are best used when working creatively with multimedia content and in calculation tasks, for example, financial ones. For general activity, you can easily limit yourself to the Core i5-4690K, which after overclocking lags behind the overclocked Core i7-4790K by only 7 percent. A similar recommendation can be extended to games, but let’s not get ahead of ourselves.
Gaming Performance
As you know, the performance of platforms equipped with high-performance processors in the vast majority of modern games is determined by the power of the graphics subsystem. That is why, when testing processors, we select the most processor-dependent games, and measure the number of frames twice. The first pass tests are carried out without turning on anti-aliasing and with settings that are far from the most high resolutions. Such settings allow you to evaluate how well processors perform with a gaming load in principle, and therefore allow you to speculate about how the tested computing platforms will behave in the future, when more quick options graphics accelerators. The second pass is performed with realistic settings - when selecting FullHD resolution and the maximum level of full-screen anti-aliasing. In our opinion, such results are no less interesting, since they answer the frequently asked question about what level of gaming performance processors can provide right now - in modern conditions.
Results in FullHD resolution with maximum quality settings:
We have long said that older processors of the Core i5 and Core i7 series are approximately equal in modern games. The release of Devil's Canyon didn't change anything. From the graphs above, you can clearly see that the maximum difference in frame rate is 3 percent. In other words, for gaming systems, purchasing a Core i7-4790K instead of a Core i5-4690K doesn’t make too much sense. Moreover, during overclocking the frequencies of these processors are compared, which further brings the results they show closer together.
Naturally, it is almost impossible to see in 3D games any progress in gaming performance when moving from the old Haswell to the new Devil’s Canyon of the same class. Therefore, the only clear fact that can be noticed by looking at the above diagrams is that AMD processor The FX-9590 performed weaker than the Core i5 and Core i7 processors.
Results at reduced resolution:
If you lighten the load on the graphics subsystem a little by lowering the resolution, there will be much more food for thought. Here the Core i5-4690K processor has become faster than its predecessor by an average of 2 percent. The older Core i7-4790K, whose frequency has stepped far ahead, is ahead of the Core i7-4770K much more significantly - by 12 percent. The difference in performance between the members of the Devil’s Canyon duo is about 13 percent on average, but depending on the optimization of the game engine for multi-threading, it can vary from 3 to 33 percent. It turns out that the gaming potential of the Core i7-4790K is significantly higher than that of its brother and than that of its predecessors. However, to reveal it, the power GeForce video cards GTX 780 Ti is not enough. But never mind, soon we will meet a new generation of graphics accelerators.
Tests in applications
In Autodesk 3ds max 2015 we test the final rendering speed. Measures the time it takes to render at 1920x1080 resolution using the renderer mental ray one frame of the standard Space_Flyby scene from the SPEC test package.
Processing Performance Testing graphic images takes place in Adobe Photoshop CC 2014. Measures the average execution time of a test script that is a creative reworking of the Retouch Artists Photoshop Speed Test, which involves typical processing of four 24-megapixel images taken with a digital camera.
Due to numerous requests from amateur photographers, we conducted performance testing in graphics program Adobe Photoshop Lightroom 5.6. The test scenario includes post-processing and export to JPEG with a resolution of 1920x1080 and maximum quality two hundred 12-megapixel RAW images taken with a Nikon D300 digital camera.
IN Adobe Premiere Pro CC 2014 is tested for performance in non-linear video editing. The time for rendering a Blu-Ray project containing HDV 1080p25 video with various effects applied is measured.
Performance of websites and Internet applications built using modern technologies, we measure it in Internet Explorer 11. For this purpose, we use a specialized test, WebXPRT 2013, which implements algorithms actually used in Internet applications in HTML5 and JavaScript.
To measure the speed of processors when compressing information, we use the WinRAR 5.0 archiver, with which we archive a folder with various files with a total volume of 1.7 GB with the maximum compression ratio.
To evaluate the speed of video transcoding into the H.264 format, the x264 FHD Benchmark 1.0.1 (64bit) test is used, based on measuring the time the x264 encoder encodes the source video into MPEG-4/AVC format with a resolution of 1920x1080@50fps and default settings. It should be noted that the results of this benchmark are of great practical importance, since the x264 encoder underlies numerous popular transcoding utilities, for example, HandBrake, MeGUI, VirtualDub, etc. We periodically update the encoder used for performance measurements, and this testing involved version r2453, which supports all modern instruction sets, including AVX2.
In addition, we have added to the list of test applications a new x265 encoder designed for transcoding video into the promising H.265/HEVC format, which is a logical continuation of H.264 and is characterized by more efficient compression algorithms. To evaluate performance, a source 1080p@50FPS Y4M video file is used, which is transcoded into H.265 format with a medium profile. The release of the encoder version 1.2 took part in this testing.
There are no surprises in resource-intensive applications either. The increased clock speeds of Devil’s Canyon make them leaders in all tests, and the superiority over the older Haswells released a year ago is very noticeable. Despite the fact that in the case of the Core i5-4690K it is about 2 percent, the Core i7-4790K, operating in nominal mode, is 12 percent ahead of the Core i7-4770K. This means that if you are interested in high performance right out of the box, the Core i7-4790K is the best option. Its speed in resource-intensive tasks not only significantly exceeds the performance of its predecessor. It is more than a quarter faster than its Core i5 series brother.
By the way, in its marketing materials, Intel also points out that the older Devil’s Canyon processor, Core i7-4790K, runs applications more than 30 percent faster than the older Sandy Bridge, Core i7-2700K. Moreover, in some tasks, for example when creating and processing video content, the level of this advantage can be much higher and even reach 65 percent.
The scalability of performance during overclocking is rather weak. Overheating issues that keep Devil's Canyon processors from running at frequencies well above stock mean that overclockers can only benefit from 5 to 15 percent of extra performance, depending on whether the processor is a Core i7-4790K or a Core i5-4690K. - we are talking.
Energy consumption
In fact, everything is clear: Devil’s Canyon has no fundamental improvements in power consumption, because these processors are based on exactly the same semiconductor chip as all other quad-core LGA 1150 processors. Therefore, it is obvious that they will require more power than their predecessors, simply due to the higher clock speed. Moreover, for the Core i7-4790K and Core i5-4690K the manufacturer increased the thermal package limits by 4 W. However, all these theoretical calculations need practical testing. Let's move on to it.The following graphs, unless otherwise noted, show the total power consumption of systems (without monitor), measured at the outlet into which the test system's power supply is connected, and representing the sum of the power consumption of all components involved in the system. The total indicator automatically includes the efficiency of the power supply itself, however, given that the power supply model we use, Seasonic Platinum SS-760XP2, has an 80 Plus Platinum certificate, its influence should be minimal. To correctly assess energy consumption, we activated turbo mode and all available energy-saving technologies: C1E, C6 and Enhanced Intel SpeedStep.
First of all, idle consumption was measured.
When idle, no fundamental differences in the consumption of regular Haswell and Devil’s Canyon are noticeable. This is understandable: the activation of processor energy-saving technologies reduces the energy appetite of the CPU to almost zero. Therefore, what is reflected in the diagram in the case of LGA1150 systems is the total consumption of the motherboard, video card and SSD.
We then measured the maximum consumption under load created by the 64-bit version of the LinX 0.6.5 utility with support for the AVX2 instruction set, based on the Linpack package.
What can be seen in the diagram is the maximum level of consumption observed in practice. However, some concern is caused by the fact that the difference between the consumption of Haswell and Devil’s Canyon at the level of several watts is observed only in Core i5 series processors. The Core i7-4790K turned out to be more power-hungry than the Core i7-4770K, by as much as 14 W! This is logical, based on the significant increase in the clock frequency and supply voltage of the older Devil’s Canyon, but does not quite agree with its TDP level. In other words, it seems that you need to get used to the fact that this characteristic of Intel is associated with the maximum achievable power consumption and heat dissipation according to some implicit law. And in fact, modern Intel processors are capable of demonstrating real power consumption that significantly exceeds their thermal package.
Considering that the power consumption initiated by the LinX utility, based on the Linpack package, greatly exceeds the average realistic level, we measured the consumption under a more “mundane” load - video transcoding using the 64-bit version of the x264 codec version r2453.
Qualitatively the picture does not change. Devil's Canyon again show higher consumption than their predecessors. Meanwhile, the absolute numbers of energy consumption do not look as scary as in the previous case. Moreover, the increase in power consumption of new products turns out to be less than the performance advantage that they can offer in resource-intensive offerings. This means that the Core i7-4790K and Core i5-4690K are more profitable than the Core i7-4770K and Core i5-4670K, in terms of specific performance per watt of electricity consumed.
conclusions
Users desktop computers, who face resource-intensive tasks in their daily activities, are always happy to see new processors with increased performance. However, in the last few years, each subsequent generation of Intel CPUs has given only a slight increase in performance at the level of 5-10 percent, which was provided by small microarchitectural optimizations and nothing else. Clock frequencies remained almost unchanged. But this year Intel was unable to bring its next generation of processors, Broadwell, to market. Therefore, finally, an increase in operating frequencies came into play, through which new processors increased their speed. In other words, the new Devil’s Canyon provides a traditional increase in performance, but does it in a slightly unusual way, which can be described as “legal overclocking”.However, the matter was not limited to just increasing the frequency. Intel also slightly changed the power supply, adding additional stabilizing capacitors, and also tried to combat the high temperatures inherent in Haswell generation processors. This struggle, however, was carried out on its own the simple way– improving the thermal conductivity of the internal thermal interface material. But this also had some effect: under high load, temperatures dropped by about 10 degrees relative to the temperatures of the original Haswell. But did this make Devil's Canyon more interesting than the old Haswells? The answer to this question will depend on from what position you look at the new products.
Intel itself, when introducing the new Devil’s Canyon, emphasized that these processors were specifically intended for overclocking enthusiasts. However, overclockers most likely won't like Devil's Canyon. What Intel did to improve overclocking potential actually had virtually no effect. Under normal overclocking conditions, which involves constantly operating the processor at an increased frequency, we were able to see only a 100 MHz improvement in the frequency potential, and even then, only for one processor from the tested pair. At the same time, the temperatures of the processor cores remained as wild as before, and removing heat from the processor chip is still associated with significant problems. This means that if you already have an overclocker Haswell, there is no point in switching to Devil’s Canyon. More or less normal overclocking in both cases requires removing the cover from it and replacing Intel thermal paste with, for example, liquid metal.
It’s a different matter if you approach the new products as ordinary flagship processors for the LGA 1150 platform. In this case, the older model, Core i7-4790K, looks very attractive due to the greatly increased clock speeds. This processor was the first among Intel products to cross the 4-GHz bar, thanks to which it can offer approximately 10-15 percent higher performance than its predecessor. At the same time, the energy consumption and heat dissipation of the new product increased quite slightly, and the price generally remained at the old level. As a result, the Core i7-4790K can be called the fastest quad-core processor to date and one of the most advantageous offers upper price category. The second processor reviewed today, Core i5-4690K, increased its frequencies quite slightly. But in price it is equivalent to the Core i5-4670K, so there will be buyers for this model too.
In conclusion, I would like to say that the announcement of the next generation desktop processors, Broadwell, according to the latest data, has been postponed until the second quarter of next year, which means that Devil’s Canyon will remain the best options there is still a very long time for the LGA 1150 platform. However, even with the release of Broadwell, they may retain their appeal. The big question is whether Intel's promising flagship processors will be able to operate at clock speeds as high as the Core i7-4790K, or whether they will have to fall back to the traditional 3.5-3.9 GHz. This means that it may well be that buyers purchasing the older Devil's Canyon today are making a very shrewd investment.
ASUS motherboard Maximus VII Formula turned out to be good choice both for fans computer games, and for a large circle of IT enthusiasts. Excellent equipment, rich BIOS CPU overclocking capabilities, three PCI-e x16 slots for video cards and three more PCI-e x1 for additional function controllers, the ability to install ten SATA3 drives, support for processors Socket LGA1150– all this makes the ASUS motherboard a desirable basis for building a powerful computer.
There is already an article on our website where we looked at it in some detail (and sorted it out a little). It's time to test the overclocking capabilities of this motherboard, which is designed for installing a Socket LGA1150 processor. We tried our best overclock CPU Intel Core i7 4790K. What came of this - you will find out further as the article progresses. If you want to compare it with your existing solutions in standard operating modes, you can do this too - we present the results of several popular benchmarks without overclocking the computer.
In the graphical shell BIOS there is an impressive list of items dedicated to CPU overclocking. There are also several presets that can make the overclocking procedure easier for inexperienced beginners. The developers gave users the ability to choose a language GUI BIOS. There is even a Russian language in the list, but when you select it, very little changes... I think that most readers will agree that it would be very unusual to see abbreviations like “HDD” instead of “HDD” in the BIOS. If the translators had tried and translated everything, then no one except the translators themselves would have understood such an “adapted” interface. Please note that the screenshots show that “LN2 Mode” is in the “Enabled” state. This is what the unmarked jumper on the motherboard in the area of the “ProbeIt” pad is responsible for. This mode allows you to stabilize the system at subzero temperatures (hello, liquid nitrogen) and push back the Cold Boot and Cold Bug limits.
Motherboard test ASUS Maximus VII Formula and was carried out in a Corsair Graphite 600T case with an open side wall, with an Intel Core i7-4790K processor, a Noctua NH-D15 cooler, Kingston HiperX DDR3 2800 memory (2 sticks of 4 GB), an ASUS Strix GeForce GTX 970 OC video card, SSD disk Samsung 840 Pro 512GB. Meals on test bench supplied from an Enermax Revolution 850 Watt power supply.
Before measuring the speed of the system, Windows 7 x64 SP1 and all the necessary drivers were installed on it. During the experiment, all tests/measurements were carried out 5 times and the average result was calculated for them. It is these that we will share with you further in the form of graphs reflecting performance in normal mode and with overclocked processor.
As for overclocking our Intel Core i7-4790K: without much difficulty we managed to overclock it to 4800 MHz. The 4800 megahertz bar most likely could have been raised higher, but it all came down to air cooling. It is very likely that the cause of overheating was in the thermal interface under the CPU cover since it was not carried out, and this procedure “removes” from 4790K about ten degrees. There was clearly not enough cooling: without overclocking, the test subject heated up no higher than 93 degrees Celsius, and after increasing the frequency (with a slight increase in voltage), the sensors recorded 100 degrees. Temperature environment at the same time it was constant and amounted to 25 degrees Celsius.
Note that on no other system board It was with this CPU instance that we were unable to achieve stable operation at 4.8 GHz. From this we can conclude that the board is quite suitable to overclock the processor. But again, it’s worth taking care of good cooling. Otherwise, the situation will repeat itself, and you will be limited by insufficient heat removal from the CPU. Due to the fact that at a hundred degrees our Intel Core i7-4790K was already starting the mechanism throttling, then we reduced the frequency multiplier from 48 to 47.
Temperature conditions of transistor assemblies in the processor power circuit from overclocking practically did not change: instead of 48 degrees, they began to warm up to 51. Let us recall that overclocking was carried out with air cooling, and when combined radiators are included in the water cooling circuit, the sensor readings should drop. Operation speed disk subsystem The ASUS Maximus VII Formula motherboard has not undergone any changes due to CPU overclocking.
When in 2006 Intel switched to using the Core microarchitecture in its processors and introduced new design principles that received the self-explanatory nickname “tick-tock,” it was assumed that the company would introduce fundamentally new processors for personal computers to the market annually. However, later this plan underwent certain changes: as it turned out, there is no point in such a high intensity of microarchitecture development. AMD has gradually left the high-end processor segment, and this has allowed Intel to make significant concessions to its original schedule without any risk of losing market share. And now, today no one is particularly surprised by the fact that Haswell processors are going to last as advanced solutions for personal computers for at least eighteen months, and as the most relevant options for desktops - for almost two years.
However, initially nothing like this was envisaged. The Broadwell processor microarchitecture was supposed to replace Haswell in the middle of this year, and life cycle fourth generation Core processors would have a completely normal duration. However, an unpleasant surprise came from the 14nm process technology, which must be implemented for the release of Broadwell. Something went wrong, and the production schedule had to be adjusted, delaying the start of mass production of promising semiconductor crystals by about six months. Now the announcement of mobile energy-efficient versions of Broadwell is expected only on the eve of the New Year, and processors based on this design for mass computers will become available only next year. Moreover, the commonly used Broadwell desktops will appear on the market only in May-June 2015.
And although the shift in announcement schedules for six months to a year does not threaten any special troubles for Intel, the company still considered it necessary to carry out some kind of update of its own platform within the previously established time frame - in the middle of this year. This is a kind of nod to leading partners, who are given the opportunity to refresh their own product lines. This promotion, codenamed Haswell Refresh, includes two components. Firstly, the actual launch of new processor models with the old microarchitecture, but with increased clock speeds, and, secondly, the presentation of new ninth series system logic sets.
The official announcement of new processors and chipsets was scheduled for May 11 - and it has already taken place. At first glance, the Haswell Refresh promotion turned out to be very large-scale. 42 new processors have been added to Intel's price list, 25 of which are aimed at desktop systems of various classes. In addition, three new logic sets have appeared among the company's offerings. However, we will try to find out what actually lies behind the release of such a significant array of new products in this review. For this purpose, our laboratory managed to obtain two older Haswell Refresh desktop processors, Core i7-4790 and Core i5-4690, as well as a motherboard based on the Z97 chipset, ASUS Z97-Deluxe.
⇡ Haswell Refresh processors
Strictly speaking, the Haswell Refresh family of processors is nothing special; in fact, we are talking about a simple increase in clock speeds of the Haswell processors that are well known to us. The only unusual thing here is that a large group of processors with increased frequencies are being introduced to the market simultaneously, as part of a single announcement. Previously, Intel preferred to increase the frequencies of its CPUs separately, without being tied to one date. The logic of the strategy used this time is that, in the absence of real new products, the company wants to find a reason to attract additional attention to its products.
In other words, all the hype around Haswell Refresh is artificial, it is specially generated by Intel itself, trying to create the impression of continuous innovation, even despite the postponement of the Broadwell announcement to a later date. In fact, the release of Haswell Refresh is a completely ordinary update, and the new processors differ from the old Haswell processors, which have been on the market for almost a year, only by increasing the frequency by a ridiculous 100-200 MHz, which occurred within the framework of the old thermal packages. In fact, we are talking about a slight increase in productivity, amounting to about 2-3 percent, and nothing more. Nevertheless, numerous Intel partners will take advantage of the emergence of Haswell Refresh and will undoubtedly offer consumers new models of ready-made systems.
To be fair, it should be noted that the update of Intel processors did not become a reason for their rise in price. Haswell Refresh took the old positions in the price list, displacing Haswell from last year. The full list of new CPUs for desktop systems looks like this:
| Cores/threads | Clock frequency | Turbo frequency | L3 cache | Graphic arts | TDP | Price | |
|---|---|---|---|---|---|---|---|
| Celeron G1840 | 2/2 | 2.8 GHz | - | 2 MB | HD | 53 W | $42 |
| Celeron G1840T | 2/2 | 2.5 GHz | - | 2 MB | HD | 35 W | $42 |
| Celeron G1850 | 2/2 | 2.9 GHz | - | 2 MB | HD | 53 W | $52 |
| Pentium G3240 | 2/2 | 3.1 GHz | - | 3 MB | HD | 53 W | $64 |
| Pentium G3240T | 2/2 | 2.7 GHz | - | 3 MB | HD | 35 W | $64 |
| Pentium G3440 | 2/2 | 3.3 GHz | - | 3 MB | HD | 53 W | $75 |
| Pentium G3440T | 2/2 | 2.8 GHz | - | 3 MB | HD | 35 W | $75 |
| Pentium G3450 | 2/2 | 3.4 GHz | - | 3 MB | HD | 53 W | $86 |
| Core i3-4150 | 2/4 | 3.5 GHz | - | 3 MB | HD 4400 | 54 W | $117 |
| Core i3-4150T | 2/4 | 3.0 GHz | - | 3 MB | HD 4400 | 35 W | $117 |
| Core i3-4350 | 2/4 | 3.6 GHz | - | 4 MB | HD 4600 | 54 W | $138 |
| Core i3-4350T | 2/4 | 3.1 GHz | - | 4 MB | HD 4600 | 35 W | $138 |
| Core i3-4360 | 2/4 | 3.7 GHz | - | 4 MB | HD 4600 | 54 W | $149 |
| Core i5-4460 | 4/4 | 3.2 GHz | 3.4 GHz | 6 MB | HD 4600 | 84 W | $182 |
| Core i5-4460S | 4/4 | 2.9 GHz | 3.4 GHz | 6 MB | HD 4600 | 65 W | $182 |
| Core i5-4590 | 4/4 | 3.3 GHz | 3.7 GHz | 6 MB | HD 4600 | 84 W | $192 |
| Core i5-4590S | 4/4 | 3.0 GHz | 3.7 GHz | 6 MB | HD 4600 | 65 W | $192 |
| Core i5-4590T | 4/4 | 2.0 GHz | 3.0 GHz | 6 MB | HD 4600 | 35 W | $192 |
| Core i5-4690 | 4/4 | 3.5 GHz | 3.9 GHz | 6 MB | HD 4600 | 84 W | $213 |
| Core i5-4690S | 4/4 | 3.2 GHz | 3.9 GHz | 6 MB | HD 4600 | 65 W | $213 |
| Core i5-4690T | 4/4 | 2.5 GHz | 3.5 GHz | 6 MB | HD 4600 | 45 W | $213 |
| Core i7-4785T | 4/8 | 2.2 GHz | 3.2 GHz | 8 MB | HD 4600 | 35 W | $303 |
| Core i7-4790 | 4/8 | 3.6 GHz | 4.0 GHz | 8 MB | HD 4600 | 84 W | $303 |
| Core i7-4790S | 4/8 | 3.2 GHz | 4.0 GHz | 8 MB | HD 4600 | 65 W | $303 |
| Core i7-4790T | 4/8 | 2.7 GHz | 3.9 GHz | 8 MB | HD 4600 | 45 W | $303 |
Unfortunately, in the above list you will not find a single overclocking processor with index K. This means that the Core i7-4770K and Core i5-4670K, released a year ago, remain unaffected by the Haswell Refresh campaign and still remain relevant.
The explanation for this fact is quite simple. Enthusiasts, who usually buy processors with an unlocked multiplier, are unlikely to be interested in a slight increase in the clock frequency stated in the specifications, especially considering that it can be obtained through basic overclocking. The processors included in the Haswell Refresh family do not carry any other changes. They are based on the same C0 stepping semiconductor core as their predecessors and have absolutely no additional advantages. The increase in frequency that has occurred in Haswell Refresh is based only on the maturation of the 22-nm process technology, which was introduced by Intel in 2012. Overclockers need something more.
And there is more in Intel’s immediate plans, it’s called Devil’s Canyon. This code name refers to the new Haswell K-series, which will be presented in the very near future, but will not go on general sale until the fall. These offerings will take longer from Intel to prepare for release, but will be worth the wait. Devil's Canyon plans to make a major change to the processor packaging, which has caused serious criticism at Haswell. The thermal conductive material laid between the processor chip and the heat spreader cover will be replaced with a more efficient one, and the cover itself will begin to be made of a different alloy with better thermal conductivity. In addition, changes will be made to the electrical wiring of the processor chip in order to improve the “purity” of the supply voltages. As a result, representatives of the Devil’s Canyon series, which will be called Core i7-4790K and Core i5-4690K, will become noticeably more overclockable even without performing the “scalping” procedure. In addition, in Devil’s Canyon the nominal clock frequencies will also increase noticeably. For example, in the case of the Core i7-4790K they will reach the 4-GHz mark even without the help of turbo mode. True, the calculated heat dissipation will also increase along the way - it will be not 84, but 88 W.
In the meantime, the most senior processors for the LGA1150 platform - those available for sale - are the regular, non-overclocking Core i7-4790 and Core i5-4690. These CPUs are completely devoid of any overclocking capabilities and do not allow increasing operating frequencies above the nominal values. Even a slight increase in the multiplier, which was possible in representatives of the Sandy Bridge and Ivy Bridge families, is not available. The only characteristics of Haswell Refresh that can be improved by enthusiasts are the memory and graphics core frequencies. In other words, new CPUs are intended to be used only in their normal mode. But in the absence of any other options, we tested just such a pair of older representatives of the Haswell Refresh family.
Detailed characteristics of these models are given in the following table:
| Core i7-4790 | Core i5-4690 | |
|---|---|---|
| Cores/threads | 4/8 | 4/4 |
| Hyper-Threading Technology | Eat | No |
| Clock frequency | 3.6 GHz | 3.5 GHz |
| Maximum frequency in turbo mode | 4.0 GHz | 3.9 GHz |
| TDP | 84 W | 84 W |
| HD Graphics | 4600 | 4600 |
| Graphics core frequency | 1200 MHz | 1200 MHz |
| L3 cache | 8 MB | 6 MB |
| DDR3 support | 1333/1600 | 1333/1600 |
| Technologies vPro/TSX-NI/TXT/VT-d | Eat | Eat |
| Instruction Set Extensions | AVX 2.0 | AVX 2.0 |
| Package | LGA 1150 | LGA 1150 |
| Price | $303 | $213 |
The Core i7-4790 increases the clock frequency of the older line of processors for the LGA1150 platform by 100 MHz, thus surpassing both the overclocker Core i7-4770K and the regular Core i7-4771 by one step. Otherwise, this is a typical Core i7 of the Haswell generation: it has four cores, supports Hyper-Threading, has a spacious 8 MB L3 cache and is compatible with the new AVX2 instructions. The graphics core, like its predecessors, belongs to the GT2 class, that is, it has 20 actuators. It should be noted that thanks to Turbo Boost 2.0 technology, the typical operating frequency for the Core i7-4790 is 3.8 GHz.
The supply voltage of our sample under load was 1.225 V, while in idle state the frequency dropped to 800 MHz and the voltage to 0.717 V.
Security technologies, including vPro, TXT and VT-d, are fully supported by this processor. In other words, the Core i7-4790 is a new flagship for the LGA1150 platform, but without overclocking capabilities.
The Core i5-4690 is a simpler processor: unlike the representatives of the flagship series, it does not have Hyper-Threading technology, the L3 cache is reduced by 2 MB and the clock frequency is slightly lower. Nevertheless, the Core i5-4690 was able to take the position of the older model in the Core i5 series: its frequency is 100 MHz higher than that of the Core i5-4670, ranging from 3.5 to 3.9 GHz. Thanks to turbo mode, the most typical operating frequency of the Core i5-4690 becomes 3.7 GHz - this is only 100 MHz less than that of the Core i7-4790.
The Core i5-4690 voltage under load was 1.195 V, but at idle it, like its older brother, reduced the frequency to 800 MHz and the voltage to 0.718 V.
The graphics core in the Core i5-4690 is exactly the same as that of the Haswell Refresh Core i7 series processor; there are no additional restrictions in either the supported instruction sets or security technologies.
The new Haswell Refresh processors do not impose any special requirements on motherboards. They are compatible with any LGA1150 platforms, including those that were released a year ago, at the time of the announcement of conventional Haswell processors. But, despite this, simultaneously with the release of new processors, Intel also released new chipsets for LGA1150 systems - Z97 and H97. They should be discussed separately.
⇡ Chipsets of the ninth series
In general, the story with the announcement of the Z97 and H97 chipsets is approximately the same as with Haswell Refresh. These chipsets do not bring any noticeable improvements, and are not needed to accompany new processors. The only more or less logical explanation for their appearance is preliminary preparation of the ground for the next generation of processors, Broadwell, which are expected to be only partially compatible with the LGA1150 ecosystem.
According to available information, Broadwell processors are Additional requirements to the power system implemented on the motherboard. Therefore, they most likely will not be compatible with LGA1150 boards released last year. The emergence of new and more modern logic sets of the ninth series should give impetus to the development and release of LGA1150 boards of the second stage, which should already have Brodawell compatible without any reservations. Therefore, Intel calls compatibility with future CPUs one of the main properties of the Z97 and H97 chipsets: now it is included at the level of the reference design of the motherboard power converter. It is not surprising that the first lines of the characteristics of the new chipsets include “support for 4th and 5th generation Intel Core processors,” while the 8th series chipsets are formally compatible only with Haswell.
There is one more nuance regarding processor support. For some reason, when talking about its Devil’s Canyon overclocking chips, Intel points to their performance in boards with the Z97 chipset and remains silent regarding compatibility with the Z87. It is quite possible that this is a marketing gimmick, but it is not yet possible to rule out that the new LGA1150 boards will be able to work with the Core i7-4790K and Core i5-4690K better than the old ones in some aspects.
If you close your eyes to the situation with processor support that is not yet completely clear, then the new chipsets look like a simple evolutionary update of the Z87 and H87. At the same time, the main advantage of the Z97 and H97 is the emergence of new opportunities related to the construction of the disk subsystem. In particular, these logic sets take the first step towards the introduction of promising interfaces for connecting solid state drives- SATA Express and M.2.
At the same time, the characteristics of the Z97 and H97 are almost the same, the difference between them lies only in their positioning. The Z97 traditionally targets enthusiasts and overclockers, supporting CPU overclocking and multi-GPU configurations. H97 is a more conservative option, which is cheaper, but does not allow division of PCI Express processor lines, does not allow overclocking of the processor, but supports the Intel Small Business Advantage administration system.
The main innovation in the ninth series chipsets is Rapid Storage technology version 13. In the new version, this technology becomes compatible with drives connected not only via the SATA interface, but also via the PCI Express bus. Simply put, this means that the Intel disk driver will now see both AHCI and NVMe PCI Express devices, allowing almost all the functions built into Rapid Storage Technology to be applied to them. For example, they can be made bootable disks, hybrid configurations based on Smart Response Technology (SRT) can be created with their participation, and the like. The only limitation is drives with PCI interface Express cannot be included in RAID arrays.
As for the RAID modes themselves, for SATA drives almost everything remains the same. The chipsets support arrays of levels 0, 1, 10 and 5, while in an array with mirroring of two or four solid-state drives (level 0), the TRIM command can be run for all members included in it.
Another improvement in Rapid Storage Technology 13 is the introduction of support for configurations with 16 GB of RAM in Intel Rapid Start Technology. Let us remind you that this technology allows you to save a memory dump on the SSD when the system goes into a state of deep sleep - so that the process of “waking up” occurs faster, and during sleep itself the system does not require power to maintain the state of the RAM. Previously, this technology was only allowed to work on systems with 8 GB of memory, but now the restriction has been lifted.
Otherwise, no significant innovations are visible. The Z97 and H97, like their predecessors, support up to six SATA 6 Gb/s interfaces, up to six USB 3.0 ports and up to eight PCI Express 2.0 lanes. At the same time, obviously, a pair of PCI Express chipset lines can be used not only to connect additional controllers or implement expansion slots, but also to install one M.2 slot or one SATA Express port on boards. This means that interfaces for connecting new generation drives, implemented on the basis of ninth series chipsets, can provide maximum throughput about 1 GB/s, which is 67 percent more than the usual SATA 6 Gbit/s ports.
However, the implementation scheme for all high-speed ports in the new chipsets remains as unclear as it was before. It again obeys the proprietary Flex IO scheme, but now another unknown has been added to the equation - an M.2 slot or a SATA Express port. The bottom line is that in total Z97 and H97 have 18 channels each for implementing high-speed ports. Four channels are for USB 3.0 ports, four for SATA 6 Gb/s and six for PCI Express 2.0 lines. The remaining four channels do not have strictly fixed functionality: two of them can be used for both PCI Express and USB 3.0, and the second pair can play the role of either PCI Express or SATA 6 Gb/s. Therefore, motherboard manufacturers are forced to choose between different configurations when designing their products. For example, if the board has six chipset USB 3.0 and six SATA 6 Gb/s, then you can only get a maximum of six chipset PCI Express 2.0 lines. In addition, there is one more limitation: the total number of PCI Express lanes should not exceed eight.
Now M.2 or SATA Express interfaces are also being added to the described scheme, which Intel is talking about supporting in new chipsets. As you know, to implement the M.2 slot that is currently available on this moment Type M requires one SATA port and at least two PCI Express lanes. The SATA Express interface requires two SATA ports and at least two PCI Express lines. At first glance, adding such interfaces to a motherboard with a chipset that has only eighteen channels for high-speed ports should significantly deplete its expandability. But, fortunately, in M.2 and SATA Express SATA interfaces and PCI Express are never used at the same time. Therefore, Intel decided to give them channels numbered 13 and 14, the same ones that can switch their functionality between SATA and PCI Express. Of course, board manufacturers are free to use other SATA ports and PCI Express lines for M.2 and SATA Express, or even resort to external controllers. But support for Rapid Storage Technology 13 for drives connected via the PCI Express bus is only possible if they operate through the 13th and 14th channels of the chipset. And this means that the presence on a board based on a ninth-generation chipset of a PCI Express drive, which is connected in accordance with Intel’s scheme, reduces the number of working SATA 6 Gb/s ports to four.
Despite the fact that there are not many new features in the ninth series logic sets, motherboard manufacturers have received their announcement with great enthusiasm. All leading brands have completely updated their lines of LGA1150 motherboards, but most of the new motherboards do not offer any fundamentally new features. Nevertheless, 3DNews is monitoring the emergence of really interesting new products - a couple of reviews have already been published, plus we will continue to publish them on our website.
Just a month and a half ago, Intel updated its line of processors for personal computers. A large-scale improvement in specifications has affected almost the entire Haswell family, including offerings for desktop systems in LGA1150 packaging. However, processors for enthusiasts with the letter K in their designation were not affected by the Haswell Refresh campaign. The point here is not at all that Intel has decided once again to leave the elite part of the computer community without its attention and force it to continue to be content with the Core i7-4770K and Core i5-4670K models. On the contrary, the company took an additional pause to prepare improved overclocking modifications with completely unique properties – Devil’s Canyon. And now, finally, these flagship CPUs for the LGA1150 platform are ready to hit the market.
We can only welcome the appearance of Devil’s Canyon processors, which support overclocking and are intended for high-performance desktop systems. Over the past few years, the desktop market has been under serious pressure from the mobile segment. Sales of desktop systems are declining, and this entails a shift in the interests of developers from CPUs for computers of traditional form factors to cost-effective and energy-efficient architectures used in compact laptops, tablets, smartphones and all kinds of hybrid devices. For example, last time we wrote about overclocking products Intel last fall, and at that time we were talking about Ivy Bridge-E processors, which, frankly speaking, did not carry anything new and offered only an incremental modification of the old LGA2011 platform. The Haswell processors released a year ago also do not seem like a breakthrough; they did not significantly increase their performance compared to their predecessors in the Ivy Bridge series and, moreover, received significantly worse overclocking potential. In fact, the latest Intel product, which is remembered with warm words by all fans of desktop systems without exception, remains Sandy Bridge. An excellent illustration of the lack of attractive offers for advanced desktop users since the beginning of 2011 is the fact that many of them still continue to use platforms based on second-generation Core processors and are not going to invest in their modernization.
Fortunately, however, Intel has not forgotten that there is a layer of desktop enthusiasts who want to stay on the cutting edge of technology. Not long ago, Intel came to the understanding that this audience consists almost entirely of its either ideological or forced fans, who one way or another require special attention. Therefore, Intel has now decided to increase its emphasis on products for overclockers, as a result of which Devil’s Canyon appeared.
It is curious that such an unexpected turn towards advanced desktop users occurred at Intel quite recently - in December last year. It was then that the Devil’s Canyon project was given the green light by the company itself. upper level. The developers were given a rather difficult task: they had to create a fundamentally new overclocking processor in a short time, which would have significantly higher clock speeds and better overclocking potential, but would not lose compatibility with the existing ecosystem and would not go beyond the usual thermal package. And this task was solved: the Devil’s Canyon class processors presented earlier this month meet all the stated conditions. Of course, they are not perfect, as the developers had to make numerous compromises to speed the product to market, but it should be understood that Devil's Canyon is just a harbinger of more interesting overclocking products that will continue to be released throughout the coming fall and winter.
However, let's not be distracted from the main topic of this review and see what Intel intends to please enthusiasts with right now. We have the eldest of Devil’s Canyon in our hands – Core i7-4790K.
More about Core i7-4790K
In short, Devil’s Canyon processors are Haswell Refresh for overclockers. They do not have any improvements at the microarchitecture level, and their superiority over their predecessors is due only to such engineering solutions that could be implemented without delving into the inside of the processor. But while Haswell Refresh differs from Haswell of the “first wave” only by slightly increased clock frequencies, there are three fundamental differences between Devil’s Canyon and previous overclocker Haswells.
Increase
Firstly, these are, of course, clock frequencies. If we talk specifically about the Core i7-4790K model, then the increase in its frequency looks very impressive. This processor is designed to operate at 4 GHz, and in turbo mode it can accelerate to 4.4 GHz. Thus, thanks to Devil’s Canyon, Intel’s line of desktop processors now includes products that have confidently surpassed the 4-GHz mark. At the same time, the calculated heat dissipation of the Core i7-4790K has hardly increased: its thermal package is set at 88 W, which is only 4 W higher than the TDP of older Haswell processors that have existed so far. This change in thermal characteristics will be almost imperceptible to end users: the Core i7-4790K remains fully compatible with the existing fleet of LGA1150 motherboards and cooling systems in the mid-to-high price category.
Secondly, in Devil’s Canyon processors, the manufacturer changed the composition of the heat-conducting material located between the semiconductor crystal and the processor cover. This step is a direct response to criticism from overclockers who expressed dissatisfaction with the poor thermal conductivity of the internal thermal interface of older Haswells, due to which the processor cores overheated even in systems with very efficient cooling systems. However, the joy from the change that has occurred will still not be complete. The fact is that the most effective method of mating the crystal and the cover - through flux-free soldering - has not returned to Devil’s Canyon. In fact, Intel simply replaced one type of thermal paste with another. The new material, called NGPTIM (Next-Generation Polymer Thermal Interface Material) in the documentation, is distinguished by its chemical composition and has better thermal conductivity than before. However, it is obvious that to achieve maximum overclocking results, radical enthusiasts will have to “scalp” Devil’s Canyon.
Thirdly, in Devil’s Canyon some changes were made to the processor power system. In order to improve the stability of supply voltages, Intel has developed a new electrical wiring diagram, which is reflected in the appearance of additional capacitors on the “belly” of the processor.
With all this, the Haswell semiconductor core used in Devil’s Canyon remains unchanged. The new overclocking processors are based on the same C0 stepping crystal, which is used in both Haswell Refresh and regular Haswells from a year ago. Therefore, all improvements in frequency potential are solely the result of changing the internal thermal interface and improving the power supply circuit.
Thus, the Core i7-4790K remains a completely ordinary LGA1150 processor in the upper price range with four processor cores and support for Hyper-Threading technology, with 8 MB of third-level cache, support for dual-channel DDR3 memory, with a built-in GT2-class graphics accelerator and PCI Express 3.0 controller for 16 lanes. In comparison with the previous overclocking flagship, Core i7-4770K, and with the older Haswell Refresh, Core i7-4790, its characteristics look like this:
| Codename | Devil's Canyon | Haswell Refresh | Haswell |
| Cores/threads | 4/8 | 4/8 | 4/8 |
| Hyper-Threading Technology | Eat | Eat | Eat |
| Clock frequency | 4.0 GHz | 3.6 GHz | 3.5 GHz |
| Maximum frequency in turbo mode | 4.4 GHz | 4.0 GHz | 3.9 GHz |
| Unlocked multiplier | Eat | No | Eat |
| TDP | 88 W | 84 W | 84 W |
| HD Graphics | 4600 | 4600 | 4600 |
| Graphics core frequency | 1250 MHz | 1200 MHz | 1250 MHz |
| L3 cache | 8 MB | 8 MB | 8 MB |
| DDR3 support | 1333/1600 | 1333/1600 | 1333/1600 |
| Technologies vPro/TSX-NI/TXT/VT-d | Eat | Eat | No |
| Instruction Set Extensions | AVX 2.0 | AVX 2.0 | AVX 2.0 |
| Package | LGA1150 | LGA1150 | LGA1150 |
| Price (Boxed/OEM) | $350/$339 | $312/$303 | $350/$339 |
Since the Core i7-4790K does not have any changes at the microarchitectural level, it is compatible with any LGA1150 motherboards, both the new, ninth series, and the old, eighth. But keep in mind: older boards may require a firmware update.
The Core i7-4790K processor we received for testing functioned in full accordance with the stated specifications - at a frequency of over 4.0 GHz. With Turbo Boost 2.0 technology activated, with a multi-threaded load, its frequency was 4.2 GHz, but in the case of single-threaded activity it increased to the promised 4.4 GHz. The operating voltages were in the range of 1.19-1.29 V, which, it should be noted, is higher than the voltages that are characteristic of both conventional Haswell and new Haswell Refresh modifications.
It should be emphasized that in the Core i7-4790K, unlike the Core i7-4770K, Intel decided not to disable vPro and TXT security technologies, VT-d virtualization and TSX-NI instructions. This makes Devil’s Canyon interesting also in the corporate sector, where users may be tempted by significantly higher clock speeds than the regular Core i7-4790.
In addition to what has been said, it only remains to add that Devil’s Canyon is not only a Core i7-4790K. This family includes two more representatives: a quad-core processor without Hyper-Threading support, Core i5-4670K, designed for frequencies of 3.5-3.9 GHz; and also very interesting, thanks to its tempting price, the dual-core Pentium G3258 - the first budget Intel processor aimed at overclocking. We'll get to know these new products a little later, but for now let's look at the features of the Core i7-4790K in practice.
Temperature test
Since in terms of architecture, Devil’s Canyon is absolutely similar to Haswell, the main question is: how much could replacing the thermal interface under the processor cover improve heat dissipation from the processor chip?
To answer this, we conducted a simple experiment - we compared the operating temperatures of the fresh Core i7-4790K and the old Core i7-4770K when operating at the same clock frequency with the same supply voltage. In particular, at 4.2 GHz, which for the Core i7-4790K is a typical operating frequency, but for the Core i7-4770K is already quite a noticeable overclocking. The voltage of both CPUs was set at 1.2 V, Turbo Boost 2.0 technology was disabled. To remove heat, our test system used the new Noctua NH-D15 supercooler installed on the processor using Arctic Cooling MX-2 thermal paste. The load was created by the LinX 0.6.5 utility, which supports the AVX2 instruction set.
The following graph displays the instantaneous temperatures observed during the test, measured every 2 seconds.
In fact, no miracle happened. By “scalping” Haswell and replacing the standard thermal paste with liquid metal, you can gain back more than two dozen degrees under high load. There is nothing like this in Devil's Canyon. It seems that the new thermal interface material is not very different from the old one, so the difference is Core temperatures i7-4790K and Core i7-4770K under the same conditions is no more than 7-10 degrees. Of course, this will be enough to slightly improve the overclocking potential, but one should clearly not expect a serious effect from the implementation of the notorious NGPTIM.
By the way, the difference between the temperatures of processors released with an interval of a year could well be due to a decrease in heat generation that occurred due to the aging of the 22-nm process technology used by Intel to produce its CPUs. However, in this case the influence of this factor is minimal. This is what the power consumption of both processors measured in parallel with the temperature looks like:
As you can see, the consumption of the Core i7-4790K and Core i7-4770K at the same frequency and at the same voltage is almost the same. Both processors are based on semiconductor crystals of the same C0 stepping, and this is reflected in the graph above. In other words, there are no improvements in heat dissipation for Devil’s Canyon. Extreme heating of Haswell processors is a problem whose roots go back to the microarchitecture, and new CPU modifications are no better than the old ones in this regard. It's no joke, with a high AVX2 load, Haswell's consumption at 4.2 GHz exceeds 150 W. Of course, the LinX utility, based on the Linpack package, squeezes all the juice out of the processor's computing resources, in which it is not very similar to real applications. However, the Core i7-4790K needs high-quality cooling even when operating in nominal mode. Otherwise, owners of this Devil’s Canyon variety will have to deal with, if not instability, then at least temperature throttling.
Overclocking
Devil's Canyon is aimed at the enthusiast audience by Intel. Therefore, all processors included in this series, including the Core i7-4790K, have unlocked multipliers that allow you to independently overclock the processor, bus, memory and graphics core. However, there is nothing new in this list, and in terms of overclocking functions, the Core i7-4790K is similar to its K-series predecessor, the Core i7-4770K.
Past Haswells with an unlocked multiplier were somewhat disdainful among overclockers. This is due to high heating, due to which the maximum frequency suitable for everyday use of the processor did not exceed 4.3-4.5 GHz. Considering the fact that Sandy Bridge generation processors from three years ago were often able to reach frequencies above 4.8 GHz, this looked like a definite step back.
But Devil’s Canyon is positioned by Intel as “working on mistakes” primarily in terms of heat removal from the processor chip. That is, the manufacturer directly promises that new CPUs should heat up less and, due to this, conquer higher frequencies without any surgical intervention (scalping) or replacing the internal thermal interface. However, we had certain doubts about the significance of the improvements that had occurred from the very beginning. Firstly, even when operating at the nominal frequency, the Core i7-4790K uses very high voltages, reaching 1.29 V in turbo mode. At the same time, even Intel itself did not previously recommend installing voltages above 1.3 V on Haswell for long-term use in order to avoid degradation of the semiconductor crystal. So there is little room left for manipulating voltage during overclocking. Secondly, the power consumption indicators we measured for Devil’s Canyon clearly indicate that the problem with heat dissipation for these processors remains very critical, and replacing one polymer thermal interface with another, as practice shows, does not give a very noticeable gain in temperatures.
Actually, all these fears were confirmed in practical experiments. The maximum frequency at which we managed (and with considerable difficulty) to achieve stable performance of the Core i7-4790K when testing it in LinX 0.6.5 turned out to be 4.5 GHz. To ensure that there were no failures in the test, we had to set the voltage to 1.25 V, but the maximum temperature almost reached the critical limit and reached 97 degrees, even despite the use of a very efficient Noctua NH-D15 cooler. Let us remind you that temperature throttling for representatives of the Devil’s Canyon family, as well as for regular Haswells, turns on at 100 degrees.
In fact, this is a very sad result. It turns out that the Core i7-4790K is not much better at overclocking than its overclocking predecessor, the Core i7-4770K. Simply changing the chemical composition of the polymer thermal interface is not enough to make Haswell a new Sandy Bridge, and we can only hope that flux-free soldering of the semiconductor chip to the processor cover with low-melting indium solder will return in future LGA1150 products for Broadwell generation enthusiasts.
It must be said that our result of overclocking the Core i7-4790K - 4.5 GHz - may seem too low given the fact that many foreign sources indicate the possibility of overclocking this processor to 4.6-4.7 GHz or even higher. However, there is nothing strange about this - the LinX 0.6.5 utility we use to check system stability, due to the active use of AVX2 instructions, creates a very high load that seriously warms up the processor. If, for example, we take for testing the previous version of this utility (0.6.4), which uses AVX commands instead of AVX2, then the performance of our Core i7-4790K copy can be achieved at a frequency of 4.6 GHz with a supply voltage of 1.275 V. C With these settings, the processor is able to pass all of our performance tests and shows no signs of instability in general use.
However, be that as it may, all Haswell’s problems that appear during overclocking have moved to Devil’s Canyon. When a load occurs, the temperature of the processor cores increases sharply and significantly, and keeping it within acceptable limits requires considerable effort. As before, increasing the performance of the cooling system does not help too much, and it is quite obvious that the thermal interface material located under the cover remains the bottleneck in heat removal from the processor chip.
Loading...
