Overclocking engineering samples of Intel Skylake processors or greetings from China. Theory and practice of overclocking Intel Skylake processors via the BCLK bus Overclocking and testing

On the Internet you can find an engineer for sale (Intel Confidential): a quad-core processor of the sixth generation Core i7-6400T for the LGA 1151 platform, built on the Intel Skylake architecture. It runs on the bus (BCLK), without losing the integrated graphics (read Intel Quick Sync Video) and AVX/AVX2 instruction sets, and the thermal sensors are not disabled.
When overclocking, there are two subtleties: Enhanced Intel SpeedStep technology must be forcibly disabled, and the Boot Performance Mode parameter must be set to Turbo Performance.
Throttling in engineering Skylake is activated at 100 degrees, just like in regular Core i7 processors.
The processor is equipped with a more efficient thermal interface: NGPTIM polymer paste thermal interface. Therefore, the best overclocking of the Core i7-6400T can be achieved using scalping.
Processor specifications:
Supported connectors: FCLGA1151.
4 cores and 8 logical threads (NT).
Processor clock frequency: 2.2 GHz.
Maximum clock speed with Turbo Boost 2.0 technology: 2.6 GHz (2.4 GHz for all cores).
Intel Smart Cache technology - 8 MB (third level cache).
Core to bus ratio: 26.
DMI (Direct Media Interface) 3 bus operates at a frequency of 8 GT/s (versus previously 5 GT/s).
Command set: 64-bit.
Instruction set extensions: SSE 4.1/4.2, AVX (Advanced Vector Extensions) 2.0, FMA3 and TSX.
Technologies supported: Turbo Boost, vPro, Virtualization (VT-x), Intel Virtualization Technology for Directed I/O (VT-d), Intel Trusted Execution, Intel Anti-Theft, Fast Memory Access, Flex Memory Access.
New AES commands, Intel 64 architecture, Intel VT-x with Extended Page Tables (EPT).
Advanced Intel SpeedStep Technology.
Idle states, thermal control technologies.
Execute Disable Bit function.
Lithography: 14nm.
Maximum Design Power (TDP) = 65W.
Maximum memory capacity (depending on memory type): 64GB.
Memory types: DDR3L-1333, DDR3L-1600, DDR4-1866, DDR4-2133.
Built-in dual-channel memory controller.
Maximum memory bandwidth: 34.1 GB/s.
Graphics system built into the processor: Intel HD Graphics 530 Gen 9 (GT2). 24 execution units (versus 20 for i5 with HD Graphics 4600 core). Performance: 844 GFLOPS. Peak Pixel Fill Rate: 5.4 GPixels/s. Peak Texel Rate: 10.8 GTexels/s. Peak Polygon Rate: 675 MPolys/s.
Base frequency of the graphics system: 350 MHz. Maximum dynamic frequency of the graphics system: 1.15 GHz.
Maximum Dynamic Graphics Frequency - The maximum rendering frequency (MHz) supported by Intel HD Graphics with Dynamic Frequency.
Supports DirectX 12, OpenCL 2.0 and OpenGL 4.4.
Technology support: Intel Quick Sync Video, InTru 3D, Intel Insider, Intel Wireless Display, Intel Clear Video HD.
Intel Quick Sync Video - supports video encoding in the following formats: AVC/H.264 (Blu-ray), MPEG2 (DVD), MVC HW (Stereo 3D), JPEG/MJPEG. Transcoding video in 4K HEVC/H.265 format.
Maximum supported GPU memory = 1.7 GB.
Intel Flexible Display Interface (Intel FDI).
Number of supported displays: 3.
PCI Express edition: 3.0, speed: 8 billion transfers per second.
PCI Express configurations: 1x16, 2x8, 1x8 & 2x4.
Cooling system specifications: PCG 2015C (65W).
Case temperature: 72.7C.
Case size: 37.5 x 37.5 mm.

There are processors on sale with markings (S-Spec): QHQG, QHQF, QHQJ and QH8G.
L452, L448, C445, L501, information about what week, what year the processor was released.

News from the site: www.intel.com.

As you understand from the cat, these processors are Engineering versions of full-featured INTEL processors, which, apparently, were used for testing purposes by motherboard manufacturers and are now being gradually stolen and leaked by the cunning Chinese.

Let me remind you right away: Engineering versions of processors are the property of Intel, and after testing must be sent to the company for recycling. The distribution and use of processors is a violation of the company's rights and is punishable by law.
(Z.Y. in fact, Intel does not always destroy them, and after proper testing, the most extreme versions by production date are sold to assemblers of finished computers, such as LENOVO, HP and DELL)

The characteristics, in a sense, are a lottery; the frequency overclocking potential may differ from stone to stone, but by accepting such features, you can get the latest generation processor two or three times cheaper than in Russian stores! It has everything: 4 cores/8 threads (thanks to Hyper-threading), 8 megabytes of L3 cache, support for all the latest instructions, etc. The only difference initially (for the processor under review) is a reduced multiplier and a correspondingly lower clock frequency.

Let me clarify that these processors were called 6400T by the Chinese and in fact there are no restrictions inherent in T-versions of Intel processors (like the even more limited Tjmax), they don't have it.

I will also briefly indicate the differences in the ES models of these processors known to me (from the Internet) and their stepping:

QH8F is the earliest stepping, does not work with an external card, only built-in, in principle, can be used as an option for office assemblies and HTPC, due to the low heat dissipation and quite powerful graphics. processor.

QHQJ is another HTPC candidate, clocked at 1.6 GHz and TDP 35w, coupled with the latest generation of integrated graphics from Intel, HD530, looks very tempting. If only the price were even lower. 🙂

The QHQG reviewed in the review, also known as the i7-6400T, is the most likely candidate for the “folk” stones. Works with an external video card, multiplier 8-24, turbo boost 26, overclocked on the bus from 3.5 to 4.5 GHz. Full support for all instructions. Stepping is desirable no less than l448.

QHQF - the last revision before sending “for gold”. The multiplier is 8-40, i.e., in fact, this is a full-fledged i7-6700. No problems noticed

Also briefly about the notorious decoding of the stepping name:

Okay, let's finish with the lengthy preamble and move on to the subject of the review.
The lot was ordered on 11/07/16 and shipped the next day. The seller sent SPSR by mail, as I requested, the package took 10 days, and was given to me by courier around lunchtime on 11/18/16.

For fans of tracks and other things

The product was packed in a cardboard box, the processor itself was placed in a plastic blister and wrapped in bubble wrap, the seller put a sachet with cheap thermal paste “for 1 use.” I think the seller packed everything quite well.

Packaging photos

Here it is, handsome (already installed in the socket):

The seller sent me a newer version of the processor than shown in the photo. After thinking and googling, I realized that the current situation is more to my advantage, because the newer the version, the less chance there is of performance problems.

An ASROCK Z170 PRO4 board was purchased for the processor, as the most successful option in terms of price/quality ratio.

Well, let’s apply thermal paste (not the one included in the kit, of course! 🙂 and preferably something better than KPT-8), install the cooler, pinch our fingers and start the PC.
Everything was determined the first time, and the computer will be able to show us everything about the processor.

Processor idle and under load:

Everything doesn’t look as rosy as we would like (although even in stock it is more powerful than the i3-6100, which is sold in Russian stores for the same amount). But all the pulp - in acceleration!

Additional information about the situation with overclocking processors

Those who follow events on the “iron front” are aware that starting with the Sandy Bridge generation processors, Intel eliminated bus overclocking in hardware, which was the bread and butter of overclockers around the world, leaving only the possibility of overclocking with a multiplier in expensive “K” versions of their products. Moreover, with the release of the third generation Core, the situation worsened. Instead of the solder used in Sandy Bridge, Intel began adding thermal paste of very mediocre quality under the cover of Ivy Bridge processors, and other generations of processors adopted this bad innovation.
Now, with the advent of the Skylake generation, Intel engineers have separated the PCI Express bus and chipset into a separate domain, the frequency of which remains fixed, regardless of changes in BCLK, which was the main problem of overclocking on the bus. But anyway, the engineers added the so-called BCLK Governor - a mechanism that limits changes in BCLK. As it turned out, it was implemented at the software level, which theoretically provided loopholes to bypass it.
What finally happened! The first company to officially implement the loophole was the aforementioned ASROCK, rolling out the SKY OC function - and opening up the opportunity to overclock any processor of the SKYLAKE generation, but with one limitation - it works (as it was at the beginning) only on Z170 chipsets. Then this feature was picked up by ASUS, MSI and some others.
Intel, of course, didn’t like this, and threatened to demand that these functions be removed from the BIOS, and also promised to take a closer look at limiting BCLK overclocking in the new generation of KabyLake. But it’s too late - the bios have leaked onto the Internet and can be easily downloaded from there. Moreover, the craftsmen from Korea were able to correct some shortcomings when overclocking on the bus (such as the dump of some AVX instructions and temperature sensors).
Maud. BIOS beavis and overclocking instructions are taken, for example, here: overclocking.guide/category/intel-oc-guides/skylake-non-k-oc/
In general, to achieve full parity with the I7-6700, we need to help the processor by overclocking via BCLK. To do this, you need a modified BIOS (on ASUS boards, new BIOSes support everything out of the box, and you can download them on the website) corresponding to your device and a little time. Also, after flashing, do not forget to reset the BIOS with a jumper, this is highly recommended.
There are many videos on the Internet and YouTube with instructions on overclocking.
If they ask in the comments, I’ll post my BIOS settings, which I have tested and are quite successful.

As a result, we get: disadvantages of overclocking:
It's not so scary
1. We are losing all Intel power saving technologies, including frequency and voltage reset when idle. This feature is preserved for overclocking above ~3.5 GHz; if you don’t push it higher, the technologies remain functional and delight the owners. I want to emphasize that without a load on the processor, it cools down quite quickly and returns to idle temperatures even without these technologies, they are aimed only at reducing energy consumption, which is not as important here in Russia than in Europe.
2. Increased need for heat dissipation from the processor; according to my estimates, TDP has increased under load from 65w to 100-125w. Which is easily solved with an entry-level tower for 700-1000 rubles from Avito. Well, it’s still a top-end processor, after all, and yet it’s smaller than some 8-core processors from AMD.

As a result: We got a processor similar in power to the top-end stone from Intel, with support for everything and everyone, paying 2.5 times less!
Of course, I cannot recommend this option to everyone, because there is no guarantee compared to the store. But for those who are not afraid to take risks and who have a little knowledge of hardware, I highly recommend this option for an upgrade. But make no mistake, take the QHQG stepping l501.

This is my first review, I will accept adequate criticism and requests for points of interest to you regarding the subject. If everyone likes everything, I’ll write a similar review about the Intel X3440 processor, which will breathe life into socket 1156 (opens video cards up to gtx1060), and can become a good basis for a budget PC gaming platform.
Thank you for your attention.

The progress that Intel processors undergo when changing generations of microarchitecture has recently slowed down noticeably. Indeed, if we compare similarly positioned processors from the next few years of release, it turns out that their computing performance differs by 3-7 percent at best, and this despite the fact that developers constantly talk about huge steps (“ticks” and “ takah") in the development of microarchitecture. Therefore, it is not surprising that many owners of systems even five years ago simply do not see the point in upgrading their computers and continue to stay with Sandy Bridge, sincerely believing that in the absence of real competition, the development of desktop processors has seriously slowed down.

However, in reality everything is far from so clear. We will not argue with the thesis about the absence of a noticeable increase in productivity. But it’s impossible to agree with the fact that nothing interesting has been happening in the development of desktop systems for a long time. And we are not talking here at all about the graphics core built into processors and not about the introduction of new interfaces into modern platforms, but about the fact that over the past couple of years Intel has come a long way towards returning overclocking to its original meaning.

With the transfer of processors to the Core architecture, Intel tried to change the idea of overclocking processors, making it not a means of saving money, but, on the contrary, a very expensive sport. And to some extent, Intel (with the active assistance of the rest of the industry) succeeded in this: there are noticeably fewer processors capable of overclocking, and they have become noticeably more expensive. However, mass practical overclocking, fortunately, has not disappeared. It was partly fueled by the efforts of AMD, and sometimes some gifts (for example, Pentium Anniversary Edition) were presented by Intel itself. But the most important event for the overclocking renaissance happened quite recently - with the advent of the LGA1151 platform and sixth generation Core processors.

The fact is that, as it suddenly turned out, in the LGA1151 ecosystem any processors can be overclocked: both models specially designed for this with the index K in the name, and all other Skylake with a locked multiplier. Yes, overclocking ordinary processors is not as straightforward as with expensive K-versions. It is performed not by changing the multiplier, but by increasing the frequency of the base clock generator. Therefore, not all motherboards are suitable for such overclocking; special BIOS versions are needed, and, in addition, AVX/AVX2 instructions are disabled as a side effect. Nevertheless, the results are quite positive: in a number of applications - primarily gaming - you can get a noticeable increase in performance.

Of course, we can say that Intel has no direct connection to the revival of mass overclocking, because the company still does not give free rein, but is trying to introduce restrictions on the overclocking of Skylake models without the K index through organizational measures. But firstly, it is not particularly successful in combating overclocking of mass-produced processors, and open loopholes still remain to this day. And secondly, who, if not Intel, laid the hardware foundation for such overclocking? This whole story became a reality solely due to the fact that in the LGA1151 platform it became possible to use two clock generators at once: one for the processor frequency, and the second for generating interface frequencies. Before the advent of Skylake, such frequency division was not included in the design of platforms, and therefore it was simply impossible to overclock processors in any way other than by changing their multipliers. In other words, Intel is directly involved in the return of overclocking for low-cost processors, and it would be unfair to deny its role in this.

But the most convincing evidence that Intel was seriously thinking about the possibility of opening up complete overclocking freedom in Skylake came recently from China. Pre-release processors of the Skylake generation, which are freely overclocked by the clock generator frequency without any tricks or restrictions, have begun to be actively sold on numerous Chinese trading platforms. That is, the overclocking block for non-K processors was added by the microprocessor giant immediately before Skylake was launched on the market, and it is not surprising that such artificial restrictions were eventually circumvented.

This whole story is not only about the fact that Intel wanted to add overclocking to Skylake, and then changed its mind, but cunning motherboard manufacturers were able to implement Intel’s original plan without the participation of the microprocessor giant. It adds another important touch to the already known picture of Skylake overclocking: in nature there are pre-release LGA1151 processors that, by increasing the clock generator frequency, can be overclocked “out of the box”, without any additional conditions. And this fact has enormous practical value, since these processors can be obtained if desired, and at a very attractive price.

Actually, this article will be devoted to the story of testing such an easily overclockable pre-release processor. Recently, computer forums have been filled with an avalanche of enthusiastic reviews from users who purchase low-frequency engineering samples of quad-core Skylake on Chinese Internet sites and raise their frequency to the level of Core i7-6700K.

It’s not very easy to believe that this is possible, since the Chinese sell such processors at about two to two and a half times cheaper than a full-fledged modern quad-core Core i7 level. Therefore, we decided to figure it out ourselves and ordered from Aliexpress one of the processors with the mysterious and non-existent name, in Intel’s understanding, Core i7-6400T, for which sellers from the Middle Kingdom promise the ability to easily overclock and operate at a frequency of about 4.0 GHz.

Since the multiplier of the vast majority of engineering processors is locked, they have to be overclocked at the base frequency. In the case of Skylake, this is quite possible on motherboards that have an external clock generator, and this is almost any board based on the Intel Z170 chipset. If we were talking about overclocking serial processors, then success would also require a special, tweaked BIOS. But to overclock pre-production samples of Skylake steppings A0 and Q0, it is not needed - such processors can be freely overclocked in frequency on regular firmware versions. This means that raising the frequency of the Core i7-6400T is not so difficult. And even more than that, engineering versions of Skylake stepping Q0 compare favorably with their later production counterparts in that when overclocking the clock generator frequency, thermal sensors and AVX/AVX2 instructions are not disabled in them. That is, overclocking is even more complete.

When overclocking Skylake in frequency, there are two subtleties: Enhanced Intel SpeedStep technology must be forcibly disabled, and the Boot Performance Mode parameter must be set to Turbo Performance. These settings in most cases eliminate problems with cold system starts. Otherwise, the algorithm is very simple: the processor multiplier is fixed at the maximum allowable value, after which the frequency of the base clock generator BCLK is increased and, if necessary, the processor voltage is added to ensure stability. You just need to remember that the memory operating frequency is also related to BCLK, so when overclocking with a basic clock generator, you simultaneously need to adjust the multipliers responsible for generating the DDR4 SDRAM frequency.

Overclocking of engineering Skylake stepping Q0 occurs precisely according to this algorithm, but one should not expect from them the same overclocking results as those obtained on serial processors. The Q0 stepping is preliminary, and the maximum frequency at which its carriers are capable of operating lies in the vicinity of the 4-GHz mark.

In the process of studying the capabilities of our Core i7-6400T QHQG instance, we plotted the dependence of its maximum frequency on the level of the supplied voltage. The graph below clearly shows that the frequency potential of engineering Skylake is indeed worse than that of serial processors, but this is not due to some artificial limitations, but to the design of the Q0 stepping core itself.

The nominal voltage of the tested Skylake instance is 1.12 V, but we started our overclocking experiments with a lower value of 1.0 V. And as practical tests have shown, even in this case the engineering Core i7-6400T easily reaches the 3-GHz bar. However, you shouldn’t delude yourself too much. Increasing the voltage does not push the limit of stable overclocking too far, but the operating temperatures increase very sharply.

As a result, an increase of 0.4 V to the starting voltage made it possible to push the limit frequency only by 850 MHz. And the maximum result that we managed to squeeze out of our Core i7-6400T QHQG sample was only 3.9 GHz. Attempts to further increase the frequency had to be discarded due to excessive heating of the processor crystal in stability tests in LinX 0.7.0, even despite the fact that all experiments were carried out with a fairly good cooler Noctua NH-U14S.

As you can see in the above screenshot, overclocking was carried out with the multiplication factor set to 24x. This is the maximum turbo multiplier at which the processor under study can work with a load on all cores at once. The BCLK frequency was raised to 162.5 GHz, which ultimately brought the Core i7-6400T to the 3.9 GHz mark. However, in order to be able to conduct a full cycle of stability tests in this state, the Vcore voltage had to be increased to 1.425 V. And this led to almost critical heating - throttling in engineering Skylake is activated at 100 degrees, just like in regular Core i7s.

It is obvious that, like in serial processors, engineering samples also use a polymer thermal interface, and it is clearly not of the best quality. But for a processor with a nominal frequency of 2.2 GHz this is not surprising. However, you should keep in mind that the best overclocking of the Core i7-6400T can be achieved using scalping.

At first glance, overclocking Skylake to 3.9 GHz does not seem like a particularly successful overclocker. However, do not forget, we are talking about preliminary stepping and about a processor whose nominal frequency was 1.8 times lower than the achieved overclocking. Therefore, the result obtained is actually not that bad. In the end, having taken a processor costing about $130-$150 for the experiment, we eventually came to the frequency that $300 CPUs provide. And this, generally speaking, is not an illusory gain at all.

In addition, any overclocking is always a lottery. And with another Core i7-6400T the result could have been completely different. For example, on the Internet you can find many reviews indicating that engineering CPUs have conquered the 4-GHz mark or even the possibility of stable operation at frequencies of about 4.2 GHz. In other words, the Core i7-6400T is a completely worthy object of effort for a budget-conscious overclocker.

As for our specific case, we were able to overclock without any special effort and without tedious selection of secondary parameters. Only the single voltage V CORE was changed, and in general the set of applied settings can be found in the following screenshot.

The overclocking was done on an ASUS Maximus VIII Ranger board, but that's not a big deal. Chinese engineering processors are overclocked on other motherboards in approximately the same way. True, in some cases you may encounter disabling AVX/AVX2 instructions and temperature monitoring, for example this happens on some ASRock motherboards. But in this case, BIOS versions modified by enthusiasts, which are regularly posted on the Korean website, can come to the rescue

Good day to all.
Briefly: so-called engineering samples (ES - Engineering Samples) of INTEL processors have appeared on foreign trading platforms in significant quantities. These processors are, in fact, variations of the Intel i7-6700.

Below the cut I’ll tell you about buying an ES processor with QHQG L501639 stepping (aka I7 - 6400T), about the pros (of which there are more) and cons, and about some of the pitfalls of such a purchase.
UPD1: Added information on how to preserve the processor's power-saving functions during overclocking! ( Yes, WE CAN!)

UPD2: At the request of readers, I am posting a link to ready-made BIOS settings for ASROCK Z170 pro4 (at the end of the review).

A short note for your attention about such a relatively new trend on foreign sites as the modern generation Skylake engineering processors. Yes, of course, it was possible to get an engineering sample of the processor abroad before, but what happened this summer-autumn is something new, in terms of the massiveness of the leak. YouTube is already full of videos about these processors, most often about their gaming power, but few people talk/film about the intricacies of the purchase itself and choosing a processor among a bunch of items.

Processors, due to the hype and huge benefits, are quickly sold out. For example, when I ordered the subject, the seller had more than 800 pieces in stock, but now they are no longer in stock.
Under the spoiler I will add analogues from other sellers:

Additional information about INTEL Skylake ES processors

Additional Information

Let me clarify that the name 6400T was given to these processors by the Chinese and in fact there are no restrictions inherent in the T-versions of Intel processors (like the even more limited Tjmax), they don't have it.

I will also briefly indicate the differences in the ES models of these processors known to me (from the Internet) and their stepping:

QHQJ is another HTPC candidate, clocked at 1.6 GHz and TDP 35w, coupled with the latest generation of integrated graphics from Intel, HD530, looks very tempting. If only the price were even lower. :)

QHQF - the last revision before sending "for gold". The multiplier is 8-40, i.e., in fact, this is a full-fledged i7-6700. No problems noticed

Also briefly about the notorious decoding of the stepping name:

Additional Information

For fans of tracks and other things

Packaging photos

Here it is, handsome (already installed in the socket):

CPU

An ASROCK Z170 PRO4 board was purchased for the processor, as the most successful option in terms of price/quality ratio.

Briefly other PC characteristics

where the new arrival is installed: Zalman Cnps11x Performa cooler tower, 8GB DDR4 Ram Patriot VIPER, KFA2 Gtx 1060 6gb EXOC, SSD 120 KingFast F9

Well, let's apply thermal paste (not the one included in the kit, of course! :) and preferably something better than KPT-8), install the cooler, pinch our fingers and start the PC.
Everything was determined the first time, and the computer will be able to show us everything about the processor.

Processor idle and under load:

Everything doesn’t look as rosy as we would like (although even in stock it is more powerful than the i3-6300, which is sold in Russian stores for the same amount). But all the pulp - in acceleration!

Additional information about the situation with overclocking processors

In general, to achieve full parity with the I7-6700, we need to help the processor by overclocking via BCLK. To do this, you need a modified BIOS (on ASUS boards, new BIOSes support everything out of the box, and you can download them on the website) corresponding to your device and a little time. Also, after flashing, do not forget to reset the BIOS with a jumper, this is highly recommended.
There are many videos on the Internet and YouTube with instructions on overclocking.
If they ask in the comments, I’ll post my BIOS settings, which I have tested and are quite successful.

As a result we get:

Processor after overclocking, as well as temperatures under load and a small benchmark

Cinebench R15 Score

More details about overclocking settings options:

overclocking to 4ghz without power saving: 167.7 bus, multiplier 24 (cache multiplier 23), EIST/TurboBOOST/C-states - Disabled. Power supply fixed 1.29 v, LLC - type 1

overclocking to 3.8 GHz-Max. frequency while maintaining stability and maintaining energy savings: 158.3 bus, multiplier 24 (cache multiplier 23), EIST/TurboBOOST/C-states - on, C1E - off. Power supply fixed 1.270 v, LLC - type 1

In both cases, you need to monitor the memory speed, which, although it is set to AUTO for most, grows as the bus increases. I recommend MANUALLY forcing the standard settings for your memory, which can be viewed in the SPD or on the sticker.

Well, that's much better! And this is not the limit, because the Internet is full of examples with overclocking it to 4.2-4.8 GHz. Productivity, there is no doubt, has increased. I worked stably for several days 24/7 - no problems in work, no glitches or BSODs in games. I think this is a good claim for stability.

UPD: The processor, as stated, is able to save energy saving technologies at BCLK less than 150MHZ, which I checked at the request of the workers. By setting the bus frequency to 145MHZ with a x24 multiplier, we have a frequency of 3.5 Ghz and a fairly cool processor (after all, the required voltage has also decreased to a completely standard 1.184), sacrificing only about 13% of performance (what a figure)!

Tests 3500mhz(24x145Mhz) and Cinebench

Now about cons overclocking:

It's not so scary

1. (24x167.7=4Ghz), we lose all Intel power saving technologies, including frequency and voltage reset when idle. Korean, which modified the bios m/n to support these processors, asserts that this feature is preserved for overclocking above ~3.5 GHz (more precisely, to a BCLK frequency above 150 mhz), and if you do not push it higher, the technologies remain operational and delight the owners. I plan to test this statement, and will soon supplement the review with results and confirmation/refutation of this statement. Well, the Korean was right, and the processor is capable save energy saving technologies at BCLK 145MHZ!

I want to emphasize that without a load on the processor, it cools down quite quickly and returns to idle temperatures even without these technologies, they are aimed only at reducing energy consumption, which is not as important here in Russia than in Europe.

2. When overclocked to i7-6700 level(24x167.7=4Ghz), we have an increased need for heat dissipation from the processor; according to my estimates, TDP has increased under load from 65w to 100-125w. Which is easily solved with an entry-level tower for 700-1000 rubles from Avito. Well, it’s still a top-end processor, after all, and yet it’s smaller than some 8-core processors from AMD.
Which, again, is not relevant at a frequency below the 150Mhz bus - in this mode, the voltage returns to the standard limits, EIST and C-States work, resetting the frequency and voltage as they are intended for.

This is my first review, I will accept adequate criticism and requests for points of interest to you regarding the subject.
If everyone likes everything, I’ll write a similar review about the Intel X3440 processor, which will breathe life into socket 1156 (opens video cards up to gtx1060), and can become a good basis for a budget PC gaming platform.
Thank you for your attention.

UPD2: At the request of readers, I am posting a link to ready-made BIOS settings for ASROCK Z170 pro4. Here is the link to Google Drive: drive.google.com/file/d/0Bw2ZBpGcW0a1Q05Ha1ZPUzZxTms/view?usp=sharing
I hasten to remind you that your RAM may not be able to handle the set settings (2666mhz 15-15-15-35), so when entering the BIOS, adjust the frequency and timings to suit your memory.

Planning to buy +133 Add to favorites I liked the review +194 +370

If we abstract from current market trends, when full-fledged overclocking capabilities are found mainly in expensive solutions (Intel K-series CPUs or AMD Black Edition processors, motherboards with older system logic sets, top-end versions of video cards with monstrous coolers and power converters), it all comes down to to one thing: if you want to overclock it, please pay for it.

advertising

At the same time, the overclocking capabilities of lower-end models (where performance is often lacking) are artificially limited, because if you can take something affordable and get the performance of a more expensive product, then they simply won’t want to buy the latter. As a result, the trend of overclocking from a niche of a truly useful activity designed to save money on purchasing a computer, over the years, has flowed into the segment of additional marketing tricks of manufacturers.

And the appearance in retail of something that allows a potential buyer to save money has moved from routine to the category of truly significant events. In my opinion, this is the emergence of engineering samples of Intel Skylake processors on sale, with prices in the region of 8-9 thousand rubles for a full-fledged Core i7. Let them function normally at very low frequencies, but is this a fixable issue? How much it can be fixed will be determined based on the results of the review.

What kind of processors are these and what are they used for?

Detailed information about Intel Skylake engineering processors can be found.

These CPU models are analogues of the i7-6700, are made on full-featured cores (that is, with the full amount of cache memory, all cores, etc.), and differ from officially sold commercial copies by extremely low multiplication factors, requiring a significant overclocking of the base frequency.

List of markings of engineering processors:

From the point of view of price-quality ratio, the most interesting and widespread models are the samples marked QHQG, designated by the Chinese as i7-6400T. It was a couple of such solutions that ended up in the test laboratory.

Yes, among the two steppings on sale (L448 and L501), I got earlier copies of L448, judging by the reviews, which have slightly less potential, but it is all the more interesting to evaluate the capabilities of the processors and squeeze “all the juice” out of them.

I note that no special version of the motherboard BIOS was used to test/overclock the processors. All review participants started up out of the box, and there were no artificial restrictions on overclocking.

Test stand

Testing was carried out with the following configuration:

Motherboard: ASUS Z170I Pro Gaming;
Processor: Intel Core i7-6400T, QHQG, 2.2 GHz;
Cooling system: Thermalright SilverArrow SB-E Extreme;
Thermal interface: Prolimatech PK-1;
RAM: G.Skill Ripjaws4 F4-3000C15Q-16GRR, 2 x 4 GB, DDR4-3000 15-15-15-35 1.35 V;
Hard drive: Western Digital Caviar Blue (WD500AAKS), 500 GB;
Power supply: Corsair GS Series GS800, 800 Watt;
Housing: open stand.

To measure voltages, a Mastech MY64 multimeter was used; energy consumption measurements were carried out using the same multimeter and a 50 A 75 mV shunt (75SHIP1-50-0.5) at the positive break of the 8 pin power cable.

Brief introduction to the motherboard

Such a motherboard has already been tested in the laboratory by a colleague wildchaser, however, before starting to study the capabilities of the processor, you first needed to understand what you would have to face and how it behaves in certain situations.

advertising

This material should be useful not only to those who will overclock the processor with this motherboard, but also to owners of models from other manufacturers. Therefore, the first step was to compare the voltage values set in the BIOS with the multimeter readings in order to obtain a stable reference point that is not tied to a specific hardware.

Noctua NH-D14 on the ASUS Z170I Pro Gaming motherboard (photos taken from the corresponding review).

The first thing that was done was to set the processor supply voltage to 1.2 V, then turn off all energy-saving technologies and check the operation of various Load-Line Calibration modes.

The measurement results are summarized in the table:

advertising

Load-line Calibration mode	Simple, measured with a multimeter, V
Level 1	1.193	1.158	1.112	1.127-1.145
Level 2	1.194	1.165	1.112	1.128-1.148
Level 3	1.195	1.178	1.116	1.125-1.146
Level 4	1.196	1.192	1.115	1.123-1.147
Level 5	1.197	1.205	1.109	1.121-1.142
Level 6	1.198	1.221	1.109	1.120-1.140
Level 7	1.200	1.236	1.109	1.121-1.138

Software monitoring lives a life of its own, while measurements using a multimeter show that Level 4 and Level 5 modes are optimal for use; For further experiments with overclocking, Level 5 mode was used.

I note that the software monitoring readings were influenced by the value of the set multiplication factor; when testing the motherboard to overclock the base frequency without overclocking the CPU, one could see voltage readings below 1 V, while the multimeter showed the same stable 1.2 V, as with maximum multiplier.

Additionally, secondary stresses were measured:

It might be useful to read: