Intel felt good for a long time, not feeling any serious competition from AMD. The release of the Core family in the past and the abandonment of the NetBurst architecture have allowed it to strengthen its dominant position in all market segments. The “Tick-tock” strategy introduced at the same time, following which the chipmaker updated the architecture and technical process every two years, made it possible for the processor giant to easily outperform the opponent, but after mass products based on the Broadwell core, it failed. Problems with the transition to more subtle technological standards and the lack of the same competition stretched Tak for two generations with minimal productivity growth. And we would have continued to observe such a calm in the processor market if AMD, after vague attempts with its Bulldozer, had not finally declared itself with the Zen architecture.
The release of Ryzen processors forced Intel to reconsider its plans for the release of new products, and the mass user finally got six-core CPU models that only the HEDT platform could boast of previously. But the Zen / Zen + solutions turned out to be so successful, occupying all the niches in the market, that they left the chipmaker no other chances, except to further increase the power of their available processors, especially since “eight cores in every home” are now in trend.
Coffee Lake Refresh
New products, which are still produced according to 14-nm technical standards (and this, oddly enough, is already the fourth generation), in terms of microarchitecture, do not differ in any way from their predecessors of the 8000 series. In principle, we have before us another iteration of Skylake, which has its roots in the ever-living Sandy Bridge. This is not surprising, given the very successful architecture, which is only enough to optimize after a couple of generations. The main optimization was made last year, with the release of the first Coffee Lake cores, which made it possible to increase the frequency potential relative to Kaby Lake, which underlay the 7000 series solutions. Despite the increased number of compute units and L3 cache, 8000-series processors, with proper cooling during overclocking, could already reach 5 GHz. This time, there are even more cores, and the operating clock speeds have grown. Below is a snapshot of the new Coffee Lake die compared to the previous generation.
As you can see, the difference between them lies in the addition of two more computing modules and the corresponding amount of L3 cache, which can reach 16 MB that has never been seen before in the mass segment. The remaining blocks, with a greater degree of probability, remained unchanged. There is support for SSE4.1, SSE4.2, AVX, FMA3 instructions, hardware acceleration of AES encryption, as well as AVX2.0 vector instructions – everything is the same as in previous generations of processors.
But what really appeared new, while the old forgotten long ago, is the type of thermal interface between the crystal and the heat-distributing cover. As we remember, Intel, starting with Ivy Bridge-based CPUs, uses the usual thermal paste instead of solder, which worsens the temperature regime of the processor and for serious overclocking, you have to resort to the so-called scalping, which allows you to replace the standard thermal interface with liquid metal. Now the chip maker promises to return the solder, which should significantly affect the level of heating of new products.
As for the latest processors themselves, only three of them are provided for the mass platform, at least for now. These are the two older models Core i9-9900K and Core i7-9700K, and the more affordable Core i5-9600K.
|Processor||Core i9-9900K||Core i7-9700K||Core i5-9600K||Core i7-8086K||Core i7-8700K||Core i7-8700||Core i5-8600K||Core i5-8600||Core i5-8500||Core i5-8400||Core i3-8350K||Core i3-8300||Core i3-8100|
|Core||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S||Coffee Lake-S|
|Process technology, nm||14||14||14||14||14||14||14||14||14||14||14||14||14|
|Number of cores (threads)||8 (16)||8||6||6 (12)||6 (12)||6 (12)||6||6||6||6||4||4||4|
|Rated frequency, GHz||3,6||3,6||3,7||4,0||3,7||3,2||3,6||3,1||2,8||2,8||4,0||3,7||3,6|
|Turbo Boost frequency, GHz||5,0||4,9||4,6||5,0||4,7||4,2||4,3||4,3||4,0||4,0||–||–|
|Multiplier unlocked up||+||+||+||+||+||–||+||–||–||–||+||–||–|
|L1 cache, KB||8 x (32 + 32)||8 x (32 + 32)||6 x (32 + 32)||6 x (32 + 32)||6 x (32 + 32)||6 x (32 + 32)||6 x (32 + 32)||6 x (32 + 32)||6 x (32 + 32)||6 x (32 + 32)||4 x (32 + 32)||4 x (32 + 32)||4 x (32 + 32)|
|L2 cache, KB||8 x 256||8 x 256||6 x 256||6 x 256||6 x 256||6 x 256||6 x 256||6 x 256||6 x 256||6 x 256||4 x 256||4 x 256||4 x 256|
|L3 cache, MB||16||12||9||12||12||12||9||9||9||9||8||6||6|
|Graphics core||n/a||n/a||n/a||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630||UHD Graphics 630|
|Graphics core frequency, MHz||n/a||n/a||n/a||350–1200||350–1200||350–1200||350–1150||350–1050||350–1050||350–1050||350–1150||350–1100||350–1100|
|Recommended cost, $||488||374||262||n/a||370||312||258||224||202||187||179||147||117|
Noticeable differences between the current generation and the past are visible only in the Core i7 series – the number of computing cores has increased to eight, while Hyper-Threading technology is not supported and the Core i7-9700K processor is capable of processing only eight threads per clock. As such, it is very similar to the Core i5 series and, more recently, the Core i3 series, which do not have additional processing power options. Multithreading has already been reserved for the Core i9-9900K, Intel’s most powerful modern desktop processor for the mass market. In addition, it boasts a round figure in its operating frequency – as much as 5 GHz, just like the elite Core i7-8086K. True, this value is available with Turbo Boost 2.0 technology and only for a couple of cores, while for all eight it will be 4.7 GHz. But the consumer is unlikely to delve into such details; Taking into account the pathos of AMD in terms of presenting its products, Intel did not stand aside and for the recommended $488 the user can get a brand new CPU in a very original package in the form of a dodecahedron.
The operating frequency of the Core i7-9700K is declared less than 100 MHz for all modes of operation of Turbo Boost technology and will be 4.9 / 4.6 GHz for one and eight cores, respectively. Compared to the predecessor Core i7-8700K, the characteristics of the new product have grown by 200 MHz. The Core i5-9600K model turned out to be an exact copy of the Core i5-8600K with increased frequencies up to 4.6 GHz in the maximum mode.
All new processor models officially work with DDR4-2666 memory and are installed in motherboards based on the 300 series system logic, subject to appropriate support. In addition, the release of a new chipset – Intel Z390, which was announced last week, is timed to coincide with the CPU 9000 series.
All the difference between the new PCH and the Z370 came down to support for six USB 3.1 Gen2 ports and the ability to work with wireless network adapters. You can learn more about it in the material on the motherboard based on it.
We got an engineering version of the Core i7-9700K for testing, which outwardly does not differ from retail copies, except for the marking.
The nominal frequency of the processor is 3.6 GHz, but thanks to the work of Turbo Boost technology, it rises to 4.6-4.9 GHz, depending on the number of cores involved, while the frequency of the Uncore block is 4.3 GHz. You can also see from the screenshots below that the associativity of the third level cache has become 12-channel, while the Core i7-8700K had 16 channels.
In old and unoptimized applications, one of the CPU cores in the new product will operate at a frequency of 4.9 GHz, while with increasing load and using more computing units, it will gradually decrease by 100 MHz per step. Ultimately, the user will most often observe 4.6 GHz, but this is subject to the Multi-Core Enhancements feature activated in the UEFI motherboard. If the processor goes beyond TDP during operation, then the frequency can be reset below this value, up to its nominal value.
In idle mode, the processor frequency drops to 800 MHz, as does the supply voltage, which is about 0.7–0.8 V. To work at higher frequencies during Turbo Boost operation, the voltage will already be 1.14–1.2 V, with the maximum frequencies of one core are already 1.3–1.4 V.
The nominal temperature of the Core i7-9700K during the Prime95 stress test causes conflicting feelings – 79 ° C under the Noctua NH-U14S supercooler is somehow too much, especially when the room is about 24 degrees. Perhaps we came across such a copy of the processor, and retail solutions will have a less hot temper. Or it may be that eight cores with such a technical process are simply not able to work with less heat generation. Naturally, the Multi-Core Enhancements feature in the motherboard was activated, and it does not allow the frequency to drop below the minimum value that Turbo Boost technology allows. If it were disabled, then we would probably get a fairly cold processor, but its performance would suffer.
Now we come to the most interesting – will the new processor be able to operate at a frequency of 5 GHz with the involvement of all cores? Let’s try to answer this question a little later, but for now let’s look at the possibilities with more “sparing” modes.
So, manually setting the voltages, both on the processor and on the cache with a memory controller, allowed the Core i7-9700K to function using all the computing units at a frequency of 4.8 GHz, while the hottest core warmed up to 93 ° C.
An attempt to further reduce the supply voltages made it possible for the processor to operate at 4.9 GHz at an even lower temperature – just some 90 degrees.
The next experiment for the cherished 5 GHz required a serious increase in the supply voltage, up to 1.3 V and higher, which ultimately led to a fiasco – the processor instantly overheated and went into throttling. Even CBO be quiet did not help! Silent Loop 280mm.
It is possible that the whole problem is in our pre-production sample, and it is likely that users will have much better luck with retail processors than we do. But for now, we can safely say that the latest eight-core CPUs to work at 5 GHz with all active cores fit at a stretch.
Why with a stretch? Well, at 5 GHz our Core i7-9700K overheated corny in LinX and Prime95, but it easily passed all tests in benchmarks without a single error, while its temperature was at the level of 70-75 ° C. If we speed up the memory at the same time, then the processor warms up to 80 degrees.
Given this behavior of the CPU under consideration, it was decided to make its further comparison with a competitor at this frequency – after all, there is some hope that retail copies will be able to show themselves better.