The Intel Core i7-8086K (Part 1)
- Transfer
Part 1 >> Part 2 >> Part 3 >> Part 4
Intel announced the release of a new processor, and we in turn asked the company to provide us with one sample for testing. Having received in response - “no samples for the press”, we simply bought the processor ourselves. In honor of the 40th anniversary of the x86 microarchitecture and 8086 processor, Intel released the Core i7-8086K based on Coffee Lake with a turbo frequency of up to 5.0 GHz. Intel has released a limited batch of 50,000 processors, 8086 of which participate in regional lotteries and are distributed free of charge to the winners. This is a “daring” marketing move from Intel in promoting its main product line. Like most processors, the whole secret lies in binning, and it really is the most productive mainstream processor ever released by Intel. However, the 8086K may not be as good as it seems.
Jubilee processors, they are also processors with a limited edition, for many years remain a "dark horse." Back in June 2014, Intel released the version of the Pentium Anniversary Edition G3258 - an overclocking dual-core processor - with a lot of noise and pathos, but no matter how overclocked the processor was, it never came close to a full quad-core processor. In 2009, AMD launched a limited edition of 100 copies of the Phenom II X4 TWKR - a processor designed to surpass world performance records. It has never been available in retail. However, in honor of the 50th anniversary of Intel and the 40th anniversary of the x86 architecture, the new Core i7-8086K is available for 50,000 happy users.
Applause to David Shora from Wikichip, who predicted the appearance of the sixth-generation Coffee Lake of the 8th generation with a frequency of 5.0 GHz and called i7-8086K on the anniversary date of June 8.
Whether David has spies at Intel, or access to a time machine, or an incredible ability to influence Intel's manufacturing solutions, but he was able to predict the product’s release on January 18th.
Core i7-8086K, as predicted by David, is a six-core processor based on Coffee Lake, running on a turbo frequency of 5.0 GHz. Although the batch of processors, in fact, are simply the best binaries of the Core i7-8700K.
The main thing to note is that the processor does not issue 5 GHz on all cores. The CPU still has an official TDP rating of 95 W, so instead of raising the TDP and squeezing a bit more, Intel plays within its TDP range. We have already seen that the recent transition to six-core processors as the main platform led to the fact that Intel processors exceed their TDP ratings, thereby causing confusion among consumers and the press. It is important to note that Intel's TDP rating is valid only at the base frequency, in this case it is 95 watts at 4.0 GHz, and any turbo modes can consume as much power as necessary. We will talk about this moment in the current review.
Therefore, one of the key questions to this “faster” processor is how much faster it is than Core i7-8700K. A few years ago, Intel had already launched a number of Devil's Canyon processors as the "best version for overclocking" of the main Haswell processors, which provided a higher frequency and better thermal response to frequency and voltage. If you expected the same thing from the Core i7-8086K, I'm afraid you might be disappointed.
The Intel Core i7-8086K has the same core frequencies as the Core i7-8700K, with the exception of one core. Only one core in turbo mode has a frequency increased from 4.7 GHz to 5.0 GHz, but it is precisely this that is the cause of all the hype.
Raising a turbo to just one processor core has limited benefits. For all tests, except for the purest single-core, there should be no difference in the results. The problem is that in very few situations on modern machines only one core is loaded: almost every user has programs running in the background, such as other browser tabs, virus scanners or unfair updates. Gamers should not expect any noticeable benefits.
Technically, the base frequency increased from 3.7 GHz to 4.0 GHz, which means that Intel guarantees a higher minimum frequency for 95 W TDP, which is practically unimportant for users who simply connect the processor to their system. The fact is that the processor will never use this base frequency. An interesting part of this review: the base clock with a clock frequency of 4.0 GHz means that it is a processor from the best binning, and should accelerate better.
Based on the results, it will become obvious that an increase in the frequency of the turbo of just one processor core did not become a big breakthrough, which is indicated by the figure "5.0 GHz" on the side panel of the box. And the pricing policy points to this fact: Intel installed the MSRP for the Core i7-8086K at $ 425, compared to the retail price of $ 350 on the Core i7-8700K. This is a difference of $ 75 +, or even more, because the 8700K periodically goes on sale at an even lower price.
The processor tested here was not provided to us by Intel at no charge. At Computex, we were told that Intel does not plan to provide samples for the media. This was somewhat disappointing when you consider how much excitement was raised by the company around the new processors. There is an assumption that Intel will sell all jubilee processors without problems, and therefore there is no point in stimulating additional sales by attracting the media. After the review, there may be other suggestions why the company is not interested in conducting a survey by the media.
I was in Taiwan - it was necessary to attend several meetings after Computex, and would not be able to return home to my office for several days. Therefore, when Intel announced that they did not provide a processor, we had two options: try to win one in the lottery and wait eight weeks, or wait for my return home and buy a processor. We found our own, third option - AnandTech option: buy a processor in one of the local stores and check it out here.
Some guy with a processor in the box.
We turned to partners to borrow a system for a couple of days for testing in a hotel room, and ASRock kindly provided the motherboard. Already in the evening a completely ready system from the exhibition hall was delivered to my hotel.
Many thanks to ASRock for their help, making testing possible. They also provided a Core i7-8700K. This test system is obviously different from my test system in the office because, as a rule, for the sake of fairness of reviews, we use a permanent system for testing processors. To be objective, we will compare only 8700K and 8086K. There is still some good news - I took my standard OS for testing, and the tests themselves were with me to Taiwan. Never leave home without them guys.
As already mentioned, many thanks to ASRock for providing the system in such a short time. Without them, this review would not have been possible in the desired time frame.
It should be noted that, as in previous reviews, our basic CPU settings include setting the memory to the maximum supported processor frequency. In this case, we use DDR4-2666 for Intel Coffee Lake processors.
It should be noted that ASRock could not provide us with the same graphics processor that I usually use to test games. Instead, we were able to install the RX 580, so this means that our game testing data will also have only two points of reference: Core i7-8700K and Core i7-8086K.
As mentioned earlier, one of the key differences of this test is the motherboard. In the office, we used the ASRock Z370 Gaming i7 (P1.70 BIOS) to test our Coffee Lake, while here we use the ASRock Z370 Taichi (P1.80 BIOS). Different motherboards, even from the same company, use different methods to control the internal frequencies on the board (for example, Uncore) or power limits (PL2), which may differ from BIOS to BIOS. It is difficult to maintain these sequences in different systems, so some differences are expected in the results.
In the short time that we spent with the processor, we sketched a few pages to interest you:
Part 1
Part 2
3 part
4 part
One of the main questions associated with the new processor is whether Intel has decided to make changes to the way CPU connects and the heat sink. The best contact method is to use an Indium-Tin solder or liquid metal to ensure that the heat load from the processor is transferred directly to the cooler. A cheaper method (but also more reliable) is with thermal paste, which is more resistant to thermal expansion coefficients during the life cycle of the processor. Ideally, we expect the highest performance processors to use the soldering method, while cheaper processors can use thermal paste. However, Intel recently makes its processors exclusively with thermal paste, as a result of which extreme enthusiasts resort to “opening” the processor and replacing the thermal paste with liquid metal.
The elimination of the AMD Ryzen 5 2400G APU:
The method of opening the Intel chip is generally the same. However, the question posed to us: what if Intel switched from the thermal paste used on the Core i7-8700K to a more accelerating and thermally conductive material for the Core i7-8086K. The idea is that if Intel is focused on enthusiasts, you need to use soldering, right?
We are extremely grateful to Alva Jonathan, also known as “Lucky_n00b”, to overclocker and Jagat Review journalist. I know Alva for almost 10 years now, and, like me, he also bought his Core i7-8086K during Computex this week, except that he went straight to opening the processor and cooling it with liquid nitrogen. He allowed us to share our results with our audience, so many thanks to Alve!
Alva makes an impressive review of overclocking on all the new platforms in the Jagat Review (in Indonesian), and also shows great performance in overclocking competitions all over the world. This week he finished third in the overclocking G.Skill event at Computex, winning good equipment and a cash prize.
Suffice it to say that Intel made zero changes to the thermal interface on the Core i7-8086K. It is completely identical to the Core i7-8700K, using the same heat conductor as in previous generations of chips. With modern Coffee Lake processors, the removal of thermal paste and its replacement with the implementation of liquid metal usually lead to a decrease in temperature from 5 to 15 ° C (depending on the quality of the application) or another 100-300 MHz depending on the effect of voltage on the chip.
Alva recommends disassembling the processor for higher frequencies or lowering the temperature only if you plan to supply more than 1.30 volts per processor. With this voltage, with a good cooler, the processor temperature will be around 80 C with a full CPU load (we can confirm that our results are similar), which is a turning point for those who consider disassembly of the processor as an option.
On its processor, the Alva reached 5.0 GHz at 1.20 volts, and the CPU was stable enough to score 1627 points on the CineBench R15 (compared to 1424 on the stock processor with fast memory). The processor also coped with a frequency of 5.2 GHz at 1.35 V, which increased its scores to 1692. Alva used Kingpin Cooling KPX as a thermally conductive material to replace thermal paste.
Extreme overclocking is an interesting pastime, but for users who are at the top of this sport, every MHz is important. The result is achieved not only by cooling, but also by physically modifying the systems, by improving the power supply or by adjusting the voltage manually, and not through software. For those who know how to do it, this is a real competition, able to deliver a lot of thrills.
As we can see from Alva’s test notes, he started from an MSI Z370 Godlike Gaming motherboard, prepared for cooling to ultra-low temperatures, and used a heavy LN2 copper vessel to control the temperature using liquid nitrogen. After the system cooled down to -100C, it booted by adjusting the BIOS so that the processor was at 6.0 GHz (60x100), with uncore 5.0 GHz and at a processor voltage of 1.70 volts. Do not try this without over-cooling (!). Other voltages were as follows:
The CPU was supported in full 6C / 12T mode.
After loading into the OS, MSI Command Center Lite was used to set processor variables (multiplier, base clock, voltage) in real time. The system was cooled to the limit known as “full pot” liquid nitrogen benchmarking, and the multiplier was increased to find the absolute limit of the processor frequency.
Final result? 7309 MHz: here
In general, Skylake-based processors tend to show peak frequencies when cooled with liquid nitrogen around 7.1-7.4 GHz, so the new processor is no different from the usual one. Alva said that he was very pleased with the chip, but he would need to check a few more to see exactly where there are variations (if any) in the matrix / batch from Intel. When Alva publishes his full overclocking article, I will contact him.
Update: Here is Alva 's article .
To test overclocking to work on a continuous basis, 24/7, we used the system in our hotel room to evaluate how the Core i7-8086K works on closed-loop liquid cooling, passing through multipliers one by one. For this, we used our standard overclocking method.
Due to time and location, our overclocking method was as follows:
Blender includes a good combination of hard disk load, memory load and, as a result, power consumption. Any problems requiring additional voltage to stabilize can be found very quickly after the start of the test.
The Blender test lasts about five minutes on the Core i7-8086K, which is enough for our quick overclocking testing. For users who insist on confident stability in 24/7 mode, this is not the most suitable test, but it still provides an excellent load on the system.
Using this methodology, we achieved the following results:
With the default settings, our system would get into a full-scale 4.3 GHz turbo and complete the test in 311 seconds on Blender, with a CPU temperature of 62 degrees and a consumption of 115 watts. We also tested the system “on auto”, but setting the frequency to 5.0 GHz on all cores. This gave a test result of 268 seconds, but significantly higher temperatures (82 C) and power consumption (175 W).
If we manually increase the frequency, starting at 4.0 GHz, we can see that the OS stops receiving power information: AIDA64 showed a voltage that slowly increased as the multiplier increased, even if the voltage setting in the BIOS did not change. AIDA64 also shows a 1,364 volt crash, although adjusting the voltage in the BIOS helped with higher coefficients. It was strange, but I think that the indicative results in the table are Blender, Temperature and Power.
I'm going to set up Blender's output as “render per hour” - this is easier to visualize on a chart.
The key result here is 5.0 GHz, it is an excellent compromise between power and performance, as well as temperature and voltage. At this level, the system gives + 16% performance due to an additional + 16% frequency. However, there are nutritional problems.
A comparison of the 5.0 GHz manual overclocking with the default processor shows a 32% increase in power consumption. But when compared with the equivalent manual “overclocking” at 4.3 GHz, the difference in power consumption is now 68%. At this stage, we really "stretch" the micro-architectural design to the limit.
It should be noted that with the default settings, the system consumed 115 W, which is 20 W higher than the declared TDP. As mentioned earlier, the TDP is determined at the base frequency, which in this case is 4.0 GHz. We watched the power consumption of 80 W at the base frequency, showing that the processor is still - technically - under its TDP value, at least if the user optimizes the voltage. At the consumption level of 95 W, if we had maximized the frequency for this TDP, we would have to see the base frequency of 4.4 GHz on this chip.
However, think about what could happen if Intel decided to increase TDP by + 10 W or +15 W, to 110 W. In this case, we would have a chip with a base frequency of about 4.6 GHz, taking into account the results of our testing. As we will see in the results on the following pages, Intel really missed a good chance by refusing to increase the TDP.
For anyone interested in the upper limits of our chip, let's call 5.1 GHz a realistic maximum. I could not get 5.2 GHz, and keep the system stable in Blender for more than 30 seconds. As the voltage in the BIOS increased to 1,425 volts, the system showed maximum temperatures of about 100 ° C, which greatly exceeded the comfort limit. Although it is worth mentioning Alva and his excellent chip: he said that with delidding (disassembling the chip and replacing the heat conductor) the frequency of 5.2 GHz is quite achievable, but further increase can be very difficult, considering how quickly the voltage in our sample grows.
As for the absolute maximum at which we were able to load Windows - we observed 5.4 GHz. No load was applied due to fear of overheating, and at 5.5 GHz the system could not work.
As part of our testing, we were able to conduct several tests both with high overclocking and fast memory (as well as at the same time). Again, many thanks to ASRock for the system provided.
Thank you for staying with us. Do you like our articles? Want to see more interesting materials? Support us by placing an order or recommending to friends, 30% discount for Habr users on a unique analogue of the entry-level servers that we invented for you: The whole truth about VPS (KVM) E5-2650 v4 (6 Cores) 10GB DDR4 240GB SSD 1Gbps from $ 20 or how to share the server? (Options are available with RAID1 and RAID10, up to 24 cores and up to 40GB DDR4).
Dell R730xd 2 times cheaper? Only here2 x Intel Dodeca-Core Xeon E5-2650v4 128GB DDR4 6x480GB SSD 1Gbps 100 TV from $ 249 in the Netherlands and the USA! Read about How to build an infrastructure building. class c using servers Dell R730xd E5-2650 v4 worth 9000 euros for a penny?
The Intel Core i7-8086K: Review
Intel announced the release of a new processor, and we in turn asked the company to provide us with one sample for testing. Having received in response - “no samples for the press”, we simply bought the processor ourselves. In honor of the 40th anniversary of the x86 microarchitecture and 8086 processor, Intel released the Core i7-8086K based on Coffee Lake with a turbo frequency of up to 5.0 GHz. Intel has released a limited batch of 50,000 processors, 8086 of which participate in regional lotteries and are distributed free of charge to the winners. This is a “daring” marketing move from Intel in promoting its main product line. Like most processors, the whole secret lies in binning, and it really is the most productive mainstream processor ever released by Intel. However, the 8086K may not be as good as it seems.
Holiday reason: 40 years x86
Jubilee processors, they are also processors with a limited edition, for many years remain a "dark horse." Back in June 2014, Intel released the version of the Pentium Anniversary Edition G3258 - an overclocking dual-core processor - with a lot of noise and pathos, but no matter how overclocked the processor was, it never came close to a full quad-core processor. In 2009, AMD launched a limited edition of 100 copies of the Phenom II X4 TWKR - a processor designed to surpass world performance records. It has never been available in retail. However, in honor of the 50th anniversary of Intel and the 40th anniversary of the x86 architecture, the new Core i7-8086K is available for 50,000 happy users.
Applause to David Shora from Wikichip, who predicted the appearance of the sixth-generation Coffee Lake of the 8th generation with a frequency of 5.0 GHz and called i7-8086K on the anniversary date of June 8.
Whether David has spies at Intel, or access to a time machine, or an incredible ability to influence Intel's manufacturing solutions, but he was able to predict the product’s release on January 18th.
Core i7-8086K, as predicted by David, is a six-core processor based on Coffee Lake, running on a turbo frequency of 5.0 GHz. Although the batch of processors, in fact, are simply the best binaries of the Core i7-8700K.
The main thing to note is that the processor does not issue 5 GHz on all cores. The CPU still has an official TDP rating of 95 W, so instead of raising the TDP and squeezing a bit more, Intel plays within its TDP range. We have already seen that the recent transition to six-core processors as the main platform led to the fact that Intel processors exceed their TDP ratings, thereby causing confusion among consumers and the press. It is important to note that Intel's TDP rating is valid only at the base frequency, in this case it is 95 watts at 4.0 GHz, and any turbo modes can consume as much power as necessary. We will talk about this moment in the current review.
Therefore, one of the key questions to this “faster” processor is how much faster it is than Core i7-8700K. A few years ago, Intel had already launched a number of Devil's Canyon processors as the "best version for overclocking" of the main Haswell processors, which provided a higher frequency and better thermal response to frequency and voltage. If you expected the same thing from the Core i7-8086K, I'm afraid you might be disappointed.
The Intel Core i7-8086K has the same core frequencies as the Core i7-8700K, with the exception of one core. Only one core in turbo mode has a frequency increased from 4.7 GHz to 5.0 GHz, but it is precisely this that is the cause of all the hype.
Raising a turbo to just one processor core has limited benefits. For all tests, except for the purest single-core, there should be no difference in the results. The problem is that in very few situations on modern machines only one core is loaded: almost every user has programs running in the background, such as other browser tabs, virus scanners or unfair updates. Gamers should not expect any noticeable benefits.
Technically, the base frequency increased from 3.7 GHz to 4.0 GHz, which means that Intel guarantees a higher minimum frequency for 95 W TDP, which is practically unimportant for users who simply connect the processor to their system. The fact is that the processor will never use this base frequency. An interesting part of this review: the base clock with a clock frequency of 4.0 GHz means that it is a processor from the best binning, and should accelerate better.
Based on the results, it will become obvious that an increase in the frequency of the turbo of just one processor core did not become a big breakthrough, which is indicated by the figure "5.0 GHz" on the side panel of the box. And the pricing policy points to this fact: Intel installed the MSRP for the Core i7-8086K at $ 425, compared to the retail price of $ 350 on the Core i7-8700K. This is a difference of $ 75 +, or even more, because the 8700K periodically goes on sale at an even lower price.
Some features of testing
The processor tested here was not provided to us by Intel at no charge. At Computex, we were told that Intel does not plan to provide samples for the media. This was somewhat disappointing when you consider how much excitement was raised by the company around the new processors. There is an assumption that Intel will sell all jubilee processors without problems, and therefore there is no point in stimulating additional sales by attracting the media. After the review, there may be other suggestions why the company is not interested in conducting a survey by the media.
I was in Taiwan - it was necessary to attend several meetings after Computex, and would not be able to return home to my office for several days. Therefore, when Intel announced that they did not provide a processor, we had two options: try to win one in the lottery and wait eight weeks, or wait for my return home and buy a processor. We found our own, third option - AnandTech option: buy a processor in one of the local stores and check it out here.
Some guy with a processor in the box.
We turned to partners to borrow a system for a couple of days for testing in a hotel room, and ASRock kindly provided the motherboard. Already in the evening a completely ready system from the exhibition hall was delivered to my hotel.
Many thanks to ASRock for their help, making testing possible. They also provided a Core i7-8700K. This test system is obviously different from my test system in the office because, as a rule, for the sake of fairness of reviews, we use a permanent system for testing processors. To be objective, we will compare only 8700K and 8086K. There is still some good news - I took my standard OS for testing, and the tests themselves were with me to Taiwan. Never leave home without them guys.
Test system
As already mentioned, many thanks to ASRock for providing the system in such a short time. Without them, this review would not have been possible in the desired time frame.
It should be noted that, as in previous reviews, our basic CPU settings include setting the memory to the maximum supported processor frequency. In this case, we use DDR4-2666 for Intel Coffee Lake processors.
It should be noted that ASRock could not provide us with the same graphics processor that I usually use to test games. Instead, we were able to install the RX 580, so this means that our game testing data will also have only two points of reference: Core i7-8700K and Core i7-8086K.
As mentioned earlier, one of the key differences of this test is the motherboard. In the office, we used the ASRock Z370 Gaming i7 (P1.70 BIOS) to test our Coffee Lake, while here we use the ASRock Z370 Taichi (P1.80 BIOS). Different motherboards, even from the same company, use different methods to control the internal frequencies on the board (for example, Uncore) or power limits (PL2), which may differ from BIOS to BIOS. It is difficult to maintain these sequences in different systems, so some differences are expected in the results.
In the short time that we spent with the processor, we sketched a few pages to interest you:
Part 1
- Intel Core i7-8086K: 40 years of x86 architecture
- Thermal interface and extreme overclocking with lucky_n00b
- Analysis of acceleration and power
Part 2
- Performance Testing: CPU System Tests
- Performance Testing: CPU Rendering Tests
- Performance Testing: CPU Encoding Tests
- Performance Testing: CPU Office Tests
- Performance Testing: CPU Legacy Tests
3 part
- Game Performance: Civilization 6
- Game Performance: Shadow of Mordor
- Game Performance: Rise of the Tomb Raider
- Game Performance: Rocket League
- Gaming Performance: Grand Theft Auto V
4 part
- Overclocking 5.0 GHz performance: CPU Tests
- Overclocking performance at 5.0 GHz: GPU Tests
- findings
Thermal interface and extreme overclocking
One of the main questions associated with the new processor is whether Intel has decided to make changes to the way CPU connects and the heat sink. The best contact method is to use an Indium-Tin solder or liquid metal to ensure that the heat load from the processor is transferred directly to the cooler. A cheaper method (but also more reliable) is with thermal paste, which is more resistant to thermal expansion coefficients during the life cycle of the processor. Ideally, we expect the highest performance processors to use the soldering method, while cheaper processors can use thermal paste. However, Intel recently makes its processors exclusively with thermal paste, as a result of which extreme enthusiasts resort to “opening” the processor and replacing the thermal paste with liquid metal.
The elimination of the AMD Ryzen 5 2400G APU:
The method of opening the Intel chip is generally the same. However, the question posed to us: what if Intel switched from the thermal paste used on the Core i7-8700K to a more accelerating and thermally conductive material for the Core i7-8086K. The idea is that if Intel is focused on enthusiasts, you need to use soldering, right?
Make it possible
We are extremely grateful to Alva Jonathan, also known as “Lucky_n00b”, to overclocker and Jagat Review journalist. I know Alva for almost 10 years now, and, like me, he also bought his Core i7-8086K during Computex this week, except that he went straight to opening the processor and cooling it with liquid nitrogen. He allowed us to share our results with our audience, so many thanks to Alve!
Alva makes an impressive review of overclocking on all the new platforms in the Jagat Review (in Indonesian), and also shows great performance in overclocking competitions all over the world. This week he finished third in the overclocking G.Skill event at Computex, winning good equipment and a cash prize.
Open the chip
Suffice it to say that Intel made zero changes to the thermal interface on the Core i7-8086K. It is completely identical to the Core i7-8700K, using the same heat conductor as in previous generations of chips. With modern Coffee Lake processors, the removal of thermal paste and its replacement with the implementation of liquid metal usually lead to a decrease in temperature from 5 to 15 ° C (depending on the quality of the application) or another 100-300 MHz depending on the effect of voltage on the chip.
Alva recommends disassembling the processor for higher frequencies or lowering the temperature only if you plan to supply more than 1.30 volts per processor. With this voltage, with a good cooler, the processor temperature will be around 80 C with a full CPU load (we can confirm that our results are similar), which is a turning point for those who consider disassembly of the processor as an option.
On its processor, the Alva reached 5.0 GHz at 1.20 volts, and the CPU was stable enough to score 1627 points on the CineBench R15 (compared to 1424 on the stock processor with fast memory). The processor also coped with a frequency of 5.2 GHz at 1.35 V, which increased its scores to 1692. Alva used Kingpin Cooling KPX as a thermally conductive material to replace thermal paste.
Going beyond using liquid nitrogen (LN2)
Extreme overclocking is an interesting pastime, but for users who are at the top of this sport, every MHz is important. The result is achieved not only by cooling, but also by physically modifying the systems, by improving the power supply or by adjusting the voltage manually, and not through software. For those who know how to do it, this is a real competition, able to deliver a lot of thrills.
As we can see from Alva’s test notes, he started from an MSI Z370 Godlike Gaming motherboard, prepared for cooling to ultra-low temperatures, and used a heavy LN2 copper vessel to control the temperature using liquid nitrogen. After the system cooled down to -100C, it booted by adjusting the BIOS so that the processor was at 6.0 GHz (60x100), with uncore 5.0 GHz and at a processor voltage of 1.70 volts. Do not try this without over-cooling (!). Other voltages were as follows:
- SA / IO Voltage: 1.35 V
- DMI Voltage: 1.80 V
- CPU PLL Voltage: 2.20 V
- CPU PLL OC Voltage 2.20 V
- CPU ST Voltage: 1.35 V
- CPU ST V6 Voltage: 1.35 V
The CPU was supported in full 6C / 12T mode.
After loading into the OS, MSI Command Center Lite was used to set processor variables (multiplier, base clock, voltage) in real time. The system was cooled to the limit known as “full pot” liquid nitrogen benchmarking, and the multiplier was increased to find the absolute limit of the processor frequency.
Final result? 7309 MHz: here
In general, Skylake-based processors tend to show peak frequencies when cooled with liquid nitrogen around 7.1-7.4 GHz, so the new processor is no different from the usual one. Alva said that he was very pleased with the chip, but he would need to check a few more to see exactly where there are variations (if any) in the matrix / batch from Intel. When Alva publishes his full overclocking article, I will contact him.
Update: Here is Alva 's article .
Analysis of acceleration and power
To test overclocking to work on a continuous basis, 24/7, we used the system in our hotel room to evaluate how the Core i7-8086K works on closed-loop liquid cooling, passing through multipliers one by one. For this, we used our standard overclocking method.
Home acceleration, step by step
Due to time and location, our overclocking method was as follows:
- Start with a 40x and 1.05 volt multiplier
- Set line load calibration to level 1 (ASRock Z370 Taichi)
- Download OS
- Run our Blender test, take power and temperature data from AIDA
- If the system fails or the temperature exceeds 95 ° C, stop testing.
- If the system fails, add +0.025 V and go to step 3
- If the test is successful, record the result of the Blender, add a multiplier and return to step 3
Blender includes a good combination of hard disk load, memory load and, as a result, power consumption. Any problems requiring additional voltage to stabilize can be found very quickly after the start of the test.
The Blender test lasts about five minutes on the Core i7-8086K, which is enough for our quick overclocking testing. For users who insist on confident stability in 24/7 mode, this is not the most suitable test, but it still provides an excellent load on the system.
results
Using this methodology, we achieved the following results:
With the default settings, our system would get into a full-scale 4.3 GHz turbo and complete the test in 311 seconds on Blender, with a CPU temperature of 62 degrees and a consumption of 115 watts. We also tested the system “on auto”, but setting the frequency to 5.0 GHz on all cores. This gave a test result of 268 seconds, but significantly higher temperatures (82 C) and power consumption (175 W).
If we manually increase the frequency, starting at 4.0 GHz, we can see that the OS stops receiving power information: AIDA64 showed a voltage that slowly increased as the multiplier increased, even if the voltage setting in the BIOS did not change. AIDA64 also shows a 1,364 volt crash, although adjusting the voltage in the BIOS helped with higher coefficients. It was strange, but I think that the indicative results in the table are Blender, Temperature and Power.
I'm going to set up Blender's output as “render per hour” - this is easier to visualize on a chart.
The key result here is 5.0 GHz, it is an excellent compromise between power and performance, as well as temperature and voltage. At this level, the system gives + 16% performance due to an additional + 16% frequency. However, there are nutritional problems.
A comparison of the 5.0 GHz manual overclocking with the default processor shows a 32% increase in power consumption. But when compared with the equivalent manual “overclocking” at 4.3 GHz, the difference in power consumption is now 68%. At this stage, we really "stretch" the micro-architectural design to the limit.
It should be noted that with the default settings, the system consumed 115 W, which is 20 W higher than the declared TDP. As mentioned earlier, the TDP is determined at the base frequency, which in this case is 4.0 GHz. We watched the power consumption of 80 W at the base frequency, showing that the processor is still - technically - under its TDP value, at least if the user optimizes the voltage. At the consumption level of 95 W, if we had maximized the frequency for this TDP, we would have to see the base frequency of 4.4 GHz on this chip.
However, think about what could happen if Intel decided to increase TDP by + 10 W or +15 W, to 110 W. In this case, we would have a chip with a base frequency of about 4.6 GHz, taking into account the results of our testing. As we will see in the results on the following pages, Intel really missed a good chance by refusing to increase the TDP.
Climbing to the top
For anyone interested in the upper limits of our chip, let's call 5.1 GHz a realistic maximum. I could not get 5.2 GHz, and keep the system stable in Blender for more than 30 seconds. As the voltage in the BIOS increased to 1,425 volts, the system showed maximum temperatures of about 100 ° C, which greatly exceeded the comfort limit. Although it is worth mentioning Alva and his excellent chip: he said that with delidding (disassembling the chip and replacing the heat conductor) the frequency of 5.2 GHz is quite achievable, but further increase can be very difficult, considering how quickly the voltage in our sample grows.
As for the absolute maximum at which we were able to load Windows - we observed 5.4 GHz. No load was applied due to fear of overheating, and at 5.5 GHz the system could not work.
Testing at 5.0 GHz
As part of our testing, we were able to conduct several tests both with high overclocking and fast memory (as well as at the same time). Again, many thanks to ASRock for the system provided.
Thank you for staying with us. Do you like our articles? Want to see more interesting materials? Support us by placing an order or recommending to friends, 30% discount for Habr users on a unique analogue of the entry-level servers that we invented for you: The whole truth about VPS (KVM) E5-2650 v4 (6 Cores) 10GB DDR4 240GB SSD 1Gbps from $ 20 or how to share the server? (Options are available with RAID1 and RAID10, up to 24 cores and up to 40GB DDR4).
Dell R730xd 2 times cheaper? Only here2 x Intel Dodeca-Core Xeon E5-2650v4 128GB DDR4 6x480GB SSD 1Gbps 100 TV from $ 249 in the Netherlands and the USA! Read about How to build an infrastructure building. class c using servers Dell R730xd E5-2650 v4 worth 9000 euros for a penny?