Skylake-X and Kaby Lake-X
As stated before, Intel calls their new CPUs Skylake-X and Kaby Lake-X generations. The latter is quite remarkable, because Intel’s high-end platform used to be a generation behind compared with the mainstream platform up until this point. The company seems to make their most recent architecture available for the high-end market as well, although there is something to be said about this.
The Skylake-generation processors are known in laptops and standard desktops as the “6th generation Core-processors” and have been available since August 2015. This was the last generation that was released according to Intel’s tick-tock model, which means that the company releases a new or strongly improved architecture, or alternatively introduce a new fabrication process. The Skylake-predecessor Broadwell was a so-called tick, which made the previously used Haswell architecture available on a 14nm- instead of a 22nm fabrication process. Skylake is Intel’s second generation 12nm CPU and therefore is a tock, in other words a new architecture based on a mature fabrication process. Kaby Lake should have been a tick- a change to a new fabrication process. However, Intel does not use this model anymore. This is in part due to the immense challenges that come with the development of the 10nm fabrication process. This means that Kaby Lake is more or less an optimized version of Skylake at 14nm. In laptops and standard desktops we know Kaby Lake as the “7th generation Core processors”, that have been available since last year. With Kaby Lake-X this generation becomes available for the high-end desktop platform for the first time.
While Intel used to introduce new CPU-generations as server- and high-end-desktop processors and would introduce models used for less demanding applications step by step, this is the other way around now. Because of this, we had to wait for this moment for two years- the moment where the Skylake architecture makes the change from laptops and desktops to high-end desktops and servers. However, Skylake-X is not simply a model based on the Skylake CPUs with more cores. The cores within the Skylake-X chips offer functionality that the existing Skylake CPUs do not offer, such as AVX-512 instructions. They also have a different cache-construction. We will cover both of these in more detail further on in this article. Something that Skylake-X will have in common with the mainstream chips, we know from previous tests, is an approximately 8% higher IPC than predecessor Broadwell. This is the first advantage of the new chips.
Eventually Intel wants to release Skylake-X CPUs with up to 18 cores, but for now the top model is “limited” to 10 cores. Aside from the amount of computing cores, the Skylake-X chips have a quad-channel DDR4 memory controller, which officially supports up to DDR4-2667 and a PCI-Express 3.0 controller that (depending on the CPU model) offers 28 or even 44 lanes. The maximum L3-cache is 16.5 MB. This might seem small compared with the existing Broadwell-E chips, but this is a consequence of the fact that every single core has more L2-cache of their own with Skylake-X; more on this later.
Aside from that there are also two Socket 2066 processor from the Kaby Lake-X generation. Kaby Lake is, as stated before, Intel’s latest processor generation and with that slightly more modern than Skylake, although we know from our IPC-tests that the improvements made in Kaby Lake are marginal. The Kaby Lake-X CPUs are based on the exact same chips as the existing Socket 1151 Kaby Lake CPUs and offer a maximum of 4 cores, a dual-channel DDR4 memory controller and a PCI-Express controller with a maximum of 16 lanes. That said, Kaby Lake-X is placed on the same X299 motherboards as Skylake-X, which in practice means that a lot of the functionality of the motherboard, including half the memory slots, will not function.
In the table below you can find the most important high-level characteristics of Skylake-X and Kaby Lake-X next to Intel’s existing Broadwell-E high-end desktop platform.
|Amount of cores||6, 8, 10, 12, 14, 16 and 18||4||6, 8 and 10|
|L2-cache||1 MB / core||256 kB / core||256 kB / core|
|L3-cache||1,375 MB / core
(max. 16,5 MB)
|8 MB||2,5 MB / core
(max. 25 MB)
|PCIe-lanes||44 / 28||16||40 / 28|
|Multi-GPU||2x 16 / 4x 8||1x 16 / 2x 8||2x 16 / 4x 8|
|Memory||Quad-channel DDR4-2666||Dual-channel DDR4-2666||Quad-channel DDR4-2400|
|TDP||140 watt||112 watt||140 watt|
|Socket||LGA 2066||LGA 2066||LGA 2011-3|
With Kaby Lake-X, Intel releases a variant of the current “mid-range” desktop CPUs for the new high-end platform. As we will show on the next page, the Kaby Lake-X processors have a higher clock frequency of 100 MHz. and a higher TDP than their Socket 1151 counterparts, meaning that the performance will be slightly better.
The question is why Intel would release the Kaby Lake-X CPUs in the first place. The marginal performance difference, that you can easily match on Socket 1151 with a small overclock, will not convince anyone. Aside from that, the Intel X299 motherboards that you have to use with your Kaby Lake-X processors are a lot more expensive than Socket 1151 motherboards with Z270 chipset. Add that to the fact that when you place a Kaby Lake-X processor on a brand new X299 motherboard, a lot of the new functionality will simply not work. In contrast to the Skylake-X CPUs, the Kaby Lake-X chips have a dual- instead of a quad-channel memory controller, meaning that half of the present memory slots will not work. Furthermore the PCI-Express controller of the Kaby Lake-X CPUs offers only 16 lanes instead of 28 or even 44 for the Skylake-X CPUs, meaning that several PCI-Express expansion slots, M.2-slots and / or other functions will not work either. As we will see further down in this article, the X299 chipset also does not offer any functionality that is missing from Z270, meaning that the chipset does not have any added value either.
Add to that the fact that the integrated graphics card is turned off with Kaby Lake-X; Socket 2066 does not offer the possibility to directly connect a monitor. Although in this price range it is obvious that most of the users will combine their CPU with a separate graphics card, turning off the graphics card also means that the useful QuickSync video-encoder is no longer available on the Kaby Lake-X CPUs. This means that the new CPUs are missing some of the functionality that the existing ones do have! Besides that, you can always continue working using the integrated GPU if your graphics card breaks down or is having some issues; this is not the case with Socket 2066.
A clear answer to the question “why would anyone want to purchase a Kaby Lake-X CPU?” was missing from our official Intel briefings. Our colleagues of Anandtech explicitly asked Intel this question and received the following answer:
“The two Kaby Lake quad core processors were introduced for broader expandability of the platform. The Intel® Core i5-7640X and Intel® Core i7-7740X enable customers to invest in a more powerful enthusiast platform that provides more headroom and scalability up the X-Series stack when they are ready. These two processors will start with a slightly higher base frequency and provide all of the ingredients, like higher memory speed and larger socket, for better overclocking performance compared to their mainstream counterparts. That said, our Intel Core i7-7700K and the S-line are still great options and by introducing the X-series we’re giving a wide variety of consumers the benefit of choice.”
In other words: the advantage of Kaby Lake-X is that you can use a relatively cheap processor in a system based on a high-end X299 motherboard, where you can upgrade to a faster, more expensive Skylake-X processor in the future.
While this argument sounds plausible, at the same time this is not advice that we would offer as Hardware.Info. It is pretty senseless to combine a processor with an expensive motherboard, while you can reach the same performance level with a platform that is a lot cheaper, while also being able to use QuickSync. If you do want to upgrade from four cores to more, it is probably a lot cheaper and smarter to upgrade the CPU, motherboard and memory and sell your existing set second-hand. Doing so, you save a lot with your initial purchase, you will not be fiddling around with a platform that only works partly and when you do want to upgrade you still have all the freedom to choose the motherboard that suits your desires at that time.
Furthermore there are persistent rumours that Intel is already releasing their 8th generation Core-processors (codename Coffee Lake) in August (in two months’ time!), which will bring CPUs of up to 6 cores to the mid-range Socket 1151 platform. Although these are rumours and their prices are not yet known, it is not unlikely that thanks to Coffee Lake you will be able to purchase a Socket 1151 platform with a lot more performance at the same price of a quad-core Kaby Lake-X CPU on a Socket 2066 platform in two months’ time.
This means that, before we look at any benchmarks, the Kaby Lake-X chips are the first Intel CPUs of which we have to conclude that the purchase of one of these is questionable.