Core i9-9900K (8 cores), i7-9700K and i5-9600K

Among Intel's many announcements today, the launch of 9^th Generation Core desktop processors offering up to 8 cores on the mainstream Intel platform. These processors are compatible with the current Coffee Lake and Z370 platforms, but come with a new Z390 chipset and associated motherboards. The highlights of this launch are the core i9 8-core components, which include an i9-9900K Turbo 5.0 GHz processor, rated at a 95W TDP.

One, two, three, four, five, six, seven, eight. Eight hearts.

I will not go around the bush – the 9th^th The Intel processors announced today are as follows:

The Core i9-9900K, Intel's new flagship mainstream processor, is at the top of the pack. This part includes eight complete cores with hyperthreading, with a base frequency of 3.6 GHz at 95W TDP and a turbo up to 5.0 GHz on two cores. Memory is supported up to dual-channel DDR4-2666. The Core i9-9900K builds on the Core i7-8086K of the 8th^th Generating products by adding two more cores and boosting the 5.0 GHz turbo from one heart to two. The 100% turbo is 4.7 GHz, so it will be interesting to see what is the power consumption when the processor is fully charged. The Core i9 family will have 2 MB of L3 cache per core.

The Core i7-9700K also has 8 cores, but without hyperthreading. This part will also have a base frequency of 3.6 GHz for a 95 W TDP, but can only turbo up to 4.9 GHz on a single core. The i7-9700K is supposed to be the direct upgrade of the Core i7-8700K, and although both chips have the same underlying microarchitecture of Coffee Lake, the 9700K has two extra cores and a slightly higher, but lower, turbo performance. L3 cache memory by heart to the only 1.5 MB by.

The other important processor focused on overclocking is the Core i5-9600K, with six cores without hyperthreading. Users will also see a lot of similarities in this part with the previous generation Core i5, but with additional frequency.

From Tom's Hardware

At the time of writing this article, we are still waiting for the price, although the Core i9-9900K was accidentally listed on Amazon for $ 582.50 last week. The content of this accident list was also found: Intel seems to experiment a dodecahedron (figure to 12 sides) to try to compete with the developed packaging of AMD for its high-end processors. This is the first time in recent memory that Intel has extended its packaging beyond this simple box. It will be interesting to see how far it exceeds Core i9.

Turbo Ratios by heart

In our list of information, we also have Turbo Ratios per heart for each of the three processors capable of overclocking.

Users who read our review of the Core i7-8086K, which was the first Intel 5.0 GHz processor, may remember that it offered only one 5.0 GHz turbo core. The effect of this was that we never really saw the magic number of 5.0 GHz. Because of the background and other processes, we often see the basic values ​​of turbo 2-4 in most processors. This time, at least with the Core i9, Intel placed the first two cores at the peak turbo frequency.

Coffee Lake Refresh: Learn from GPU Enterprises

Intel 9^th The Generation Core family revolves around the Coffee Lake platform and, as the processors have not undergone any microarchitectural changes, they are an update of the^th generation parts but with the stack of products arranged a little differently. For those who stay the course, Coffee Lake was already a rehash of Kaby Lake, which was an update for Skylake. So we are on Skylake Refresh Refresh Refresh.

Intel has promised that its 10nm manufacturing process would increase until 2019 and has already announced its intention to introduce Ice Lake for servers to 10nm by 2020, after a further 14nm pass with Cooper Lake in 2019. On the consumer side, the status remains in limbo – hopefully, the next generation of consumer components will be an appropriate update of the microarchitecture, regardless of the process node.

Hardware and software security patches

What makes it a bit different is eight-core products. In order to manufacture them, Intel had to create new matrix masks for the production line because their previous masks only had six cores (and before that, four cores). This would theoretically give Intel a chance to implement some of the hardware solutions for Spectrum / Meltdown. At the time of writing, we were not informed that this was the case, perhaps because the basic design was essentially a variant of Skylake in a new processor. We will have to wait for a new microarchitecture to come to the fore to see the changes.

However, Intel has optimized its manufacturing process. Many have said that the 14-nm Intel family of technologies is the most cost-effective manufacturing node in the company's history. It should be noted that Intel has now avoided names such as "14+" and "14+" in the official marketing, instead choosing to call it "14-nm family" and highlighting the product optimization. The results of these optimizations, whatever they are, are generally increases of frequency and performances with iso or lower power, generally to the detriment of a reduced zone. This would mean fewer dies per slice, naturally increasing the cost of the product.

The processors advertised so far are 8-core and 6-core processors, with leaks suggesting that Intel will also produce 4-core and 2-core processors at a later date. There is no word if some of these parts share representations of the matrix (for example, if a 6-core is a natural 6-core, or an 8-core cut, or both will occur ). Making new matrix masks is expensive, but it can be advantageous depending on the number of processors for each segment. However, Intel will want to do something with these 8-core matrices that might not make it a tactic used in its business components.

More coffee, less caffeine: HyperThreading and L3 cache

All this aside, it seems that Intel is also abandoning HyperThreading on most of its processors. The only Core processors to get HyperThreading will be the Core i9 components, and maybe the Pentium as well. This helps in part to make the stack of products more linear, and cheaper chips do not work on the more expensive feet (for example, a quad-core with HyperThreading could outperform a 6-core without). The other angle, I suppose, is one of the side attacks that can occur when HyperThreading is in action. By disabling HyperThreading on volume production chips, this security problem is no longer present. This also ensures that not all threads on this chip compete for resources by heart.

One of the most interesting dissections of the new 9^th The Generation product is in the L3 cache by heart for the different models. In previous generations, the Core i7 parts had 2 MB of L3 cache per core, while the Core i5 had 1.5 MB of L3 cache per core, and Core i3 was split between some with 2 MB and the others with 1.5 MB. This time, Intel puts only the full cache on the most powerful Core i9 components and reduces the Core i7 to 1.5 MB N3 per core. This will have a slight effect on performance, which, when we have the processors, will be an interesting metric to test.

I've had an 8-Core for years!

Depending on where you draw the line for "mainstream" processors, we have technically installed Intel 8-core processors on high-end office space for several years. The Core i7-5960X was released in August 2014 and includes eight Haswell cores on the HEDT platform, with four-channel DDR4-2133 memory and 44 140-bit PCIe channels. At the time, on Intel's 22 nm process, the size of the array was about 355.52 mm2.

At the time of the launch of Intel's first Coffee Lake processors, the design of the i7-8700K's 6 + 2 dies was about 151 mm2, an increase of about 26 mm2 from the previous year. 4 + 2 design of the i7-7700K (~ 125 mm2). This also occurred during a jump from the official Intel 14+ to 14+ manufacturing nodes, which, due to a relaxed flip-flop, has still magnified the situation.

But if we take 26 mm2 to add a pair of cores to the size of the matrix, we can predict that the 8 + 2 version of the Core i9-9900K should be about ~ 177 mm2, a 17% matrix bigger. Cut. With 177 mm2, this is half the size of the Core i7-5960X, but with only half of the memory controllers and PCIe tracks. Even with that, it's a significant decrease.

This is a model of what a 9900K might look like

This is where I point out that for Intel to maintain its profit margins on its highest parts, this 17% increase in processing area could directly translate into a 17% price increase. At the time of writing, Intel had not announced a price, but with a plateau of $ 359 for the latest generation of 8700K 6 + 2 chips, a 17% increase puts it at around $ 420. Nothing more than that, and Intel increases its margins or manages the number of chips actually at the required frequency. But as shown by the Amazon link above, 582 USD represents a larger increase.

Integrated graphics

One theme on which Intel has not focused much for generations (since Broadwell, in fact) is integrated graphics. All the chips announced for the 9th^th generation family will always have the same GT2 configuration as the 8^th generation, including new Core i9 parts. Officially, they fall under the 8 + 2 designation. Intel is still convinced that having an integrated graphical form on these high-end, overclockable processors, is still an added value for the platform. The only drawback is the performance, and it will not win any prize soon.

The graphics will always bear the UHD Graphics 630 label and use the same drivers as the driver 8.^th family gen.

Motherboards and Z390 chipset

One of the worst secrets kept this year is Intel's Z390 chipset. If you believe everything that motherboard manufacturers tell me, most of them are ready for this version for several months. That's why we're going to see about 55 new motherboard models coming on the market in the next few weeks.

The Z390 chipset is an update of Z370. Both types of motherboards support 8000 and 9000 series processors (the Z370 will require a BIOS update). The updates are similar to the updates seen with B360: native USB 3.1 ports at 10 Gbps, and built-in Wi-Fi on the chipset.

Built-in Wi-Fi uses CNVi, which allows the motherboard manufacturer to use one of Intel's three complementary RF modules as a PHY, instead of using a potentially more expensive MAC + PHY combo. another manufacturer (Broadcom, for example). It has been said that the cost of implementing a CRF adds about $ 15 to the retail price of the panel. It is therefore likely that some vendors are experimenting with mid-priced models with and without a Wi-Fi network using this method.

One of the most awesome motherboards with TB3 – the ASRock Z390 Phantom Gaming-ITX / ac card

For USB 3.1 Gen 2 ports, type A ports are natively supported and motherboard manufacturers will need to use re-driver chips to support type-C reversibility. These have a cost extra, as one might expect. It will be interesting to see how manufacturers associate Gen 2, Gen 1 and USB 2.0 ports on the rear panels. They now have the choice. I suspect this will reduce the integrity of the traces on the motherboard.

For the Z390 chipset and motherboards, we will have our usual overview of each map displayed today, covering every model that manufacturers would talk to us about. Interestingly, there will be a mini-ITX with Thunderbolt 3 and a card with a PLX chip! There are also motherboards with Realtek's 2.5G Ethernet Controller, which consumes 1.6W – now, if only we also had consumer switches.

Chronology: full launch on the 19th^th October, with criticism

From there, Intel will have processors on tablets on October 19th^th in most (if not all) major markets. The embargo on the revision of the three K processors is also valid for the same date, at 9 am, Eastern time. I will run my new scripts on as many processors as possible. Let me know what you want to see in the comments below.